Description , Evaluation and Stance : Exploring the Forms and Functions of Speech Descriptors in Early Modern English

This article explores the form and function of “speech descriptors”. These features describe in some way what a speech event that is being represented was like, such as loudly in She said loudly that she was unhappy. Based on data from Early Modern English witness depositions, the study reveals that a number of aspects of represented speech can be described by speech descriptors and that such descriptors come in a number of linguistic forms. Prepositional phrases, adverbs, and adjective constructions are the most common forms, and speech descriptors can signal aspects of evaluation (e.g., angrily), clarification (e.g., meaning), and formulation hedging (e.g., or words to that effect), among other features. The article underscores the importance of further attention to these descriptors in order to gain a full understanding of the dynamics of speech representation in historical periods.


Introduction 1
A substantial body of linguistic research has been devoted to the forms and functions of speech representation, that is, how language users represent a speech event in speech or writing.Two aspects have received particular attention: the reporting expression that is used to introduce the represented speech (say, report, claim, etc.) and the different modes of representation (direct speech, indirect speech, free indirect speech, etc.).In (1), for instance, the reporting expression is "sayed theise words" (in bold) and the speech is represented in direct speech (in single underlining).
(1) at the last the sayd lawrence vearye spighte fullye and in malliceouse and manner sayed theise 1 I am grateful to James W. Hartman for commenting on an earlier version of this article, to an anonymous reviewer and the editors for helpful suggestions, and to Alex D'Arcy, Colette Moore, and Mick Short for answering my queries about aspects of speech representation.Naturally, any errors are entirely my own.
words or the like in effecte to the sayd Peter Portor viz Thowe, Thie Jorneyman did Jape thie wiffe) take witnes yf thow willt for I will prove yt by thie owne words, (ETED: London 1590-1593: F_1LD_London_004) However, there are additional linguistic cues that language users employ to describe aspects of a previous speech event.In the passages with double underlining in (1), the witness suggests that the represented words were delivered in a particular way ("vearye spightefullye and in malliceouse and manner") and that the speech event, although reported as direct speech, may not be cited verbatim; instead, the words may simply convey a meaning similar to that of the actual words used ("or the like in effecte").These types of markers that describe the speech event in some way appear not to have been discussed much in previous research, and we thus know little about their characteristics, frequency and function.The scant attention to these "speech descriptors" is the more surprising since they provide clues to crucial questions about the represented speech, such as how the reporters view the speech, what characteristics the speech event had beyond what is signaled by the actual representation and the reporting expression, and how faithful a given representation is to the original speech event.
This article is an exploratory study of speech descriptors based on witness depositions in Early Modern English, drawn from An Electronic Text Edition of Depositions, 1560-1760, or ETED (Kytö, Grund and Walker 2011).This material is very useful for an initial exploration as witness depositions consist primarily of various levels of represented speech (see Section 3), and there are thus many opportunities for speech descriptors to occur.My analysis focuses on three broad areas: 1) an outline of the different linguistic realizations of speech descriptors (e.g., adverbs, prepositional phrases); 2) an exploration of their semanticfunctional properties (such as marking manner or intention, or formulation "hedging", as indicated by the two instances in example (1)); 3) a first attempt at charting their pragmatic, contextual uses in particular types of legal cases and with particular speech representation modes.I will show, among other things, that speech descriptors often function as stance markers that enable witnesses (or the recorders of their testimonies) to signal their attitude toward or evaluation of the represented speech event.

