UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The semantics and pragmatics of English evidential expressions : the expression of evidentiality in police… Glougie, Jennifer Robin Sarah 2016

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

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
24-ubc_2016_november_glougie_jennifer.pdf [ 1.78MB ]
Metadata
JSON: 24-1.0319268.json
JSON-LD: 24-1.0319268-ld.json
RDF/XML (Pretty): 24-1.0319268-rdf.xml
RDF/JSON: 24-1.0319268-rdf.json
Turtle: 24-1.0319268-turtle.txt
N-Triples: 24-1.0319268-rdf-ntriples.txt
Original Record: 24-1.0319268-source.json
Full Text
24-1.0319268-fulltext.txt
Citation
24-1.0319268.ris

Full Text

    THE SEMANTICS AND PRAGMATICS OF ENGLISH EVIDENTIAL EXPRESSIONS:   THE EXPRESSION OF EVIDENTIALITY IN POLICE INTERVIEWS    by  Jennifer Robin Sarah Glougie  LL.B., University of Victoria, 2003 M.A., The University of British Columbia, 2001 B.A., The University of British Columbia, 1998      A DISSERTATION SUBMITTED IN PARTIAL FULFILLMENT OF  THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY  in The Faculty of Graduate and Postdoctoral Studies  (Linguistics)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)      October 2016      © Jennifer Robin Sarah Glougie, 2016  ii ABSTRACT  The goal of this dissertation is to examine how English speakers express their evidence in the context of police interviews. I show that speakers use discourse markers, in particular, actually, apparently and supposedly, to explain their evidence in a criminal investigation. The data for this research was collected exclusively from transcripts of police interviews of lay witnesses in the investigation into the disappearance and murder of Caylee Anthony that occurred in Orange County, Florida, between 2008 and 2011. I show that actually marks evidence strength and is felicitous where the speaker has the ‘best’ evidence for their proposition. Actually’s evidential contribution largely parallels the best possible grounds evidential -mi in Cuzco Quechua, and contrasts with that observed for English must. Apparently marks that the speaker’s evidence for the proposition is indirect and supposedly marks that the speaker has reported evidence for the proposition and that they distrust the report.   In addition to what evidentials mean, this dissertation considers what speakers use evidentials to do. I show that speakers use evidentials to negotiate the common ground (cg) of discourse. While a bare assertion proposes its propositional content for inclusion in the cg, speakers use actually-assertions both to propose the propositional content for inclusion and to advocate for its inclusion by marking that the speaker has best evidence for that content. Because actually highlights the strength of the speaker’s evidence, it can be used to achieve delicate discourse actions like correcting, challenging and disagreeing. In questions, actually puts the addressee on notice that the information proposed in a bare assertion cannot be included in the cg without more information; actually-questions encourage the addressee to justify their evidence either by disclosing the source of their evidence or by expressly  iii aligning as author and/or principal of that information. Speakers use apparently and supposedly to proffer information that may be relevant to the investigation but without proposing it for inclusion in the cg, because they are either agnostic about its reliability or know it to be untrustworthy.                                         iv PREFACE   This dissertation consists of original and independent work by Jennifer Robin Sarah Glougie.    A version of the material in Section 3.1.1 has been published in Matthewson, L. and J. Glougie. 2015. Justification and Truth: Evidence from Languages of the World. Submitted to McCready, Eric, Masaharu Mizumoto, Jason Stanley and Stephen Stich (eds.), Epistemology for the Rest of the World, of which I am an author.  This research did not require ethics approval.                               v TABLE OF CONTENTS  ABSTRACT  ........................................................................................................................ ii  PREFACE  .......................................................................................................................... iv  TABLE OF CONTENTS  .........................................................................................................v  LIST OF TABLES  .................................................................................................................x  LIST OF FIGURES  .............................................................................................................. xi  LIST OF ABBREVIATIONS  ................................................................................................. xii  ACKNOWLEDGMENTS  .................................................................................................... xiii  1             INTRODUCTION .......................................................................................................1 1.1 Evidentiality  ..............................................................................................4  1.1.1 What is a linguistic evidential system?  .........................................5 1.1.2 What do evidentials encode?  ......................................................13 1.1.3 English must and the requirement for ‘not best’ evidence  .........17 1.1.4 Cuzco Quechua -mi and the requirement for ‘best’  Evidence  .....................................................................................18 1.1.5 English discourse markers as evidentials  ...................................22 1.2 Police interviews as a source of evidentiality ..........................................23 1.2.1 The role of ‘facts’ in the legal process ........................................24 1.2.2 The extent and limits of previous research on police  interview language  ......................................................................27 1.2.3 Police interviews as institutional discourse  ................................28 1.2.4 Police interviews vs. police interrogations  .................................30 1.3 Actually in the literature ..........................................................................34  vi 1.3.1 Syntactic analyses of actually  .....................................................35 1.3.2 Pragmatic functional analyses of actually  ..................................36 1.4 Apparently in the literature  .....................................................................39 1.5 Supposedly in the literature  .....................................................................40 1.6 Overview of dissertation  .........................................................................41 2 THE POLICE INTERVIEW TRANSCRIPTS  ..................................................................45 2.1 The disappearance and murder of Caylee Anthony  ................................46 2.2 The Caylee Anthony investigation  .........................................................46 2.2.1 The nature of language in transcripts ..........................................47 2.2.2 The interview transcripts in the Caylee Anthony  investigation  ................................................................................51 2.2.3 The interviews  ............................................................................53 2.3 The speakers  ...........................................................................................54 2.3.1 The interviewers  .........................................................................55 2.3.2 The witnesses  ..............................................................................57 2.3.2.1 Social network witnesses  ................................................58 2.3.2.2 Non-social network witnesses  ........................................63 2.4 The timeline of the interviews  ................................................................64 2.5 The data  ..................................................................................................65 2.5.1 Data coding  .................................................................................65 2.5.2 The tokens  ...................................................................................70 2.5.2.1 The use of actually ..........................................................70 2.5.2.2 The use of supposedly and apparently ............................73  vii 2.6 Summary  .................................................................................................74 3 THE SEMANTICS OF ENGLISH EVIDENTIALS  ..........................................................75 3.1 The evidence that licenses actually, apparently, and supposedly  ..........76 3.1.1 Where actually occurs  ................................................................77 3.1.1.1 Sensory observation  ........................................................77 3.1.1.2 Trustworthy reports  ........................................................79 3.1.1.3 General and encyclopedic knowledge  ............................84 3.1.1.4 Speaker performance  ......................................................86 3.1.2 Actually does not occur where must is felicitous  ........................89 3.2 The semantics of actually  .......................................................................94 3.2.1 Actually encodes ‘best’ evidence  ................................................98 3.2.2 Evidence strength and speaker variation  ..................................101 3.2.2.1 Inter-speaker variation and the assessment of ‘best’  Evidence  .......................................................................102 3.2.2.2 Intra-speaker variation and time of utterance ................108 3.3 The semantics of apparently  .................................................................118 3.3.1 Apparently marks indirect evidence  .........................................119 3.3.2 Apparently vs. must; evidence type vs. evidence strength  ........123 3.4 The semantics of supposedly .................................................................128 3.5 Summary of the semantics of English evidentials, so far  .....................133 3.6 Legal rules of evidence  .........................................................................135 3.6.1 Knowledge, observation and experience  ..................................135 3.6.2 The Rule Against Hearsay  ........................................................136  viii 3.6.3 General and encyclopedic knowledge  ......................................140 3.6.4 ‘Best’ evidence and the legal rules of admissibility  .................142 3.7 Actually, apparently, and supposedly and the alignment of  participant roles  ....................................................................................143 4 THE PRAGMATICS OF ACTUALLY  ..........................................................................146 4.1 Actually and the negotiation of the common ground .............................150 4.1.1 Where speakers use actually versus bare assertions  .................152 4.1.2 Actually as a means of eliciting appropriate response  ..............154 4.1.3 Actually and delicate discourse actions  ....................................161 4.1.3.1 Self-correction ...............................................................161 4.1.3.2 Challenge  ......................................................................167 4.1.3.3 Persuasion  .....................................................................182 4.1.4    Negotiating the common ground, summary  .............................191 4.2 Actually and topic control  .....................................................................191 4.2.1 Introduce a new topic ................................................................192 4.2.2 Resist topic shift  .......................................................................195 4.2.3 Discourse relevant digressions  .................................................196 4.2.4 Evidentials and topic control cross-linguistically  .....................200 4.2.5 Actually and topic control; the problem for a unified  analysis  .....................................................................................204 4.3        Summary of actually’s pragmatic function  .........................................206 5 THE PRAGMATICS OF APPARENTLY AND SUPPOSEDLY  ..........................................208 5.1 The pragmatic use of apparently  ..........................................................209  ix 5.2 The pragmatic use of supposedly  ..........................................................220 5.3 Multiple markers of evidentiality  .........................................................226 