Background
There is a long tradition of research into speech representation in English linguistics, focusing on questions of form, function, frameworks of description and use in different genres and contexts (for an overview, see Walker and Grund forthcoming).In English historical linguistics, the topic has received increasing attention over the past few years.This recent work has started to patch together a picture of the varied nature of speech representation in historical periods and the important ways in which it both overlaps with and differs from speech representation in modern contexts (for an overview, see Moore 2011).We find that language users in historical periods employed different modes of representation (esp.direct and indirect speech) for a variety of pragmatic purposes in a range of genres, including newspapers and news reports, witness depositions, trial proceedings, medieval tracts and fiction (e.g., Moore 2002Moore , 2006;;Camiciotti 2007;McIntyre and Walker 2011;Jucker and Berger 2014;Walker and Grund forthcoming).Significant attention has also been paid to issues of how faithful direct speech representations are to actual speech in historical periods, and how various genres (e.g., drama comedy, trials and fiction) can be used to capture the spoken language of the past (see esp.Culpeper and Kytö 2010, and references therein).
Despite this intense interest in speech representation from a variety of perspectives, very little scholarship has been devoted to features that I have labeled "speech descriptors", that is, features other than the reporting expression and the represented speech that describe in some way what the original speech event was like.Thompson (1996: 521-523) notes that there are "many ways" for language users to indicate their attitude to the speech they represent, but he only discusses the reporting expression itself.Clark and Gerrig (1990: 775) indicate that language users can opt to signal (among other things) voice quality, pitch, emotional state and dialect of the original speaker.Most of their illustrations come from spoken examples where the reporter would use a particular pitch or voice quality to "depict" the original speaker's language (cf.also Lyons 1977: 63-67;Quirk et al. 1985Quirk et al. : 1024, note a;, note a;Brown 1990: ch. 6).Language users are more constrained in what can be "depicted" in written language.Clark and Gerrig (1990: 782-783) give examples of how orthography can be used to indicate a lisp or how capitals can signal shouting.Aspects of the original speech event that cannot be "depicted" would be described, which they do not consider in their study.
More closely related to my study are the discussions in Brown (1990), Oostdijk (1990) andde Haan (1996).In exploring present-day fiction, Oostdijk (1990: 239) and de Haan (1996: 36-37) note that adverbials can cooccur with the reporting expressions.While de Haan (1996) provides no further information on the nature of these adverbials, Oostdijk (1990: 239) notes a number of characteristics: they can appear as adverb phrases, prepositional phrases and finite and non-finite clauses, and their functions include indicating manner (e.g., angrily), circumstance (e.g., cupping her face in her hands), time (e.g., then) and person to whom the speech is directed (e.g., to Barrow). Brown (1990: ch. 6), finally, focuses on phonetic features that "contribute to the expression of attitude by a speaker " (1990: 112).In her discussion, she reviews examples from literary works where reporting expressions and, most importantly for this study, adverbs and prepositional phrases can be used to describe the same kind of effect as the phonetic features perform.The descriptive features she includes are grouped in accordance with their supposed phonetic description, such as desperately (pitch span), gloomily (voice range), slowly (tempo), in a lower tone (loudness) and stiffly (articulatory setting) (Brown 1990: 118, 121, 127, 129).Although there is thus some research on speech descriptors, there appears to be no systematic, comprehensive treatment of them as a category or group.
It is also useful to place the study of speech descriptors in the context of research on "stance".The exact definition of stance varies (as does the term used to denote the concept; see, e.g., Thompson and Hunston 2000: 13-26;Englebretson 2007: 15-20), but it is usually taken to denote the expression of "personal feelings, attitudes, value judgments, or assessments" (Biber et al. 1999: 966).Stance has received increasing attention in scholarship on the history of English.Studies have demonstrated that a broad range of features perform stance functions and that language users make strategic use of them for a variety of sociopragmatic purposes (e.g., Fitzmaurice 2003;Biber 2004;Levorato 2009;Busse 2010;Gray, Biber and Hiltunen 2011;Chaemsaithong 2012;Grund 2014).More specifically, stance has been considered in relation to speech representation (primarily in research on present-day contexts).Studies have frequently remarked on how the reporter of a speech event can signal attitude through the selection of particular reporting expressions (e.g., Thompson and Ye 1991;Thompson 1996;Bevitori 2006;cf. Semino and Short 2004: 96).Recent research has also argued that different speech representation modes (esp.direct speech) can be used as stance resources (Holt and Clift 2007).My study explores an angle on stance in the history of English and on stance and speech representation more generally that has not been studied in detail before.Many of the speech descriptors have stance functions, allowing the deponents and/or scribes to signal their attitudes or assessment of the represented speech.

Material and method
My study is based on ETED, which contains 905 witness depositions from England in the period 1560-1760.A deposition is the written version of an originally oral testimony delivered in connection with a legal case and recorded by a scribe (for a discussion of the genre, see Kytö, Grund and Walker 2011).The extract from a deposition in (2) illustrates how the genre is characterized by the representation of previous speech events.
( In (2), the deponent (John Daw) provides spoken testimony ("saith that…"), which has been recorded by the scribe.In the testimony, Daw represents a previous statement allegedly made by a Francis Hobart.
Some depositions contain only the deponent's recorded speech, but it is common for depositions to present two levels as in (2); more levels are fairly rare (Grund and Walker 2011;Walker and Grund forthcoming).ETED consists of thirty deposition collections encompassing ca.267,000 words.The collections hail from different geographical regions in England and cover depositions by men and women engaged in a range of court cases.ETED thus enables a study of the connection between language and a number of extralinguistic features.In this exploratory investigation, I will primarily consider court type and case type, as there appears to be a close correlation between these two factors and the use of speech descriptors.
ETED covers two main types of courts: criminal and ecclesiastical.Criminal courts, which are represented by seventeen collections in ETED, handled cases of theft, robbery, poaching, assault, rape, murder, treason, infanticide and threats (among others).Ecclesiastical courts, on the other hand, had within their purview cases such as defamation, contested wills, broken marriage promises, adultery, illegitimate births and unpaid tithes (see Walker 2011); there are thirteen ecclesiastical collections in ETED.
There is one aspect of the ETED depositions that requires special attention: the scribe.The scribes recorded in writing what the witnesses said, and they often made modifications to the witnesses' original language for a variety of reasons: the exact formulation may not have been important or they may have wanted to make sure that the formulation conformed to legal requirements.The language that we find in the depositions is thus best seen as a mixture of the witnesses' original language and that of the scribes, and the depositions as a whole can be seen as co-constructed by the two (see Grund and Walker 2011;Grund 2011a; Walker and Grund forthcoming; for work on scribal influence in a different trial setting, see Doty 2007;Hiltunen and Peikola 2007).In most contexts, we cannot tell with any degree of certainty whether the speech descriptor was provided by the deponent who reports the speech or by the scribe who takes down the testimony, and the distinction is not crucial for this article.But I will touch on the question of scribal vs. witness usage in a few contexts, especially in terms of the formulation or construction of the speech descriptor used (see esp.Sections 4.2.1, 4.2.3).
Although ETED is searchable electronically, the topic of speech descriptors requires a more manual approach.As noted in Section 2, there appear to be no previous in-depth studies of these features, and therefore it is not possible to search for particular lexical forms.Rather the approach adopted here is "text-driven" (Bednarek 2006: 638-639), where instances are identified manually in context.In other words, the identification process involves function-to-form mapping (Taavitsainen and Jucker 2010: 13-14), where the function or meaning of a construction is the determinant.At the same time, once the different types of speech descriptors had been identified, I ran searches to ensure that no instances had been missed.
My identification was thus guided by semantic and functional questions, such as "what features describe the nature of the speech event and/or convey the speaker's evaluation or interpretation of the speech event that is being represented?".In (3), for example, the deponent describes the tone of voice ("vearye lowde").Example (4), on the other hand, illustrates the deponent's interpretation of the character of the speech: the words were seen as tantamount to slander, and the nature of the words was "vile". 23) and divers others woords w ch he now remembreth not weare then vttered by the sayd wilson vearye lowde and openlye (ETED: London 1590-1593: F_1LD_London_002) (4) the saied Phillida spake such sclanderous vile reprochfull worde against the saied Okeshott that the saied Okeshott called this depo t (who was at worck nere the place where they were talking) to be witnes to her worde (ETED: Winchester 1600-1602: F_2SD_Winchester_027) In this exploratory study, I have been inclusive, erring on the side of incorporating features that may be omitted as the concept or categorization is refined in the future.At the same time, certain features were not included.Descriptions of the physical context in which the speech took place were not included as they do not describe the speech per se.I thus excluded examples, such as those underlined in (5), which indicate the time and place of the speech event and the person to whom the speech was directed.Such markers were considered by Oostdijk (1990: 239), but her study focused more generally on adverbials occurring in connection with speech representation.