5.4 Summary of the pragmatic use of apparently and supposedly  .............235 6 ACTUALLY IN QUESTIONS  ....................................................................................236 6.1 Evidentials in questions  ........................................................................238 6.2 Questions in police interviews  ..............................................................241 6.3 The use of actually in questions  ...........................................................246 6.3.1 Actually-questions are addressee-oriented  ................................246 6.3.2 Actually-questions are confirmation-seeking  ...........................252 6.4 Summary  ...............................................................................................265 7            CONCLUSION  ......................................................................................................267 7.1       The pragmatics of English evidentials; where do we go from here  .....268 7.2       Forensic linguistics; where do we go from here  ...................................276 7.3       The semantics of English evidentials; where do we go from here  .......277 7.4       Methodology and the study of evidentials  ............................................279  REFERENCES  ..................................................................................................................283 APPENDIX A: LIST OF WITNESS INTERVIEWS   ................................................................292               x LIST OF TABLES  TABLE 1.1 Matthewson’s evidential specifications  ........................................................16 TABLE 2.1 Speaker demographic information for Casey Anthony’s Friends  .................61 TABLE 2.2 Speaker demographic information for Casey Anthony’s Acquaintances  .....62 TABLE 2.3 Token count by marker  .................................................................................70 TABLE 2.4 Frequency of occurrence in spoken English vs. police interview                    transcripts  ......................................................................................................71 TABLE 3.1Tokens of actually by evidence type (where the source of                    information can be determined)  ..................................................................107 TABLE 3.2 English evidential specifications  .................................................................133        xi LIST OF FIGURES  FIGURE 4.1 The recurrent sequential format of the -telako evidential marker  ..............157                      xii LIST OF ABBREVIATIONS  1  first person 3  third person  ABL  ablative ACC  accusative   ADV  adverb  AFAR  translocative (at a distance) ASSOC  associative  ASSUM  assumption ATTR  attributive  CL  classifier CONN  connective   CONTR  contrastive CORREL correlative DC  declarative  DCT:RE  deductive reasoning DIR  direct EV  evidential value  FR   factual realization GEN  genitive IE  informal ending IMPFV  imperfective  IND  indirect  INF  inferential INTERROG interrogative INTROS  introspective  LOC  locative  NEG  negation  NONVIS non-visual NML   nominalizer NOM  nominative  PL  plural  POL  polite speech level POSS  possessive  PROG  progressive PST  past QT  quotative  RAPP  rapport REC.P   recent past  REP  reportative RT  retrospective  TOP  topic  TRANS  transferentive VIS  visual   xiii ACKNOWLEDGMENTS   In the words of Marlo Stanfield, I ain’t much for sentiment. Nonetheless, I would be unforgivably remiss if I didn’t take this opportunity to offer my heartfelt thanks to the various folks who have supported, encouraged, and prodded me along this journey.  To my dissertation committee, Lisa Matthewson, Hotze Rullmann, and Henry Davis, who agreed to oversee a dissertation that was, at least at first blush, a bit outside the norm for UBC’s linguistics department. Thank you for your dedication, your enthusiasm, and your invaluable feedback. This dissertation is infinitely better for your involvement with it. To my family, Rob, Wendie, Jon, Jessica, Walter, and Riley, who have the patience of saints, and Peter, Ross, Vickie and Helen, who I wish could have been here. Love you all. To my colleagues at Banister & Company, Barristers and Solicitors, Sandy, Jonathan, Jeremy, Sara, Amanda, Jaye, Janet, and Jocelyn, who tolerated my need to work flexible, part-time hours for the last decade without a word of complaint. To my new colleagues at the BC Labour Relations Board, who supported me through the last push to finish this dissertation.  To the friends who kept me motivated even when I thought this process might never end: in particular, Jonathan & Karina, Lorna, Doug, Mara & Jonathan, Tim & Breena, Brian, Kate & Nolan, Tammy, Karen, Jeremy & Erin, Anita, Jen, Molly, Dana and Natasha. To my colleagues in the International Association of Forensic Linguists and the West Coast Roundtable on Language and the Law, who have provided and continue to provide invaluable mentorship as I develop my career as a scholar, in particular: Larry, Janet, Ed, Mel, and (the late) Peter. To the good folks at Blenz on the corner of Granville and Davie Streets, who kept me caffeinated throughout. Last, and in no way least, to Jonathan: SO much better.  1 “Talkin’ ain’t knowin’, is it?” Lester Freamon, The Wire      1. INTRODUCTION  The goal of this dissertation is to examine how English speakers express their evidence in the context of police interviews. I show that English speakers use actually, apparently, and supposedly to mark evidentiality. Adopting Matthewson’s (2015a, 2015b) analysis in which evidentials contribute one or more of three possible dimensions of meaning, I show that actually’s evidential contribution is on the evidence strength dimension. To this end, actually largely parallels the Cuzco Quechua evidential -mi which Faller (2002) analyses as encoding best possible grounds. By marking that a speaker knows the proposition they express on the basis of best evidence, actually is the opposite of must, which encodes that the speaker’s evidence for the proposition is ‘not best’ (von Fintel & Gillies, 2010; Matthewson, 2015a, 2015b). I further show that apparently makes an evidential contribution on the evidence type dimension of meaning, encoding that the speaker’s information source is indirect. Supposedly contributes on both the evidence type and evidence strength dimensions of meaning, marking that the speaker’s information source is an untrustworthy report.  The traditional diagnostic to determine whether languages are properly classified as ‘evidential’ languages, advocated for by Aikhenvald (2004) among others, is whether they have grammatically obligatory evidential marking. I argue that this diagnostic is flawed. Even in those languages which have traditionally been classified as ‘evidential’ languages, evidential marking is shown not to be grammatically obligatory. Rather, even in those  2 languages, a speaker’s decision whether to use an evidential and, if so, which evidential to use depends on the discourse goals the speaker hopes to accomplish. If evidential use is driven by discourse and not by some rule of grammatical obligatoriness, then there is no reason not to search for commonalities of meaning between English evidential elements and evidentials in supposedly obligatory evidential languages, or to compare them.  The data for this study come exclusively from the transcripts of witness interviews prepared in the course of a criminal investigation. I show that the institutional role of witnesses in the criminal justice process, namely, to provide accurate, reliable information on the basis of which to advance the investigation, makes such discourse ideally suited to the study of evidential use. While both Aikhenvald (2004: 4) and Faller (2002: 2, fn. 2) are careful to distinguish the linguistic concept of evidentiality from the legal concept of evidence, the legal context should not be discounted as a potential source of non-contrived language where evidence, both what the speaker knows and how they know it, is expressed.  All of the data I relied on for this study were generated in the course of the police investigation into the disappearance and murder of Caylee Anthony which was launched in July 2008 in Orange County, Florida. The police interview data show that English speakers use the discourse markers actually, apparently, and supposedly regularly and consistently to indicate evidentiality. Specifically, I show that speakers use actually where they have ‘best’ evidence for their propositions, apparently where they have indirect evidence, and supposedly where they have reported evidence that they specifically distrust. I argue that the semantic analyses correctly predict the various functions observed in the police interview transcripts and previously attributed to these discourse markers in the literature.   3 In addition to what evidentials mean, this dissertation is concerned with what evidentials do; specifically, I show that speakers use evidentials to negotiate the common ground in the witness interviews. The purpose of interviewing non-suspect witnesses is to establish the facts which prove an offence. The interviewer knows what facts are necessary to advance the case and the witness has evidence which may prove those facts. Thus, witness interview discourse will involve considerable negotiation to ensure the common ground contains the most accurate information from the most authentic source. However, despite each having their own necessary role, there is a significant power imbalance between the discourse participants in a police interview; the interviewers are in a position of institutional authority and the witness is not. Thus, a witness may be required, for the sake of the accuracy of the record, to correct or contradict the more institutionally powerful interviewer. I show that speakers are able to use evidentials to do so while maintaining the politeness norms dictated by their limited institutional authority.   By analysing how English speakers express evidentiality in investigative interviewing, this research will allow us to make testable predictions about how discourse markers are similarly used in other (non-legal) discourse contexts. Moreover, this research is potentially useful to investigative interviewers with respect to best practices for effective witness interviewing. Specifically, by understanding their evidential contribution, interviewers will be better able to tailor follow up questions when a witness gives evidence with actually, apparently, or supposedly. This research informs the extensive formal literature on the semantics and pragmatics of English discourse markers, as well as the literature on evidentiality, discourse analysis, and forensic linguistics.    4 1.1    Evidentiality  ‘Evidentiality’ is the linguistic means of indicating the speaker’s source for and the reliability of the information they have for the propositions they express. An ‘evidential’ is a marker of evidentiality (Faller, 2002). Anderson (1986: 274-275) proposed that archetypical evidentials can be identified on the basis of the following criteria:  (1.1)        a. Evidentials show the kind of justification for a factual claim which is  available to the person making that claim. b. Evidentials are not themselves the main predicate of the clause, but are rather a specification added to a factual claim ABOUT SOMETHING ELSE. c. Evidentials have the indication of evidence as their primary meaning, not only as a pragmatic inference. d. Morphologically, evidentials are inflections, clitics, or other free syntactic elements (not compounds or derivational forms). (capitals in original)  Evidentials typically mark how the speaker knows the information in the propositions they express. For example, the following data from Tariana (from Aikhenvald, 2004) differ only with respect to the speaker’s evidence for the proposition José has played football, as reflected by the different evidentials used.    (1.2) Tariana evidentials a. Visual Juse iɾida  di-manika-ka José football 2sgnf-play-REC.P:VIS “José has played football (we saw it)”    5 b. Auditory Juse iɾida  di-manika-mahka José football 2sgnf-play-REC.P:NONVIS “José has played football (we heard it)”  c. Reported Juse iɾida  di-manika-pidaka José football 2sgnf-play-REC.P:REP “José has played football (we were told)”  d. Inference Juse iɾida  di-manika-nihka José football 2sgnf-play-REC.P:INFR “José has played football (we infer it from visual evidence)”  e. Assumption Juse iɾida  di-manika-sika José football 2sgnf-play-REC.P:ASSUM “José has played football (we assume this on the basis of what we already know)” (from Aikhenvald, 2004: (1.1)-(1.5))  The data in (1.2a-e) differ only with respect to the speaker’s evidence for the proposition; the evidentials indicate whether the speaker’s evidence is visual (1.2a), auditory (1.2b), reported from a third party (1.2c), inferential (1.2d) or by assumption (1.2e).   