Findings 4.1 Overall results and linguistic realizations
Table 1 gives an overview of the major categories of speech descriptors in ETED and their occurrence in criminal and ecclesiastical depositions.
Since there is no already-existing framework of description, the semantic-functional categories used here represent my first attempt at classifying the meaning and function of speech descriptors (as found in the ETED depositions) (see 4.2).The distribution shown in Table 1 is statistically significant (d.f.= 4; χ 2 =180.9855;p <0.01).However, these overall patterns hide a great deal of variation among the thirty deposition collections.In the thirteen ecclesiastical collections, the range in frequency per 10,000 words is 77.3 (lowest: 10.5; highest: 87.8), with a fairly high standard deviation of 20.4.The situation is similar for the seventeen criminal collections: the range is 81 (lowest: 3.3; highest: 84.3), again with a fairly high standard deviation of 20.6.
There may be several reasons for this kind of variation.Perhaps the most important consideration is the overall frequency of speech representation in the different collections: if the collections vary in terms of how frequent speech representation is, then there are potentially more or fewer opportunities for speech descriptors to occur.Quantifying all instances of speech representation in the collections is, however, a very time consuming undertaking (see Walker and Grund forthcoming), and I have therefore not attempted it here.But just a cursory exploration reveals a great deal of variation among the collections in this regard.This variation appears to be tied in with the type of cases included in the collections.For example, cases of defamation, which are almost exclusively found in ecclesiastical collections in ETED, naturally revolve around speech and the evaluation of that speech.Depositions in such cases thus often contain multiple instances of speech descriptors (see Section 4.2.1).Cases of theft or robbery, on the other hand, which were addressed in criminal courts, tend to involve less speech.Providing counts per 10,000 words may thus be quite a blunt tool, while recording frequency in relation to the number of speech representation instances may give a more fine-grained picture.There may also be other factors behind the variation, such as differences over time, in region or in scribal preference, which I have not charted systematically in this study.
In terms of their linguistic realization, we find a range of linguistic forms for speech descriptors in ETED, as seen in Table 2.The top three categories are the same in the two types of collections: prepositional phrases, adjectives or adjective phrases, 3 and adverb phrases.However, the order of frequency differs markedly: in criminal depositions, prepositional phrases are favored over adjectival and adverbal forms (the latter two being close in frequency).In ecclesiastical collections, by contrast, adjectives are the most common, closely followed by prepositional phrases, and adverb phrases are a more distant third.As we shall see, there are several reasons for this distribution.Prepositional phrases occur in two different, primary contexts in the two collections.The higher percentage in criminal depositions is because of the very frequent occurrence of upon/on…oath in some criminal deposition collections, as in (11), while a similar annotation is rare in ecclesiastical depositions (see Section 4.2.4).This usage (Emphasis) accounts for 145 out of 167 (or 87%) of the prepositional phrases in criminal depositions.In ecclesiastical depositions, the majority (113 out of 164, or 69%) express some type of Evaluation, as in ( 12).Adjectives are most commonly found in the Evaluation category (219 out 282 total examples, or 78%), almost always expressing some kind of negative aspect of the speech (e.g., angry, cruel, evil, opprobrious, scolding, slanderous, unseemly, vile).They are by and large restricted to two types of speech representation modes, which Semino and Short (2004: 44, 52) term "narrator's representation of voice" (NV) and "narrator's representation of speech act" (NRSA).NVs involve a minimal representation of speech, simply indicating that speech occurred but providing no specifics about the content or form.NRSAs, on the other hand, provide a bit more detail, signaling what speech act was involved (see Walker and Grund forthcoming).In (13), we learn little about the content of the exchange between Joan Prentice and Mary Humfrey, but we see the deponent's evaluation of the kind of exchange they had ("high wordes" and "angrye talke").Walker and Grund (forthcoming) have suggested that instances of NV with evaluative markers allow the person representing a previous speech event to highlight particular, evaluative aspects of the speech as especially important.In ( 14), exactly what Applin said was not essential, but the fact that it was scandalous and abusive was.Stressing the nature of the words rather than the words themselves may be the reason why adjectival speech descriptors are so common in depositions from ecclesiastical courts, especially those that deal with defamation.Perhaps prompted by specific questions, deponents frequently comment on the nature of speech events that are at issue in the legal case (see 4.2.1).By providing an evaluation rather than the words themselves, the deponent also removes the opportunity for the audience (i.e., the court officials) to interpret the speech, as the deponent has already provided that interpretation (Walker and Grund forthcoming).Criminal and ecclesiastical collections differ slightly in their usage of adjectival constructions.While ecclesiastical depositions have the edge on criminal depositions in using adjectives for expressing Evaluation (80% vs. 72%), the relationship is the reverse for adjectives in the Frequency/Quantity category (19% for ecclesiastical depositions and 27% for criminal depositions). 4The Frequency/Quantity adjectives (in particular quantifiers such as many, much, several, as in ( 15)) are usually used in NRSA or NV representations and are found together with Evaluation adjectives.They often stress that not only were negative words used, but there were many such words or much language of that kind (see 4.2.5).
( As in the case of adjectival speech descriptors, the majority of adverb phrases occur within the Evaluation category (119 out of 183, or 65%), more so in the ecclesiastical depositions (72%) than in the criminal depositions (59%).Similarly to the adjectival speech descriptors, they mostly express some kind of negative aspect of the speech (e.g., angrily, disrespectfully, maliciously, outrageously, slanderously, uncivilly, unnaturally).The remaining express Frequency/Quantity (63 out of 183, or 34%), especially in criminal depositions, and Formulation Hedging (1, or 1%).Unlike the adjectival speech descriptors, they are more evenly distributed among speech representation modes, occurring not only in NV and NRSA representations but also frequently with direct and indirect speech, as in ( 16) and ( 17 Or constructions exclusively belong to the category of Formulation Hedging.They indicate that the formulation should be understood as inexact or as equivalent to the previous speech event in "effect" but not necessarily in wording, as in ( 20) and ( 21).What is coordinated by or is usually a prepositional phrase (as in ( 20)) or a noun phrase (as in ( 21)), and they are found in a number of related forms, such as or to the like/that effect, or words to the like/that/this/tending to this effect and or such words/the like words/suchlike/the like in effect.
(20) speaking to the sayd <f.77v> Curtis she sayd, Curtis why doe you stand a talking with such an offscum as he (meaning meaning m ~ Skelton the sayd m ~s Skelton's Husband, who was then also there.) that never wishd me well in his life time, or to that Effect (ETED: London 1681-1682: F_3LD_London_002) (21) he did say vnto him that the blacke coate had ouer much and woalde he helpe them to more.& or worde to that effect.
(ETED: Durham 1628-1638: F_2ND_Durham_005) Sentential comments, which are all found in the Evaluation category, differ from the other linguistic realizations in that they are usually separated from the speech representation that they relate to.In ( 22), for instance, the deponent evaluates what he thinks the dialogue that he has represented previously might mean and what the speaker intended.These sentential comments may contain several evaluative words (such as "defamatory" in ( 22)).( 22 Sentential comments are exclusively found in ecclesiastical depositions and (with one possible exception) only in cases that appear to treat defamation.They may stem from specific questioning of the deponent asking him/her to interpret and evaluate the (alleged) defamatory words.Indeed, in some collections, such as Oxford 1609-1615 (see, e.g., F_2SD_Oxford_026), sentential comments are presented as answers for specific questions, where the formulation may have been presented to the deponent and he/she simply agreed (Grund and Walker 2011: 33).The way the words would have been perceived was central for the case because the court was to establish (among other things) if the words would have had the effect of defaming the target of the words (Tarver 1995: 113, 116 Overall, it is interesting to note that, while many of the linguistic realizations are used adverbially, non-adverbials are well represented in adjectival, sentential or clausal constructions.Speech descriptors are clearly a more varied phenomenon than might be gleaned from the (admittedly sparse) previous research on the topic (esp.Oostdijk 1990; cf. Brown 1990: ch. 6).

Semantic-functional categories
As I noted in 4.1, there is currently no comprehensive system of classification that could be adapted for the description of the meaning and function of the speech descriptors in ETED.Brown (1990: ch. 6) provides categories that match the supposed phonetic character of a particular speech descriptor, such as gloomily (voice range) and slowly (tempo).However, this system is not adequate for the ETED data, which shows that these types of markers can express a number of sentiments that go far beyond phonetic features.My description below is thus a first attempt at constructing categories that fit the full range of features found in the the ETED depositions.No doubt, these categories will require refinement in the future as different kinds of materials are considered, and one category in particular, Evaluation, contains a number of subcategories some of which may be possible to separate out as major uses in light of further data.My classification here is based on the meaning of the items as well as their use in context (hence a consideration of both semantics and function).Not all the fine-grained features of these categories can be treated in this article; I will concentrate on outlining the major parameters of the categories.

Evaluation
Evaluation markers are the most frequent speech descriptors and the most varied in terms of linguistic realizations: they occur as adjectives, adverb phrases, prepositional phrases, sentential comments and, rarely, participle constructions.On the most general level, these markers express the deponent's (and/or the scribe's) assessment of the manner in which a statement was made.In other words, they signal a broad range of stance notions (see Section 2).In some contexts, they reflect the stance of the person representing the speech (the deponent and/or scribe); in other contexts, the stance appears to be attributed to the original speaker.
There are a number of different subcategories of Evaluation markers.I will concentrate on the most prominent in this exploratory survey.The most common among these subcategories represent a number of related and to some extent overlapping notions: the perceived mental state (25) or intent (26) of the original speaker, the meaning of the words spoken ( 27), the character of the words ( 28) or an interpretation of the speech act involved ( 29). ( 25 In ( 25), the deponent indicates that the statement was made in anger, presumably based on an assessment of behavior, facial expressions and the like.The deponent thus also gives an indication of the perceived state of mind of the original speaker, which was a significant consideration in defamation cases (see below).Although ( 26) superficially also deals with the manner in which something was said, the deponent in fact provides an interpretation of the intention of the speaker or attributes an intention to the speaker in claiming that the speech was made very spitefully and maliciously; that is, the original speaker allegedly had in mind to insult and to malign the person to whom the words were targeted.Example ( 27) involves a similar level of interpretation, where the deponent paraphrases the meaning of what "Christopher" said in a sentential comment.Examples ( 28) and ( 29) also center on the meaning of the words or their effect.In (28), the deponent evaluates the character or status of the words used by the defendant.In (29), finally, the adjective defamatory reveals the deponent's conclusion that the words spoken amount to the speech act of defamation.
Although most of these subcategories occur in a range of case types, they are most common in depositions pertaining to defamation cases.They clearly reveal different levels of assessment by the deponents, and reveal their stances toward the previous speech event.From the modern perspective of legal practice, these kinds of subjective evaluations would seem out of place (cf. Stygall 1994: 138).However, especially in defamation cases, interpretations and evaluations were crucial.Gowing (1996: 43) has pointed out that, in ecclesiastical court proceedings, " [w]itnesses were asked by the court who they thought 'had the right'; the question was often not what had happened, but whose interpretation of it deserved to be accepted".Furthermore, in order for the alleged words to qualify as defamation, they had to conform to certain legal standards.For example, words that resulted from a heated argument or anger were not seen as an act of defamation (Tarver 1995: 113).A speech descriptor such as in an angry manner in (25) may thus act in favor of the defendant, recusing the speaker from intentional defamation.Malicious intent, on the other hand, was important for establishing that defamation had occurred (Tarver 1995: 113).Example (26) would thus have acted against the defendant.It is of course no accident that the deponents described and evaluated the speech in terms of the legal implications of the represented speech.They were likely asked to address directly their interpretation of the words spoken along the lines of the legal definition of defamation (as suggested by Gowing 1996).Whether the exact evaluative words are those of the deponent or the scribe is difficult (if not impossible) to determine.However, considering the frequent occurrence of the same vocabulary (esp., malicious(ly), spiteful(ly), envious(ly), defamatory) across collections and across time, it is likely that it reflects the usage of the scribe, although the deponent may of course have expressed the same sentiment, albeit with a different wording.The language of the court more generally may also have influenced the deponents.In ecclesiastical depositions, in particular, deponents would respond to a set of carefully formulated questions.There is evidence that deponents often simply recycled the language of the question (or, again, the scribe adopted that language whatever the deponent expressed) (see Churches 1996: 220-223;Gowing 1996: 44;Grund and Walker 2011: 50-51).
Irrespective of whether the actual wording is the deponents' or the scribes', Evaluation markers also reflect stancetaking in favor of a particular side in the litigation: they show that the deponents agreed or disagreed that the words spoken were intended to malign, were defamatory or had content that was intended to slander the plaintiff.In other words, by evaluating the words in a particular way, the deponent also aligns or disaligns with one or the other side in the case.As Du Bois (2007: 163) argues, it is not only the evaluation of an issue (or "stance object") that is important, but the positioning vis-à-vis another person is also an essential part of stancetaking.
Speech descriptors can also express an evaluation of the degree to which a speech act was performed.In (30), for instance, the deponent's denial is emphasized by utterly.In this case, utterly (or a similar expression) may have been used by the deponent, or the scribe may have interpreted aspects of the intonation, stress or similar features as amounting to emphasis (cf.Brown 1990: ch.6; Section 3).All the speech descriptors acting as degree modifiers found in ETED express a degree toward the top end of a scale (see Mittwoch, Huddleston and Collins 2002: 722).
(30) […] but vtterly denyeth that hee brought home any beanes att all w th him as is Informed (ETED: Somerset 1635-1637: F_2WC_Somerset_028) Some speech descriptors indicate an assessment of the credibility of what is expressed in the represented speech.These descriptors occur only when a represented speech event is initiated by the reporting verbs report and inform or the noun report (as in (31)).
(31) he hath likewise heard it Credibly reported that the said George Wawdy was angry w th one of his neighbo ~s for informe -ing m ~ Liueley of some tyeth caluee whereof he was likely to be Couzened.
(ETED: Durham 1628-1638: F_2ND_Durham_003) Here the speech descriptors add the deponent's confirmation that the represented speech conveys reliable information.The reason for emphasizing the credibility is probably that the deponent does not provide a specific source that could be called on to verify the statement (a type of hearsay, from a modern perspective).In studies of the depositions from the Salem witch trials in 1692-1693 and some contemporaneous collections in ETED, I have shown that depositions rarely contain instances of represented speech that are not attributed to a specific source (Grund 2012(Grund : 29-30, 2013: 330): 330).That is, it is rare to find statements such as It was reported that…, while the type X reported that… predominates.Speech descriptors that assess credibility may thus be intended to anticipate a challenge to the legitimacy and reliability of a statement with no verifiable source, reinforcing the deponent's conviction of the credibility of what was said.Finally, speech descriptors may also indicate the quality of the voice of the original speaker, although this category is surprisingly uncommon (25 out of 513 Evaluation markers).The descriptors almost exclusively comment on the quietness or loudness of the voice (as in (32)), including (very) loud(ly), aloud, with a loud voice, (very) quietly; rarely do they indicate other aspects of the voice, such as harshness or hollowness (as in ( 33)).Comments on loudness are almost exclusively found in cases of defamation.One of the stipulations of defamation was that the alleged defamatory words should have been spoken in public and heard by many people (Tarver 1995: 113, 116).Emphasizing that something was said loudly was probably intended to address this stipulation: if the words were uttered in a loud manner, they were more likely to have been heard by bystanders.
(32) as hee stood there Samuel Redhead the Defend t in this cause (coming along the high way with his wife towards this Deponent, cryed out w th a loud voyce & in a malitious & reproachfull manner.Gibbons his wife is a whore, is a whore is a whore & Ile maintain't shee's a whore, (ETED: Oxford 1667-1679: F_3SD_Oxford_003) (33) she tempted her in a hollow voyce to kill her child (ETED: Suffolk 1645: F_2EC_Suffolk_065)