1.1.1 What is a linguistic evidential system? While it is trite to say that all languages have a mechanism for expressing evidence, the question of what constitutes a ‘linguistic evidential system’ has proven controversial. A language is said to have an ‘evidential system’ when it has a paradigm of linguistic markers in which each marker uniquely encodes information about the source or reliability of the information on which they base their claim (Peterson, 2010: 39). However, certain scholars have restricted the definition of ‘evidential system’ to only those languages where the expression of evidentiality is grammatically obligatory (see, for example, Aikhenvald, 2004, and the discussion therein). On this view, English has not traditionally been characterized as  6 an evidential language despite having a “rich repertoire of evidential devices” (Chafe, 1986: 261), including lexical verbs (allege, report, observe), phrases (I guess, they say, I heard), status-adverbials (reportedly, allegedly), and introductory clauses (it seems to me that) (Aikhenvald, 2003: 1-2). Aikhenvald justifies the exclusion of English evidential expressions from the category of ‘evidentials’ as follows.  Saying that English parentheticals are ‘evidentials’ is akin to saying that time words like ‘yesterday’ or ‘today’ are tense markers. These expressions are not obligatory and do not constitute a grammatical category; consequently, they are only tangential to the present discussion. Saying that English has ‘evidentiality’ (cf. Fox 2001) is misleading: this implies a confusion between what is grammaticalized and what is lexical in a language. (2004: 10)  There are, however, theoretical and empirical objections to Aikhenvald’s claim that only languages with grammatically obligatory evidential markers should be characterized as having evidential systems.  From a semantic perspective, if what we are interested in is how speakers express concepts of evidence, then languages like English which do not have grammatically obligatory evidentiality systems will contribute to our understanding. Languages like English differ from those with grammatical obligatory evidentiality with respect to how they mark speaker information, but not with respect to what speaker information is marked. Languages which have grammatically obligatory evidential marking may form a subset of linguistic evidentiality worthy of study in their own right. However, the fact such a subset may exist does not, in itself, justify the exclusion from the broader discussion of languages like English which do not obligatorily encode information about speaker evidence. Indeed, “the motivation to make knowledge visible in discourse through the utilization of evidential resources transcends any particular linguistic system and has therefore something to  7 contribute to our understanding of human social interaction through language” (Mushin, 2013: 630). The idea that only languages with grammatically obligatory evidential marking are properly categorized as ‘linguistic evidential languages’ is particularly problematic given that the empirical evidence from some such languages suggests that evidential use is not always grammatically obligatory; rather, corpus and pragmatic studies have shown that speakers’ evidential use in so-called ‘evidential’ languages is (a) not always obligatory and (b) motivated primarily by discourse considerations.1 For example, Kim (2006) shows that Korean speakers’ use of evidentials is determined not only by the source of their information but also by the interactional context in which the use arises. Korean is traditionally considered a grammatical evidential language in which a speaker’s source of information is obligatorily marked (see, for example, Aikhenvald 2004). However, Kim shows that, in naturally occurring discourse, speakers often do not use evidentials unless the discourse context requires them to do so.2 Specifically, she argues that when stating their claim, speakers will often use a zero evidential marker in their first pass and only switch to an overt evidential marker when their claim fails to elicit the expected response from their interlocutors.                                                   1 See also the discussion of optional versus obligatory tense marking in Cable (2016) and Bochnak (2016).   2 Another example of this occurs in the First Nations languages spoken in the Pacific Northwest which mark reported speech via a reportative or quotative evidential. In traditional legends, for example, reportatives are used at the beginning of the narrative, from time to time during the course of the narrative to remind listeners of the genre, and as a stylistic device to distance the narrator from the narration. If the reportative evidential were grammatically obligatory, we would expect it to be used more robustly throughout the narrative. Thank you to Henry Davis (p.c.) for bringing this to my attention.    8 As an example, Kim gives the following data which represents a telephone conversation between two friends, S and T, in which they discuss a mutual friend, Sangmi, and her recent weight loss. In Turn 06, S introduces the topic of Sangmi’s weight loss by using the simple declarative suffix -ta which marks the utterance as being noteworthy; S does not use an evidential to indicate her source of information for p. T responds in Turn 08 using the past experiential evidential -telako,3 which combines with the quotative construction -ta kwu to indicate she knows p (‘that Sangmi has been diligently working out’) by hearsay. S reiterates the extent of Sangmi’s weight loss in Turn 10, again without an evidential, before eventually using the direct evidential marker -telako in Turn 16.  (1.3) Korean   06 S: uh kuntey  sangmi-ka (.) cincca sal-ul  manhi    uh by.the.way S-NOM  really weight-ACC much     ppay-ss-ta    lose-PST-DC   “Uh, by the way, Sangmi has really lost a lot of weight”   07 T: uh  kulay-ss-e¿=   uh be.so-PST-IE    “Uh, is that so?”  08  =wuntong [yelssimhi ha-n-ta-kwu  kule-te-la-kwu   exercise diligently do-IMPFV-DC-QT say-RT-INTROS-QT   “(I heard that) she has been diligently working out.”  09 S:   [e      “Yes.                                                       3 Kim analyses -telako as a morphologically complex evidential marker composed of the retrospective suffix -te, the introspective suffix -la and the quotative particle -ko. She argues that speakers use it to report in the current moment what they previously perceived through experience (at page 35).   9 10  kuntey cin:cca manhi ppay-ss-e   but really much lose-PST-IE   “Yes.  But (she) really lost a lot of (weight)”  11 T: uh=   “Uh huh”  12 S: =kutongan ccokkumssik ppaci-te-ni,   while  little.by.little lose-RT-DET   “(She started to) lose weight little by little,”  13 T: uh=   “Uh huh”  14 S: nay-ka han il-cwuil cen-ey  manna-ss-keteng?   I-NOM about one-week before-at meet-PST-CORREL   “I met (her) about a week ago,”  15 T: uh=   “Uh huh”  16 S: kuntey cengmal ppacy-e iss-te-la-ko:   And really  lose-CONN exist-RT-INTROS-QT    “And [I saw] (that she) really lost the weight”  17 T: Eh::   “Oh::” (adapted from Kim (2006): 39-40)  Kim observes that S’s use of -telako in Turn 16 indicates that she has direct (visual) information for p; as she explains in Turn 14, she saw Sangmi the week previously. Nonetheless, S does not use -telako to introduce p (‘Sangmi has really lost the weight’) until the third turn in which she mentions p and even then only after T has failed to exhibit the expected reaction to p in Turns 6 and 10. S’s use of a zero-marker in Turns 6 and 10 is unexpected if Korean evidentials were truly grammatically obligatory in the sense contemplated by Aikhenvald. Rather, as Kim convincingly shows, even grammatically encoded evidentials do not, in fact, occur obligatorily. The fact that, empirically, evidentials are not always obligatory, even in so-called grammatical evidential languages, supports the  10 argument that grammatical obligatoriness is not a proper diagnostic for determining which languages should be classified as having ‘linguistic evidential systems’ and which should not.   If evidentials are not always grammatically obligatory, as Kim (2006) shows for Korean, then what motivates a speaker to use a particular evidential in a particular context? Kim shows that, where a speaker does use an overt evidential marker, their choice of evidential is determined not only on the basis of the speaker’s information source, but also on the basis of the speaker’s interactional goal in the discourse. She argues that the types of interactional goals that motivate evidential use include: achieving entitlement to information, objectivity or detachment, distributing responsibility, reorganizing the participation framework of the interaction and more (Kim, 2006: 3-4). That is, the discourse context determines whether an overt evidential will be used and the source of the speaker’s information for p and their discourse goal in uttering p determine which evidential is used.  That evidential use is governed by discourse goals does not appear limited to Korean; evidence from other ‘grammatical’ evidential languages suggests that speakers’ use of overt evidential marking is governed in large part by their interactional goals in the discourse. Indeed, data from such languages suggest that the speaker’s pragmatic goal, and not their information source, is the primary motivator with respect to the speaker’s evidential choice. Data from Wanka Quechua, for example, show that a speaker’s choice of evidential marking is motivated in large part by the discourse context. For example, a speaker can use the direct evidential -mi even where they have not witnessed the event itself. Aikhenvald (2004) observes that a speaker may respond to the allegation that he fathered a child with the sentence in (1.4), which includes the direct evidential marker -mi.  11 (1.4) Wanka Quechua  mana-m chay ya’a-pa-chu   not-DIR.EV that 1-GEN-NEG  “That (the child) is not mine.”  (Floyd, 1999: 73, cited in Aikhenvald, 2004: (10.19)) Aikhenvald points out that the effect of (1.4) is not to indicate that the speaker has direct evidence for the event itself but to impose their conviction on others. As Floyd (1999) observes in respect of the same example, the speaker should have used the inferential evidential since the identity of the child’s father is, technically, a matter of inference. However, an inferred evidential would leave open the possibility of the speaker being the father; by using the direct evidential, such an alternative is disallowed (Aikhenvald, 2004: 318; Mushin, 2013). Thus, in (1.4) it is the speaker’s interactional goal (persuasion) and not his evidence source (inference) which motivates his evidential choice.  Similarly, for Japanese, Trent (1997) shows that a speaker’s use of grammatical evidential markers is primarily motivated by the discourse context. Specifically, Trent shows that evidentials are used to assert or reserve a claim to ownership of the information. Trent points out that, in Japanese, a wife may speak about her husband’s life using direct forms. Nonetheless, the speaker in (1.5b) uses two indirect forms (mitai and the quotative marker) to convey information that she knows to be both reliable and certain.  (1.5) Japanese  a. Go-shujin  no kaisha  doo?   Your-husband POSS company how  “How is your husband’s company doing?”  b. Chotto dame  mitai.  Raigestu heisasaru-koto ni    no.good it.seems next.month close-COMM  12    kimatta-tte.  Shujin  ga kinoo  itteta wa.   decided-QT husband NOM yesterday said RAPP  “It seems that it is not doing well. I heard they decided to close the company next month. My husband told me yesterday.” (from Trent, 1997: (3-30))  Trent (1997: 107) claims that speaker (b)’s intention in using indirect rather than direct forms is to show modesty and defer to her husband’s information territory. In other words, even where she has direct reliable information about the state of her husband’s business, she may use indirect forms to convey that information out of respect for the fact that it is her husband’s information to share. Speaker (b)’s choice of indirect evidence marking is unexpected if ‘source of information’ is the primary factor at play in determining evidential choice.   