Formulation Hedging
Formulation Hedging primarily occurs in various constructions involving the conjunction or.These speech descriptors indicate that the represented speech is not necessarily exactly that of the original speaker but that it may be similar to or the same in "effect", as can be seen in ( 34) and ( 35).
In a limited number of cases (x15), as in (36), the deponent (or scribe) provides alternative formulations.Unlike the examples with effect, this kind of formulation does not suggest that the represented speech is an approximation, but that it is either of two or, in (36), one of three possible representations.
( The function of these speech descriptors can be interpreted in different, though related ways.They can be seen as stance markers where the deponent and/or scribe indicates that there is some uncertainty about the accurate representation of the original formulation.6By hedging the wording, a potential attack could be warded off, where the deponent's memory of particular words or formulations could be questioned.At the same time, although they hedge the faithfulness to the original words, these markers also signal a strong claim.The presence of the word effect in most of the examples is crucial.What the formulation indicates is that, even if the statement is not a verbatim representation, it is still the same or similar in effect, and issues with individual formulations should not invalidate the evidence given by the deponent. In most case types, pinpointing the effect or general meaning of the words would seem to be sufficient.Somewhat surprisingly, this was probably even true for cases that centered on language, where one would expect the accurate representation of particular words to be especially important.Such cases involve threats and verbal abuse, seditious speech, nuncupative wills (i.e., wills given orally) and, especially, defamation, which is the most common case type based on alleged words.Regarding defamation, Moore (2011: 88-98) has convincingly argued that what mattered to early modern ecclesiastical courts was not a verbatim representation of the earlier speech but whether the speech adhered to the legal definition of defamation (as we also saw in 4.2.1).Claiming that the words had the same or a similar effect would seem to be in line with such a principle: it is not the words themselves that matter, but that the words used in the deposition carried the same kind of meaning.Of course, not all defamation depositions contain these kinds of speech descriptors, and Moore's (2011: 98) analysis suggests that, even without overt markers of hedging, "the reported words are valued less for the exactness of their representation than for their conformity to particular phrases and words adjudged defamatory".So, indicating the similarity or sameness in effect may have been a marker of emphasis of an understood convention in the representation of speech: the meaning or effect matters more than the words themselves.Of course, scribal preferences or regional distinctions may also play a part in the use or non-use of these speech descriptors, but there are no clear patterns along these lines: Formulation Hedging devices are found in most (twelve out of thirteen) ecclesiastical collections, in which cases centering on aspects of speech are most common.