Korean, Wanka Quechua, and Japanese have all traditionally been classified as having grammatical evidentiality. Nonetheless, the data in (1.3) through (1.5) suggest that, for speakers of these languages, evidential use is motivated by the discourse and not purely by the speaker’s information source. That is, the presence or absence of overt evidential marking is determined by the discourse context and not necessarily because such marking is grammatically obligatory. Moreover, the speaker’s choice of evidential is motivated as much by their discourse goal as by their information source. If the empirical evidence from at least some of the languages which have traditionally been accepted as having ‘evidential systems’ shows that evidentials are not always grammatically obligatory, then there is no principled reason to maintain the distinction between languages that do and do not have grammatical evidentiality. That is, if languages do not have truly obligatory evidential marking as the empirical evidence suggests, then  13 grammatical obligatoriness is not a compelling diagnostic for linguistic evidentiality. Moreover, without grammatical obligatoriness as a determining factor, the question becomes: what motivates a speaker to overtly mark evidentiality in any language? The answer to this question requires an examination of evidential use in all languages, not just those which have traditionally been considered to have ‘linguistic evidential systems.’ This is not to say that there are not important differences between languages with grammaticalized evidential systems, like Cuzco Quechua, and those without, like English. However, I argue that the differences between such languages is not so fundamental that they must be studied independently of each other. Therefore, I consider ‘evidentiality’ to include any system by which speakers express their evidence for the propositions they make and ‘evidentials’ to be the individual markers within those systems.   1.1.2 What do evidentials encode?   Evidentials have traditionally been analysed as marking the type of evidence a speaker has for the propositions they express. The ways in which different evidence types are marked, however, varies from language to language. A comprehensive survey of grammatical evidential languages is set out in Aikhenvald (2004). As Faller (2002) points out, cross-linguistic studies have identified three main categories of information types for which grammatical marking is attested: direct access, reports, and reasoning.  The observation that evidentials tend to mark direct sources of information versus indirect sources like reports and reasoning is reflected in Willett’s (1988) taxonomy of linguistic evidentials as follows.      14 (1.6) Willett’s (1988) classification of evidentials    Direct           Indirect        |         /          \   Attested  Reporting  Inferring        |        |        |   Visual   Second-hand  Results   Auditory  Third-hand  Reasoning   Other sensory  Folklore    Willett’s taxonomy of evidentials is a categorization of different evidence types; that is, the distinction between direct and indirect evidence is based on the nature of the evidence source. As noted above, English has not traditionally been classified as an evidential language in this type of framework. Von Fintel & Gillies (2010), however, have argued that English must contributes an evidential meaning; specifically, must entails the truth of the sentence in which it appears and signals that the speaker reached their conclusion about the prejacent through indirect means.4 To that end, they claim that must encodes indirect evidential semantics and is, therefore, a marker of evidentiality. However, they make no claims about whether the existence of a single evidential marker requires the existence of a full evidential system or what that system might look like.  Typologically, an evidential system which distinguishes between inference on one hand and ‘everything else’ on the other is unattested in the world’s languages. In her survey, Aikhenvald (2004) reports that the attested two-way evidential distinctions are of the following types:                                                     4 The term ‘prejacent’ refers to the bare proposition embedded under the evidential.   15 (1.7) Attested two-way evidential distinctions  A1: Firsthand and Non-firsthand; A2:  Non-firsthand versus ‘everything else’; A3: Reported (or ‘hearsay’) versus ‘everything else’; A4: Sensory evidence and Reported (or ‘hearsay’); A5: Auditory (acquired through hearing) versus ‘everything else’.  (Aikhenvald, 2004: 25)  At first blush, must does not correspond neatly to one of Aikhenvald’s two-way distinctions. It does not mark “non-firsthand” so as to fall within A1 or A2 because it is infelicitous with reported information.5 Aikhenvald describes ‘firsthand’ systems as those which mark that the speaker has appropriate sensory evidence for their proposition. The non-firsthand marker, on the other hand, has a variety of meanings but generally includes reports, inference, and direct participation in the event in circumstances where the speaker had no control (at p. 24). Thus, if von Fintel & Gillies are correct in analyzing must as a marker of indirect evidence, then either English is typologically unique or must is unlikely to be the only evidential marker in the system.  More recently, Matthewson (2015a, 2015b) has analysed English must as a marker of indirectness, but she argues that must’s evidential contribution is with respect to evidence strength, not evidence type. Unlike traditional analyses, which explicitly limit evidentiality to encoding information about the kind of evidence a speaker has (see, for example, Aikhenvald, 2004: 3), Matthewson proposes that evidentials encode direct/indirect values on three possible dimensions of meaning: evidence type, evidence location, and evidence strength.                                                     5 Although see section 3.3.2 for further discussion about must and reported information.   16 (1.8) 1. Evidence type:  whether the evidence is visual, sensory, reported, etc.  2. Evidence location:  whether the speaker witnessed the event itself or merely  some of its results.  3. Evidence strength: the trustworthiness/reliability of the evidence.  (Matthewson, 2015a: (21))  Matthewson argues that each of the three dimensions has direct and indirect values and that evidential morphemes may be semantically complex, encoding information about one, two, or all three dimensions.   The evidence strength dimension encodes information about the speaker’s judgment about the trustworthiness or reliability of the evidence (Matthewson, 2015a: 9). She notes that speakers’ judgments about evidence strength correlate closely with, but are distinct from, the notion of speaker certainty. Where an evidential encodes a direct value for evidence strength, then it will be licensed when the speaker has the best possible evidence for the proposition they assert. Conversely, an evidential that encodes indirect evidence strength requires that the speaker not have the best possible evidence for their proposition.   DIRECT INDIRECT EVIDENCE TYPE e.g., sensory     e.g., inference, report EVIDENCE LOCATION event itself results/precursors EVIDENCE STRENGTH best not best TABLE 1.1: Matthewson’s evidential specifications Matthewson’s multi-dimensional approach to evidentiality is intended to account for evidentials whose semantic contribution does not fit nicely within an evidence type framework. She argues that von Fintel & Gillies’ analysis of must is such an example; she shows that, as a marker of indirect evidence type, von Fintel & Gillies’ analysis of must would be problematic.    17  1.1.3 English must and the requirement for ‘not best’ evidence According to von Fintel & Gillies (2010), must provides an evidential contribution in English. They argue that must p is infelicitous when the speaker’s information source is sensory evidence, trustworthy reports, or general knowledge.   (1.9) SENSORY EVIDENCE  a. Context: The speaker hears people playing Tchaikovsky   # They must be playing Tchaikovsky.   b. Context: The speaker smells a good smell   # Something must smell good.   c. Context: The speaker tastes something good   # Something must taste good.   d. Context: The speaker feels the coat is wet   # The coat must be wet.  (Matthewson, 2015a: (10)-(13))  (1.10) TRUSTWORTHY REPORTS  a. Context: Belinda, Bob’s wife, told the speaker that Bob is home. Belinda is  a very reliable source. The speaker now tells someone else:    # Bob must be home.  (Matthewson, 2015a: (17))  (1.11) GENERAL KNOWLEDGE  a. Context: It is general knowledge that World War II ended in 1945    # World War II must have ended in 1945.  (Matthewson, 2015b: (42))  18 Matthewson observed that, from a pure evidence type analysis, the three types of information which render must infelicitous are unusual because they do not fall neatly within a particular evidence type cohort.   (1.12) Evidence types that disallow must    Direct           Indirect        |         /          \   Attested  Reporting  Inferring        |        |        |   Visual   Second-hand  Results   Auditory  Third-hand  Reasoning   Other sensory Folklore   (adapted from Willett, 1988)  If von Fintel & Gillies are correct that must marks indirect evidentiality, then we expect sensory evidence to disallow must because such evidence is ‘direct.’ The fact that ‘indirect’ evidence types like trustworthy reports and general knowledge also disallow must, however, is not easily accounted for under a traditional evidence type analysis. Nonetheless, all three must count as direct evidence in order for von Fintel & Gillies’ analysis to work.   On an evidence strength analysis, however, indirectness cares only about whether the speaker’s evidence for p is ‘best’ or ‘not best’ in the particular circumstances. If must marks that the speaker’s evidence is ‘not best,’ then all must cares about is the trustworthiness or reliability of the evidence (Matthewson, 2015b: 18). If the speaker’s evidence for p is the best available in the context, must will be infelicitous regardless of evidence type; where the speaker’s evidence for p is not the best available, must will be felicitous.   1.1.4  Cuzco Quechua -mi and the requirement for best evidence  Matthewson (2015a) observes that, by requiring that a speaker’s evidence for p be ‘not best,’ English must contributes the opposite evidential meaning to Cuzco Quechua -mi,  19 which Faller (2002) analyses as a marker of best possible grounds. Faller argues that -mi is licensed where the speaker has the best possible evidence for the proposition they are asserting. She shows that previous attempts to analyse -mi as a marker of direct evidence type were problematic for the same reason Matthewson observes for must; namely, because the evidence which licenses -mi also includes information types that are traditionally considered indirect.  Faller shows that -mi is licensed where the speaker has direct access to the event, but also where the speaker knows the information by way of reliable report or by common knowledge; as Matthewson observes, -mi is licensed in precisely the contexts which disallow English must.   (1.13) DIRECT ACCESS  a. Pilar-qa t’anta-ta-n mikhu-rqua-n.   Pilar-TOP bread-ACC-mi eat-PST1-3    p =  ‘Pilar ate bread.’6     EV =  speaker saw that p b. Warmi-taq-mi  tutal-ta  qitipiyayu-n.   woman-CONTR-mi  total-ACC jump.around-3    p =  ‘And the woman totally jumps around.’     EV =  speaker saw that p (Faller, 2002: (10(a)-(b)) REPORTED  c. Lima-ta-n viaja-n.     Lima-ACC-mi travel-3    p =  ‘She travelled to Lima.’     EV =  speaker was told by her (= speaker’s sister) that p (Faller, 2002: (10(c))                                                  6 In the data, p indicates the prejacent and EV indicates speaker’s evidence for the prejacent.  20  COMMON KNOWLEDGE  d. Yunka-pi-n  k’usillu-kuna-qa ka-n.   Rainforest-LOC-mi  monkey-PL-TOP be-3    p =  ‘In the rainforest, there are monkeys.’     EV =  p is common, cultural knowledge (Faller, 2002: (11)) Faller shows that -mi is licensed by direct, sensory evidence, as in (1.13a) and (1.13b) above. However, like with must, the difficulty with analyzing -mi as a direct evidential is that it is also licensed by a certain subset of indirect evidence types as well; namely, reliable reports (in (1.13c)) and general knowledge (in (1.13d)).   (1.14) Evidence types licensed by -mi    Direct           Indirect        |         /          \   Attested  Reporting  Inferring        |        |        |   Visual   Second-hand  Results   Auditory  Third-hand  Reasoning   Other sensory Folklore    Faller analyses -mi not as marking direct or indirect evidence type, but as marking that the speaker has the best possible evidence for their proposition.   For Faller, the determination of whether a speaker has the best possible grounds for asserting a proposition is based first on the distinction between observable and non-observable events. Concrete events such as eating, winning, being tall etc. are capable of being observed and, as a result, direct observation is the best possible evidence of the event. By contrast, internal states or thoughts of other people are not directly observable and, therefore, reports of those events are the best possible evidence. What counts as having the  21 best possible evidence for a particular event is largely decided on a case by case basis (Faller, 2002).   Faller’s data make it clear, however, that for the purposes of licensing -mi, the (non-) observability of the event itself is not the determining factor. If that were the case, then an assertion about any concrete event could only be licensed by -mi if the speaker directly observed the event; this is not borne out in the data. Recall (1.13c) from above, repeated as (1.15) below. Faller points out that the subject of (1.15) is the speaker’s sister and that the speaker knows (1.15) because her sister told her so.    (1.15) Lima-ta-n viaja-n.    Lima-ACC-mi travel-3  p =  ‘She travelled to Lima.’    EV =  speaker was told by her (= speaker’s sister) that p (Faller, 2002: (10(c)) Travelling to Lima is a concrete event, observable for example to anyone travelling alongside the speaker’s sister. If the best possible evidence of an observable event is direct observation, then an observable event should never license -mi where the speaker did not witness the event itself and has only indirect evidence about the event. This is clearly not the case, given (1.15), where -mi is licensed by a report of the event. Therefore, the determination is based not on the inherent properties of the event type (observable versus non-observable) but on the relation between the speaker and the event; where the speaker is in a position to observe the event and does not, then -mi is infelicitous in the speaker’s subsequent report of the event. However, if the speaker is not in a position to observe the event, then -mi is felicitous provided the speaker has the next best type of evidence to support their claim. In (1.15), the fact that -mi is licensed where the speaker knows p by her sister’s report indicates that the  22 speaker was not in a position to directly observe p. This is so regardless of the fact that p is inherently observable.     For Faller, then, -mi is licensed where the speaker has the best possible grounds for p, with the best possible grounds being the best possible evidence available to the speaker in the circumstances, as determined on a case by case basis. Best possible grounds may be direct sensory information or indirect information in the form of reliable reports or general knowledge, depending on the context. Matthewson adopts Faller’s idea of best possible grounds although she expresses it as ‘best’ evidence (i.e., direct evidence strength). That is, for Matthewson, what -mi encodes is that the speaker has the best evidence available for p in the circumstances and this will generally include information obtained by sensory observation, trustworthy reports, and general knowledge. In other words, as Matthewson observes, the information sources that license –mi parallel those in which English must is infelicitous.   1.1.1.5 English discourse markers as evidentials In this dissertation, I show that actually provides an even closer cross-linguistic parallel to Quechua than must does because actually and -mi both encode ‘best’ evidence whereas must encodes its opposite. That is, I argue that actually is licensed in the same contexts as Cuzco Quechua -mi and is explicitly not licensed in the contexts where English must is felicitous.7 To this end, actually fills the gap between -mi on the one hand and must on the other. In chapter 3, I give the empirical data to show that actually is licensed where                                                  7 In chapter 3, I show that there is one difference between actually and -mi. Namely, actually can occur where the speaker’s evidence for p is the fact that they performed p, for example, I actually drank the last of the milk. Faller (2002) argues that -mi is infelicitous with speaker performance. However, as I show, this distinction is not problematic for my analysis.   23 the speaker has sensory information for p, a trustworthy report of p, or general knowledge of p. Moreover, I observe that actually is unattested in the police interview transcripts where the speaker has only indirect evidence for p; that is, actually is unattested in the evidential contexts in which English must is felicitous.  I further show that, while actually and must encode evidence strength, apparently contributes evidential information on the evidence type dimension of meaning. I show that apparently p marks that the speaker’s evidence type for p is indirect in the sense of Willett (1988). Finally, I show that supposedly encodes both evidence type and evidence strength, marking that the speaker’s information source for p is reported and that the reported evidence is ‘not best’ in the circumstances.   1.2 Police interviews as a source of evidentiality My analyses of actually, apparently, and supposedly are based on the empirical evidence showing how speakers express evidence in a series of police interview transcripts generated in the course of the investigation into the disappearance and murder of Caylee Anthony. Specifically, I reviewed English speakers’ use of actually, apparently, and supposedly in the proffering of evidence during the police interview process. As I will motivate in the next sections, police interviews are a unique discourse environment where speakers are expected to tell not just what they know but how they know it. Therefore, police interviews are an excellent source of non-contrived data reflecting evidential use.    1.2.1 The role of ‘facts’ in the legal process  24 Interviews conducted during the course of a criminal investigation are the major fact-finding method available to investigators (Dando et al., 2016; Heydon, 2005; Heaton-Armstrong et al., 2006; Heaton-Armstrong, et al., 1999; Milne & Bull, 1999). Information is the bloodstream of investigation, and “effective interviewing is the heart that pumps this bloodstream through the entire body of investigation” (Buckwater, 1983: 1). The goal of the witness interview is to obtain the information necessary to advance the criminal investigation and to establish the elements of the offence at trial (Walton, 2003). That is, the goal of the interview is to obtain the evidence necessary to determine the identity of the offender and, eventually, to prove their guilt in court. The ultimate goal of any criminal investigation is to establish guilt on the part of the perpetrator. However, all persons accused of committing a crime must, at law, be considered innocent until their guilt is proven in court.8 Unless the accused pleads guilty to the offence, a person accused of a crime will only be held accountable if guilt is established in the course of a criminal trial after the prosecutor has established all of the elements of the offence beyond a reasonable doubt. For example, Section 231 of the Canadian Criminal Code defines first degree murder as follows:  231(2) Murder is first degree when it is planned and deliberate.                                                  8 The right to be presumed innocent until proven guilty is enshrined in Section 11(d) of the Canadian Charter of Rights and Freedoms and follows from the 5th, 6th and 14th amendments of the Constitution of the United States of America, as decided by the U.S. Supreme Court in Coffin v. United States, 156 U.S. 432 (1895).   25 The prosecutor proves the offence by establishing facts on which a trier of fact9 may conclude three things beyond a reasonable doubt: that the accused caused a death, that the accused planned to cause the death, and that the death was deliberate.  In rendering a verdict, the trier of fact is only entitled to consider those facts which have been ‘proven’ in court through witness testimony; only that information which has been testified to by a witness and accepted as a ‘fact’ by the trier of fact can be considered in determining whether the offence has been proven.  Similarly, the trier of fact cannot rely on any physical evidence (for example, a document or weapon) unless and until a witness has first authenticated it by testifying as to its identity. Once a witness has testified to certain information or authenticated certain physical items, the witness’s testimony becomes evidence and can be considered for the purposes of rendering a verdict. The trier of fact decides the ‘facts’ based on witness evidence. Therefore, witness evidence is crucial in proving the elements of a criminal offence.    What evidence can be adduced in a criminal trial is governed by rules of admissibility, including the Rule Against Hearsay and the Best Evidence Rule. Both of these rules reflect the requirement that only the most authentic, reliable evidence may be used at trial to establish a suspect’s guilt. Generally speaking, the more direct the evidence, the more reliable the court will consider it to be. The Rule Against Hearsay prohibits a witness from offering as evidence an out-of-court statement to prove the contents of that statement.10 For                                                  9 In a jury trial, the jury is the trier of fact and the judge is the trier of law.  In a judge-alone trial, the judge is the trier of both fact and law.  10 Hearsay evidence may be admissible if it falls within one of the recognized exceptions to the Rule Against Hearsay or is shown to be admissible under the principled approach developed by the Supreme Court of Canada in R. v. Khelawon, 2006 SCC 57, R. v. Khan, [1990] 2 S.C.R. 531; R. v. Smith, [1992] 2 S.C.R. 915, and R. v. B.(K.G.), [1993] 1 S.C.R.  26 example, John’s statement that Jane told him she saw Sue break a window is hearsay if the statement is being offered to prove that Sue broke the window and is generally inadmissible. If the goal is to prove that Sue broke the window, then Jane must give that evidence herself. Similarly, the Best Evidence Rule requires that the original of any document must be produced in order to prove the contents of that document; a copy or facsimile will not generally suffice. The rules of evidence require that the most reliable, authentic version of the evidence must be presented to the court and, as long as that evidence is available, any less reliable evidence will be inadmissible. The investigator’s goal, therefore, in preparing the case for trial is not only to determine the evidence on which to establish the elements of the offence, but to establish which witnesses will be able to give the most authentic, reliable evidence, consistent with the Rule Against Hearsay and the Best Evidence Rule.    Given the function of the witness interview in the course of an investigation, transcripts of those interviews are predicted to be an excellent source of non-contrived discourse containing expressions of evidentiality. The job of the witness is to tell the interviewer what they know and how they know it. Not only does the witness make assertions about what information they believe to be true, they also indicate what evidential support they have for their assertions. Where the witness is not explicit about the source of their information, the investigator’s job is to ask the questions necessary to prompt the witness to provide those additional details (Buckwater, 1983: 4). Moreover, while the language generated is properly characterized as institutional discourse, it is not encumbered with the specific problems of suspect interrogation language, as discussed in section 1.2.4 below.                                                                                                                                                         740. The legal approach to the reliability of evidence will be discussed in greater detail in chapter 3.   27  1.2.2 The extent and limits of previous research on police interview language  Heydon (2005: 29) noted that the body of research involving police interviews can be roughly divided into two areas: those which result from an inquiry into questionable behaviour in the interview itself and those which investigate language in a legal setting as a linguistic theoretical exercise. Generally speaking, the research in the former category has focussed on linguistic issues arising in suspect interviews and the research in the second category has focussed on courtroom discourse; very few studies have attempted to analyse police interviews as being reflective of natural language use. This likely results from the fact that police interview data is not easily obtainable for the purpose of linguistic study.   