Clarification
Clarifications occur primarily in the form of the participle meaning (or another participle such as speaking (of) or intending), as in (37)-(39).They are usually offset from the rest of the text by the use of parentheses, as seen in ( 37) and (38).As indicated by the category label, these speech descriptors act to clarify aspects of the text.They also indirectly appear to signal that the original formulation has been retained but needs some glossing.Clarifications are primarily found in contexts of specifying a person's identity or name (as shown in ( 37)), but they may also be found in discussions of particular objects/places (as in ( 38)) or dialectal/unclear language (as in ( 39)).The clarifications may have been the result of further questioning during the recording of the deposition when, for example, a pronoun with unclear reference was used by the deponent, or the clarification may have been added by the deponent without prompting.The form of the clarification, especially the use of meaning, was undoubtedly scribal, considering its widespread use in legal contexts.It is in line with several other scribal strategies aimed at clarity and specificity in terms of names and people's roles in the court proceedings (see Grund 2011b).
It is possible to see meaning as a (semi-)fixed pragmatic marker in the depositions indicating that what follows is a clarification.It is probably modeled on a similar usage of the Latin participle innuendo 'meaning' (found twice in ETED), which appears to have been common in legal writings more generally in the early modern period.In his Glossographia from 1656, Thomas Blount explains innuendo as […] a Law term, most used in Declarations and other pleadings; and the office of this word is onely to declare and design the person or thing which was named incertain before; as to say, he ( innuendo the Plaintiff) is a Theef; when as there was mention before of another person.
Black's Law Dictionary (s.v.innuendo) notes that this usage is still found in present-day legal texts, but that an English equivalent can be used in its stead, including meaning.The ETED depositions suggest that this use of meaning goes at least as far back as Early Modern English.Tracing the exact dynamics of this development is beyond the scope of this study, but it may be similar to the genre-specific grammaticalization of viz (or videlicet) as a marker introducing (direct) speech in early modern depositions, as discussed by Moore (2006).

Emphasis
Emphasis markers are the most restricted of the speech descriptors in terms of their appearance, function and context of use.In criminal depositions, in which they occur most frequently (145 out of 173 instances, or 84%), they almost exclusively appear in the form of a prepositional phrase, upon/on his/her/their oath(s), at the very beginning of a deponent's statement, as in ( 40).These descriptors of course serve as signals of the circumstances under which the deposition was made (i.e., after the taking of an oath).In that respect, they are similar to the markers of physical context (place, time, etc.) that I excluded (see Section 3).However, on another level and unlike markers of physical context, these descriptors also emphasize that the deponent's speech (recorded by the scribe) should be understood as an accurate, truthful and legally binding statement.In some depositions, the note about the swearing of an oath is found elsewhere in the deposition, and some depositions may even have two annotations, one occurring together with speech representation at the beginning of the deposition, as in (40).This kind of doubling supports the notion that the oath phrase carries a particular, emphasizing meaning in the context of speech representation, as I argue here.
Emphasis markers perhaps more indirectly describe the represented speech than the other categories.Unlike most of the other types of speech descriptors, they are also more clearly attributable to the scribe rather than the deponent, as they reflect the scribe's annotation rather than the deponent's comment.At the same time, Emphasis markers also reflect the deponent's commitment to the veracity of the testimony.

Frequency/Quantity
Frequency/Quantity markers occur more frequently in criminal than in ecclesiastical depositions.Most directly, these markers show how many times a particular statement was made or the number of words used or statements made.The frequency/quantity is mostly (x138, or 88%) given in unspecified terms such as diverse times, many, much (as in ( 41)), often (as in ( 42)), very frequently, used to + verb (see example (24) above) and many hundred times.Rarely is a more specific frequency/quantity indicated (x19, or 12%), as in ( 43).As I suggested in 4.1, descriptors that signal quantity allow the focus to be not on the exact words of the previous statement, but on aspects of the speech that the deponent wants to highlight.Much in (41) stresses the intensity of the altercation, which may have been more important than the actual words that Joan and Mary exchanged.Similar dynamics are arguably involved in speech descriptors that signal frequency.On one level, they are obviously used in lieu of the deponent repeating and the scribe recording the same statement two or more times.At the same time as they function as convenient short-hand or abbreviation devices, these markers also allow the deponent (or less likely the scribe) to emphasize the severity of a situation or the fact that a statement was not a coincidence but that there is a pattern in a person's (abusive) behavior.In (42), by using often (or some such formulation written down as often by the scribe), John Young, a bailiff, stresses the hostile environment that he and his fellow bailiffs found themselves in when a group of people attacked them to free a prisoner in their custody.In (43), by using a frequency indication, the deponent is able to suggest that this was not a one-off occurrence and hence perhaps imply that the statement was not simply made in a fit of anger.If the potentially defamatory words were simply seen as part of an angry outburst, they may not have been seen as intentionally malicious and hence not defamatory (Tarver 1995: 113; cf.Section 4.2.1).
In some cases, these Frequency/Quantity markers were clearly meant to signal emphasis rather than provide an accurate impression of frequency/quantity.In (44), it is unlikely that the defendant had actually called the deponent cuckold and his wife whore "many hundreth tymes"; instead, the expression conveys an emphasis on repeated behavior.