Police interview data is not publically available and can generally only be accessed with the express cooperation of a police force (as was the case, for example, in Fadden (2008) and Heydon (2005)). Linguists who are retained to act in a professional capacity with respect to a dispute involving linguistic evidence arising out of police interviews will have access to the data necessary to prepare their report.  However, that data is necessarily problematic to begin with, since “some exceptional circumstances must exist for the case to proceed otherwise there would be no legal challenge to be addressed by linguistic analysis” (Heydon, 2005: 30). Therefore, such data is inherently problematic for the purpose of considering natural language use.   While courtroom discourse has the benefit of being more readily available to researchers, courtrooms being publically accessible, courtroom discourse is highly constrained in ways that non-courtroom language is not. Stygall (2012: 370) observes that  [o]rdinary conversational turn-taking, for example, disappears in a courtroom.  The attorneys control the topics of conversation. Topic choice for conversational partners  28 also disappears in a courtroom.  Attorneys also control the pacing of the questioning and how long the topic they selected stays on the conversational floor. Judges control what the attorneys can say and what the jury will hear. Witnesses cannot ask questions of their own to the attorneys or the judge. And witnesses cannot refuse to answer. These are not the ordinary discourse rules of everyday life.  Further, by the time the witness takes the stand in a criminal case, they may well have already expressed their information a number of times to a number of institutional actors. In most cases, they will first have been questioned by the police interviewer in the course of the criminal investigation and then by the lawyer who intends to call them as a witness in order to prepare them for trial.  Therefore, the language used to give that evidence in court is not reflective of unplanned, unconstrained language use. As a result of these limitations, research on courtroom discourse has focussed on specific issues arising out of how language is used in the courtroom (for example, Stygall, 2012; Komter & Malsch, 2012; Marder, 2012; Heffer, 2010; Felton Rosulek, 2010) or specific issues involving interpretation and translation of witness testimony (for example, Berk-Seligson, 2012; Hale, 2010).   1.2.3 Police interviews as institutional discourse  Like courtroom discourse, police interviews constitute institutional discourse and have been analysed as such, particularly with respect to suspect interviews (Fadden, 2008; Heydon, 2005). Institutional talk is characterized by three key features, as proposed by Drew and Heritage (1992) and as summarized in Heydon (2005):  (1) At least one of the participants in an institutional interaction displays an orientation towards some core goal conventionally associated with the institution. (2) There may be constraints on what one or all participants will treat as an allowable contribution to the interaction.  29 (3) Institutional talk may be produced within context-specific inferential frameworks. (Heydon, 2005: 37) On the basis of these factors, Heydon (2005) concludes that police interviews constitute institutional language.  She finds that police interviews share with other institutional interviews a basic turn-taking system consisting of sequences of questions and answers as well as a turn pre-allocation system whereby questions are allocated to interviewers and responses to interviewees (p. 38). Heydon’s research is specifically focussed on suspect interviews and, while witness interviews arise in the same institutional context, I argue the institutional goal of witness interviews is significantly different than that of suspect interviews, which renders them more useful as examples of natural language use.  There are usually three types of people who are interviewed in the course of a police investigation: the ‘suspect’ who investigators believe perpetrated the crime, the ‘victim’ who was wronged by the crime, and the ‘witness’ who may have information about the crime but who is neither a suspect nor a victim.11 The vast majority of research into police interviewing has involved suspects and victims; significantly less linguistic research has been done on the language of witnesses.  Buckwater (1983) argues that the goal of the interview is the same, regardless of who is being interviewed; namely, to obtain factual and relevant information. For him, the                                                  11 There is a fourth type of interviewee discussed in the literature, namely, the ‘vulnerable witness.’ Vulnerable witnesses comprise a subset of witnesses who are considered particularly vulnerable to the court process. Since 2002, in England and Wales, “vulnerable witnesses” include children under 17 years of age and any witness whose quality of evidence is likely to be diminished because they suffer from a mental disorder, a significant impairment of intelligence and social functioning, or a physical disability or disorder (Aldridge, 2007, 2010: 297). Vulnerable witnesses can be suspects, victims or witnesses. None of the witnesses interviewed in the Caylee Anthony investigation were flagged as vulnerable witnesses.    30 distinction between a suspect interview and a witness interview is in the attitude of the interviewee; an ‘interview’ as the questioning of a respondent who is “ready, willing and able to tell what he knows” and an ‘interrogation’ as the “more formal questioning of a suspect, or of a reluctant or hostile witness, or of anyone who is unwilling to discuss freely any information that he or she possesses” (Buckwater, 1983: 4). In other words, where the interviewee is cooperative, they are interviewed, and where they are uncooperative, they are interrogated, regardless of whether the interviewee is a witness or suspect.  Despite Buckwater’s view that the goals of interrogations and interviews align, Oxburgh et al. (2016) observe that there are fundamental differences between interrogations and interviews. They point out that the purpose of the investigative interview is to gather information fairly and impartially in order to secure the most reliable accounts of the crime from interviewees; the investigative interview is always intended to be non-accusatory (Oxburgh et al., 2016: 145; Oxburgh & Dando, 2011). The interrogation, on the other hand, involves active persuasion on the part of the interrogator where they are reasonably certain of a suspect’s guilt; the interrogation is always intended to be accusatory (Oxburgh et al., 2016: 146; Inbau et al., 2013). I use the terms ‘interrogation’ to describe the police questioning of suspects and ‘interview’ to describe the police questioning of non-suspect witnesses in order to reflect these institutional differences.     1.2.4 Police interrogations vs. police interviews  I argue that the fundamental difference between interviews and interrogations can have a significant impact on the nature of the institutional discourse. Both the institutional goals and the institutional ends differ significantly depending on whether the information is being sought in an interrogation or an interview. For an interrogation, the institutional goal is  31 to obtain information that shows the suspect to be guilty and the institutional end is a conviction (Heydon, 2004). On the other hand, for an interview, the institutional goal is to advance the investigation and the institutional end is to uncover the facts necessary to prove the elements of the offence at court.  The distinction between the institutional aims of interrogations and interviews is significant with respect to the potential legal repercussions facing the person being questioned. A witness being interviewed is not suspected of the crime and, as a result, does not generally face legal repercussions as a result of their participation in the interview process.12 The suspect, on the other hand, faces serious legal repercussions if their guilt is established through the interrogation or otherwise.13 As such, the suspect has significantly more at stake in an interrogation than a witness does in an interview.  Because there is more at stake in an interrogation than an interview, the participants’ goals in an interrogation are more likely to diverge than they are in an interview. Those diverging participant goals are potentially more likely to result in deceptive or manipulative language being used in an interrogation than in an interview. Since guilt is best established through a voluntary confession, given spontaneously by the suspect without any prompting or leading by the interviewers (Shuy, 1998), the interrogator’s goal is to uncover information regarding the suspect in order to prompt a confession. On the other hand, the suspect’s goal is to conceal that information or otherwise provide information that is not helpful in achieving that institutional goal (Fadden, 2008: 8-9). As a result of these diverging                                                  12 A witness is entitled to rely on Section 13 of the Canadian Charter of Rights and Freedoms or the 5th Amendment of the U.S. Constitution to refuse to answer a question which tends to incriminate them. This is colloquially referred to as “pleading the 5th”.   13 In certain jurisdictions of the United States, including Florida, where the Caylee Anthony investigation occurred, suspects may face the death penalty if convicted of a serious crime.   32 participant goals, a suspect may choose to remain silent (generating no linguistic data at all), or they may choose to mislead or deceive the interviewer.    Even where the suspect is speaking truthfully, the language used by the interrogator may be deceptive or misleading depending on the jurisdiction in which the interview takes place. Different jurisdictions authorize different types of interview methods. In Canada and elsewhere in the Commonwealth, law enforcement have adopted investigatory practices consistent with the P.E.A.C.E. (Planning & Preparation, Engage & Explain, Account, Closure, Evaluation) model of interviewing, which eliminates, at least in theory, any differences in the way investigators conduct interviews as opposed to interrogations. Under the P.E.A.C.E. model, interviewers are prohibited from using unethical behaviours like deception, making threats or promises, or using oppressive or coercive tactics in an effort to secure a confession (Oxburgh et al, 2016; Leo, 2008). To that end, the P.E.A.C.E. model is significantly different than the Reid Technique employed widely in the United States which permits such behaviours (Inbau et al., 2013; Shuy, 1998).14 Although investigators are not required to do so, the Reid Technique permits them to use manipulative behaviours to pressure a subject into compliance (Oxburgh et al, 2016: 146; Leo, 2008).15 Thus, the Reid Technique explicitly retains the distinction between an interview and an interrogation. For suspect interviews conducted using the Reid Technique, it may be difficult to determine                                                  14 The investigation into the disappearance and murder of Caylee Anthony took place in Florida. As such, the investigators had the Reid Technique available to them as a permissible interview method.   15 I leave to another forum the discussion of the ethical issues that arise in jurisdictions which both allow manipulative and deceitful interview tactics designed to coerce confessions and also have the capacity to impose the death penalty as punishment on those whose guilt is established through a confession or otherwise.    33 when the interviewers are speaking honestly and when they are using deception to obtain a confession. Thus, neither the language generated by the interrogator nor by the suspect can be relied on to be trustworthy in the context of an interrogation.16 While this will not necessarily affect the results from a natural language use perspective, it adds an additional variable to the discourse.   The difference in institutional function between witnesses and suspects translates into a difference in language use in the interviewing of witnesses as opposed to the interrogating of suspects. The job of the witness is to tell the interviewer what they know and how they know it.  Thus, not only does the witness make assertions about what information is true, they also indicate what evidential support they have for those assertions. While, of course, speakers may use deceitful or manipulative language in witness interviews, the institutional goal of the interview makes it less of a certainty than in interrogations. For this reason, I limit my analysis to the data generated in the witness interviews in the investigation into the disappearance and murder of Caylee Anthony and not the interrogations of the primary suspect, Casey Anthony.   As I show in chapter 2, the transcripts show that witnesses use actually, apparently, and supposedly robustly in giving their information throughout the interview process. However before moving on to discuss those empirical findings, I summarize previous treatments of these three discourse markers in the literature.   1.3 Actually in the literature                                                  16 See Oxburgh et al. (2016), Leo (2008) and Milne & Bull (1999), for example, for further critique of the Reid Technique.   34 The function of actually and related discourse markers like in fact have received significant attention in the literature, both in terms of their structural role (for example, Aijmer, 1986; Watts, 1988; Lenk, 1998; Tognini-Bonelli, 1993; Taglicht, 2001) and the interactional role they play in discourse (for example, Smith & Jucker, 2000; Clift, 2001; Fox, 2001).    However, not since Watts (1988) have researchers attempted a unified analysis of actually; recent discussion has focused on its multiple structural forms and/or multiple pragmatic functions.   Most if not all of the literature reports that actually represents two distinct grammatical objects. Greenbaum (1969) initially distinguished two types of actually on the basis of examples like the following.   (1.16) A: How was the bus ride? – B: Actually, we went by train.  (1.17) This isn’t (just) hearsay, you know; I [actually saw it with my own eyes].  (from Taglicht, 2001: (1) and (2))  Greenbaum defined the actually in (1.16) as an “attitudinal adjunct” and that in (1.17) as an “intensifying adjunct”.  While the terminology has changed over time (see, for example, Quirk et al., 1985; Taglicht, 2001), the empirical distinction has been substantially unchallenged in the literature. Subsequent research has attempted to motivate the distinction on the basis of both syntax and pragmatic function. Where it is important to distinguish between the two in this dissertation, I refer to the actually in (1.16) as ‘parenthetical actually’ and the actually as ‘clause-internal actually’. 1.3.1 Syntactic analyses of actually  The most comprehensive syntactic account of the distinction between the two actuallys observed in Greenbaum (1969) is that proposed in Taglicht (2001).  He argues that  35 parenthetical actually in (1.16)17 can only appear as an adjunct to the clause or a parenthetical and not as an immediate constituent to the VP.    (1.18)18   you’ve |gone ‘over the ‘M thr/ee− (a) you |actually go\ under it (CEC 1.11, TU 1059)  (b) \/actually you go \under it  (c) you go \under it/ actually (Taglicht, 2001: (4))  Using a Head-driven Phrase Structure Grammar framework, Taglicht argues that in (1.18a) actually is a parenthetical interposed between the Subject node (you) and its Head node (the VP go under it).  In (1.18b) and (1.18c), actually is an adjunct node, whose Head is the S node you go under it (2001: 4).  Conversely, clause-internal actually given in (1.17)19 will always appear as the initial element of a constituent, never as a parenthetical or an adjunct to the clause, as in the following.  (1.19) This isn’t just hearsay; I [actually saw it with my own eyes]. (VP) (1.20) All these substances are harmful, and some of them are [actually poisonous]. (AP) (1.21) All these substances are harmful, and one is [actually a poison]. (NP) (1.22) Of course what he’s doing is scandalous, but is it [actually against the law]? (PP) (1.23) The ointment was even found to contain some [actually poisonous] substances. (AP)  (Taglicht, 2001: (8)-(12))  In part, Taglicht motivates his syntactic analysis by observing how each actually interacts with negation; notably, he argues that parenthetical actually in (1.16) cannot fall within the scope of negation regardless of where it occurs in the linear structure, as shown in (1.24)                                                  17 Taglicht refers to this actually variously as actually1, “mild actually”, and “marginal actually”.  18 The notation used in (1.18) reflects that commonly used in Conversation Analysis (Drew & Heritage, 1992). See, for example, Grant et al. (2016) for a discussion of the use of Conversation Analysis in the analysis of police interviewing.   19 Taglight refers to this as actually2 or “strong actually”.  36  (1.24) (a) Actually, no one had any objection.  (b) No one had any objection, actually.  (c) No one, actually, had any objection. (Taglicht, 2001: (7))  On the other hand, Taglicht argues clause-internal actually always falls within the scope of negation.  He shows the distinction as follows.  (1.25) (a) He should have cut out the parts not essential to his argument actually. (b) He should have cut out the parts [not [actually essential to his argument]]  (Taglicht, 2001: (17))  Taglicht was not alone in attempting to distinguish the two observed meanings of actually on the basis of syntax.  Aijmer (1986), Lenk (1998) and Tognini-Bonelli (1993) provide structural accounts of actually. However, rather than relying on traditional syntactic tests, they motivate their analyses by appealing to the different pragmatic functions performed by actually, given its syntactic position; these will be discussed in 1.3.2.    1.3.2 Pragmatic function analyses of actually Though the distinction between parenthetical actually and clause-internal actually is widely recognized, the former has received significantly more attention in the literature.  Nonetheless, there have been some attempts to analyze the meaning or function of clause-internal actually.  Notably, Watts (1988) and Lenk (1998) argue that actually, basically, and really share a core fundamental meaning, namely “something like ‘genuine, real, basic’” (Watts, 1988: 251) or “in fact” (Lenk, 1998: 157). Oh (2000: 260) finds that, when it takes local scope, clause-internal actually is a clause-emphasizer which has a reinforcing effect on the truth value of the clause. She also observes a global use of clause-internal actually which she argues functions to contradict prior expectation.    37 In its earliest treatments, parenthetical actually was thought to mark “contrastive avowal”; that is, both actually and in fact were argued to introduce a proposition contrary to expectation (Halliday & Hasan, 1976: 253).  Aijmer (1986) argues that clause-final actually “expresses the speaker’s incredulity” about what was previously said in the discourse (1986: 126).  Lenk posits three functions for parenthetical actually: as a marker of personal opinion, a marker of correction or contradiction, and a marker of topic shift.  However, she notes that, at their core, all three express “that the following will be slightly (moderately/definitely) different from the expected normal course of the conversation” (1998: 160).  Oh (2000: 266) similarly proposes a common core meaning of parenthetical actually, namely “unexpectedness”, which she argues occurs when actually takes global scope, regardless of whether it occurs initially, finally, or medially.  Martin (1992) argues that counter-expectation is not part of actually’s inherent meaning.  Rather, he argues actually is a reformulation marker, which implies something in the original formulation was not quite right.  The concessive relation, he suggests, arises explicitly only when actually co-occurs with a contrastive conjunction like but. Williams (1996) argues that the same holds for in fact.  Previous studies of parenthetical actually have been concerned with assigning it a meaning or function at the speech-act level (Clift, 2001: 247-48), rather than the truth-conditional level.  Given the number of empirical studies devoted to the various pragmatic functions actually can perform, it is perhaps unsurprising that a significant number of different functions have been observed. As Waters (2011: 115)) observed, previous studies of actually have described its functions as including linking utterances (Oh, 2000: 254; Traugott & Dasher, 2002: 171), marking self-correction (Tognini-Bonelli, 1993: 209; Lenk, 1998:  38 156), reopening a previously closed topic (Clift 2001: 267), emphasizing (Quirk et al., 1998: 1415), and signalling disagreement (Oh, 2002: 256; Taglicht, 2001: 2).  A handful of studies have discussed actually as a possible marker of evidentiality. These studies, however, consider ‘evidentiality’ as marking speaker attitude. That is, in these studies, actually is analysed as marking a speaker’s attitude toward the information they have for their proposition rather than their evidence for the proposition. Chafe (1986) claims that actually signals “a fact beyond what one might have expected” (1986: 271), given the discourse.  Similarly, Biber & Finegan (1988) note actually expresses “greater certainty [or] truth than expected” (1988: 8).  Although not specifically couched as an evidential argument, Tognini-Bonelli (1993) distinguishes actual/actually from real on the basis that real contrasts semantic units whereas actual/actually contrast the speaker’s perspective.    (1.26) … the summit here last week concluded that deep emission cuts are both possible and urgently needed.  So the real problems still lie ahead … (BBC Corpus)  (1.27) … women’s magazines that matter get grottier and grottier if Claire will forgive me saying so.  The actual problems that people are allowed to ask advice about have become disgusting beyond belief … (Spoken Corpus)  (Tognini-Bonelli, 1993: (3)-(4))  Tognini-Bonelli claims that, in (1.26), real has a selective role; it sets the “real problems” against “other problems”, which are implicitly labeled as less relevant or less important as a result (1993: 196).  Conversely, in (1.27), actual emphasizes the meaning already realized by the noun, rather than modify the meaning of problems by selecting one type of problem  (1993: 196-197).   The idea that actually marks propositional attitudes was further developed in Smith & Jucker (2000), which focused specifically on actually’s function in the negotiation of  39 common ground.  They argue that actually is used to shift ground; it signals a deviation from what is expected, given the discourse, with respect to the speaker’s commitment to, the newsworthiness of or evaluation of the claim being proffered. Smith & Jucker (2000) limit their inquiry to actually’s role in conveying meaning beyond mere factual information; they specifically focus on its function to establish the speaker’s certainty regarding the factual information, their estimation of its newsworthiness, and their effective evaluation of the information (at p. 208). I will argue, however, that the latter derives from the former; it is because actually requires the speaker to have ‘best’ evidence for the proposition that permits the speaker to evaluate it, estimate its newsworthiness, and express any sort of certainty with respect to it.   1.4 Apparently in the literature  Apparently has received less scholarly attention than actually. Indeed, even in discussions of English evidentiality which specifically reference apparently, it rarely garners more than a simple mention. For example, Fox (2001) includes apparently as part of the list constituting the evidential markers of conversational English on the basis of the following.   (1.28)  apparently: Apparently Marcia went shopping for all those things.  (Fox, 2001: 171)  However, beyond that example, apparently is not otherwise mentioned or discussed. Similarly, Gisborne & Holmes (2007) simply observe that English contains lexical evidentials like apparently, which they argue incorporates the evidential sense of appearance but about which they say little further. Biber et al. (1999: 855, 871) classify apparently as an adverbial which marks a speaker’s knowledge source while Huddleston & Pullum (2002: 768-769) classify it as a modal adjunct concerned with the speaker’s commitment to the truth  40 of the adjoining proposition. Mushin (2001) analyses apparently as a reportative, though Tantucci (2013) observes it is felicitous in non-reportative environments as well.   With respect to the type of evidence apparently encodes, it has been described as both a marker of hearsay (Chafe 1986) and of inference (Hoye, 2008). Kamio (1997: 24) suggests that apparently functions to mitigate the directness of an expression by putting distance between the speaker and their proposition. Thus, for Kamio, apparently marks indirectness although no analysis is provided to explain this function.   