Conclusion
So, what has this study shown and what are the next steps in exploring speech descriptors?Speech descriptors clearly occur in a variety of linguistic forms, and they can perform a number of different textual and pragmatic functions.Speech descriptors are not obligatory: they are far fewer than the number of instances of speech representations.At the same time, they allow language users to add information about represented speech that could not be signaled by other means when such aspects are seen as significant for a specific purpose.Importantly, speech descriptors frequently function as flexible stance resources.Many of them are overtly evaluative, showing the deponent's (or, in some contexts, the scribe's) assessment or interpretation of the intention, character or implication of the speech event and hence also the deponent's alignment with the plaintiff or defendant in the case.This study has opened up a number of avenues of research.On a general level, it shows that our study of speech representation cannot be limited to the reporting expression or the speech representation itself.Rather, representing speech involves a number of linguistic strategies working in conjunction, including descriptive markers as discussed in this article.We still have much to learn about the nature of speech descriptors and their interplay with other speech representation features.The categories and functions of speech descriptors in ETED are clearly connected with the larger legal context: the descriptors are used to respond to demands of particular case types (esp.defamation) or the need for clarity in the legal narrative.We need studies of other genres (historical and present-day) to see whether the same categories are found elsewhere and what other possible functions speech descriptors can perform.Another desideratum is a comparison between reporting expressions that indicate attitude (see Section 2) and speech descriptors to see how they complement or overlap in function.A comprehensive study of speech descriptors across genres will give us a fuller understanding of the choices language users make when representing or suppressing aspects of previous speech events.

( 40 )
Who (vpon her Oath) saith y t Stephen: Shepheard went to water her horse about eight or Nine of y e clock of y e same Night y t Robert: Emotts Wife was supposed to be Murthered, (ETED: Lancaster 1696-1698: F_3NC_Lancaster_008)

Table 1
. Categories of speech descriptors in criminal and ecclesiastical depositions (per 10,000 words within parentheses)

Semantic/Functional categories Criminal court depositions Ecclesiastical court depositions
As seen in Table1, the five major semantic-functional categories of speech descriptors are Evaluation, Emphasis, Frequency/Quantity, Formulation Hedging and Clarification.I will discuss these categories in detail in later sections (4.2.1-4.2.5), but (6)-(10) give prototypical examples by way of a first introduction.Evaluation markers indicate the speaker's evaluation of a number of different aspects of the represented speech.In (6), maliciously signals the speaker's evaluation of the intent of the speech (i.e., to malign).Markers of emphasis (as in (7)) are limited to the context of emphasizing that the deponent's spoken testimony (recorded by the scribe) is in accordance with legal procedure and hence an accurate, legally binding statement.Frequency/Quantity markers (as in (8)) simply indicate how many times or how often a statement was made or the number of words, expressions or terms used.Formulation Hedging, as in (9), signals that the represented speech may not be a verbatim representation.Finally, Clarifications provide a gloss that more closely identifies a place, object or person, as in (10).(6) in conclusion Pickford maliciously called the sayde ffraunces Robins whore (ETED: Winchester 1566-1577: F_1SD_Winchester_011) (7) Thomas Longley of Lepton in the county of yorke shooemaker saith vpon oath that vpon saturday the third of this instante October about eight of the clocke in the nyght, he this informer mett one Edward Greenehold in Kyrkeheaton feild in the way goeinge towards his lodgeinge, (ETED: Northern 1646-1649: F_2NC_Northern_001) (8) James Ewens Came after him & drew his Bagonett & Swore Seu ~all times y t he would Kill this Inform t (ETED: Somerset 1682-1688: F_3WC_Somerset_036) (9) […] vtterringe and Speakeing vnto him as followeth./Gode Curse, gode plague, light of the, ^{and all thine} speakeinge vnto y e said George ffenwick and

Table 2 .
Linguistic realizations of speech descriptors in criminal and ecclesiastical depositions Inform t further saith that the said Thomas Applin att the time aforesaid spoke seu ~all other words very scandalous & abusesive ag t the Justices of the peace that was for puttinge out of his ffather as aforesaid (ETED: Somerset 1706-1716: F_4WC_Somerset_009) Then the said Catford swore very fearefully thatt hee would Cutt off this Ex ates mares legges, & foorthw th vseinge many revileinge lewde ).5(16) then spake y e said Rayff to y e said John Rosse very angrely ^{slave} what is y t y u saies Participle construction" in Table2involves primarily the participle meaning (118 out of 127 instances, or 93%), as in (18) and (19), and less commonly other participles such as speaking (of), intending, and innuendo 'meaning, that is to say' (see Section 4.2.3).These participles are primarily used within the category of Clarification, specifying an identity or name.Meaning may in fact have become a (semi-)fixed pragmatic marker of clarification in depositions (see Section 4.2.3).