Although sparse, the literature suggests that apparently is generally understood as marking that the speaker’s evidence for the proposition they express is indirect (either reportative or inferential) and that it functions to limit the speaker’s liability for the truth of that information.  In chapters 3 and 5, I motivate an analysis of apparently whereby it encodes that the speaker’s evidence for the prejacent is indirect on the evidence type dimension of meaning and I argue that the function attributed to it by Kamio falls out from this analysis.    1.5 Supposedly in the literature  As a possible evidential, supposedly has received even less scholarly attention than apparently, although Chafe (1986) and Hoye (2008) have both analysed it as one of English’s indirect hearsay devices. Chafe further notes that supposedly appears to suggest the hearsay it introduces is unreliable.   (1.29) I think it’s supposed to be the most expensive place in Europe to live.  (Chafe, 1986: (41s)) I argue, following Chafe, that supposedly is, indeed, a reportative marker. However, while  41 Chafe suggests that supposedly indicates the speaker distrusts the information introduced, I show that it can also be used where the speaker is agnostic about the truth of the information but where the source of the information is untrustworthy.  1.6 Overview of dissertation  This chapter presented an introduction to the literature on evidentiality. I showed that the diagnostic by which Aikhenvald (2004) and others have classified grammatical evidential languages, i.e., grammatical obligatoriness, is flawed. I follow Mushin (2013) and others and propose that our understanding of how speakers express evidentiality is enriched through the study of languages like English which have not traditionally been characterized as containing linguistic evidential systems. I motivated my decision to use language generated in the course of witness interviews to study speakers’ expressions of evidentiality. I showed that the institutional goal of a witness interview is to develop the informational background on which to advance the investigation. To that end, speakers in witness interviews are expected to tell interviewers not just what they know, but how they know it. Finally, I gave an overview of the literature on actually, apparently, and supposedly, the English expressions which will form the basis of this study.   In chapter 2, I set out how I obtained and analysed the data for this study. I present a brief background on the investigation into the disappearance and murder of Caylee Anthony; it is in the context of that investigation that all of the witnesses interviews I rely on were generated. I describe the speakers and show that, to a large extent, they come from a connected social network and a constituent speech community. I show that the speakers whose language was reviewed consistently used actually, apparently, and supposedly in the expression of their evidence.  42  My empirical findings and analyses are set out in chapter 3. I show that the evidential contexts in which actually occurs largely parallel those in which the evidential -mi in Cuzco Quechua is felicitous and those in which English must is infelicitous. I draw from previous analyses of -mi and must (in particular, Faller, 2002; von Fintel & Gillies, 2010; and Matthewson, 2015a, 2015b) and argue that speakers use actually where they have the ‘best’ evidence. Thus, I argue that actually makes an evidential contribution on the evidence strength dimension of meaning, marking ‘best’ evidence for their proposition. I show that apparently encodes an indirect value on the evidence type dimension of meaning. Supposedly encodes indirectness on both the evidence type and evidence strength dimensions of meaning. Specifically, supposedly marks that the speaker’s evidence is reported and that it is ‘not best’ in the circumstances. Finally, I show that legal concepts of ‘best’ evidence largely mirror the evidence contexts that permit linguistic ‘best’ evidence in both English and Cuzco Quechua.  Having shown what evidentials mean in chapter 3, I show what they do in chapters 4 and 5. From an informational perspective, speakers use actually, apparently, and supposedly to negotiate the common ground of discourse. In chapter 4, I show that speakers use actually to both propose propositional content for inclusion in the common ground and also to advocate for the inclusion of that propositional content. Speakers use actually to accomplish this in discourse contexts where a bare assertion does not suffice. Moreover, because actually marks that the witness has best evidence for their proposition, they can use actually to accomplish delicate discourse actions like disagreeing, challenging, and persuading, while maintaining the politeness norms consistent with their limited institutional authority. I show a second use of actually that operates at a discourse level to control the discourse topic;  43 specifically, speakers use actually to introduce or change topics, resist a topic shift, or embark on a discourse-relevant digression. I show that the topic control function is only felicitous with parenthetical actually and present some syntactic arguments that may explain that sensitivity.   In chapter 5, I show that speakers use apparently and supposedly to put information up for discussion without proposing it for inclusion in the common ground. In terms of Goffman’s (1974, 1979) participant roles framework, apparently and supposedly explicitly mark the speaker as a mere animator of the information and not its author or principal. By so doing, they allow the speaker to distance themself from the information and to share information that they either know is untrue or where they are agnostic as to its veracity.   Chapter 6 focusses on the use of English evidentials in questions. This discussion is limited to actually because there were no attested instances of apparently or supposedly in questions in the police interview transcripts. All of the questions in the police interview transcripts were asked by interviewers and none by witnesses, consistent with Heydon’s (2004) observations about the pre-allocation of discourse roles in police interviews. I show that actually-questions are addressee-oriented and, following Murray (2010) cannot be used to introduce new evidence into the discourse. Rather, interviewers use them to encourage the witnesses to commit to information they previously gave with a bare assertion. The use of an actually-question puts the witness on notice that the information they have proposed for inclusion in the common ground cannot be accepted without explanation. As such, actually-questions encourage the witness to either disclose the source of their information or expressly align as its principal and/or author.   I summarize the dissertation in Chapter 7 and discuss where this research fits within  44 the existing literature. I elaborate on the consequences of my research methodology and explain why neither an elicitation-based approach nor a traditional corpus-based approach would likely have been successful in achieving this research goal. I summarize issues that remain for future research and conclude.    45  “That, that, that’s ah, recording is gonna capture  more content that you’re gonna be able… then, then  you’ll be able to verbally articulate to me.”  Cpl. EE (JJ1.8:7-9)    2. THE POLICE INTERVIEW TRANSCRIPTS  The data analysed in this study was obtained from transcripts of police interviews conducted in the course of the investigation into the disappearance and murder of Caylee Marie Anthony in Orlando, Florida, USA. I chose to analyse the police interview transcripts from this particular investigation for purely practical reasons; because of the high profile nature of Casey Anthony’s trial, a significant number of documents, including the transcripts, were made accessible to the public. This is highly unusual. As such, the availability of the witness interview transcripts created in the course of the Caylee Anthony investigation provided a rare opportunity to analyse police interview discourse to determine how witnesses express their evidence.  This chapter is organized as follows. First, I provide a brief factual background to the investigation into the disappearance and murder of Caylee Anthony. I discuss some preliminary issues with the use of transcripts generally and, specifically, the transcripts in the present case. I then discuss the interviews themselves and set out the sociolinguistic information available with respect to the various speakers whose language is recorded in the interview transcripts. I go on to describe how I collected and analysed the data from those transcripts. Finally, I show that actually, apparently, and supposedly occurred regularly and consistently in the data, which is why I chose to focus my study on those markers specifically.    46 2.1 The disappearance and murder of Caylee Anthony On July 15, 2008, Cindy Anthony telephoned 9-1-1 to report her two-year old granddaughter, Caylee Marie Anthony, missing. Caylee Anthony and her mother, Casey, lived with Casey’s parents, Cindy and George Anthony, in Orlando, Florida, though at the time Caylee was reported missing, her grandmother had not seen her for over 30 days. During this period, Casey Anthony offered various accounts of Caylee’s whereabouts but ultimately confessed to her mother on July 15, 2008 that she had not seen Caylee for weeks. Despite being concerned about certain inaccuracies in Casey Anthony’s statement and despite their suspicions that Casey Anthony had murdered her daughter, the police investigated Caylee Anthony’s disappearance as a missing person case for a number of months.  However, on October 14, 2008, before Caylee’s remains were discovered, Casey Anthony was indicted by a grand jury on charges of first-degree murder, aggravated child abuse, aggravated manslaughter of a child, and four counts of providing false information to the police.  Meanwhile, on August 11, 12, and 13, 2008, a civic employee named Roy Kronk called the police to report a suspicious object located in a wooded area near the Anthony home. Those reports were not investigated in any serious way by the police. On December 11, 2008, Kronk reported the suspicious object again and, after a more thorough investigation, the object was found to be a garbage bag containing Caylee’s skeletal remains.  2.2 The Caylee Anthony Investigation The investigation took place between July 2008 and May 2011 in Orange County, Florida. Various preliminary motions were argued before the court between January 2009  47 and May 2011 in advance of Casey Anthony’s murder trial. Jury selection was ultimately completed in May of 2011 and the trial began on May 24, 2011.20  During the course of the investigation, the police interviewed a number of witnesses. Those interviews were transcribed by the police department for their records.21 Each interview was audio-recorded, typed by a police transcriber, then at least theoretically reviewed for accuracy against the initial recording. It is unclear how many transcribers were employed in creating the transcripts although from the different transcription conventions used, it appears there were multiple. The transcripts resulting from those interviews comprise the data on which this study is based.   2.2.1 The nature of language in transcripts The language represented in the police interview transcripts that form the data for this study cannot properly be characterized as either written or spoken; rather, it is the product of converting spoken language into written language. The problem with such a process is that spoken language and written language are different media with different properties (see, for example, Biber, 1988; Halliday, 1989). Gibbons (2003: 28) observes that “it is virtually impossible to accurately record in a single visual representation all the sound detail of speech, including pitch/intonation, breathiness, voice quality, accent, pausing and pace” and, were the transcriber to attempt to do so, the transcript would be virtually unreadable (see also Ochs, 1979).                                                   20 On July 5, 2011, Casey Anthony was acquitted of first-degree murder, aggravated child abuse, and aggravated manslaughter of a child. She was found guilty of four misdemeanor counts of providing false information to a law enforcement officer.   21 All of the transcripts were obtained, in .pdf format, from the (now defunct) website www.docstoc.com. I was unable to find any additional transcripts through any other website or internet search engine.  48 Gibbons further notes that the usefulness of transcripts is measured against two competing and incompatible criteria: readability and accuracy. On one hand, the transcript must be readable to be useful; “if a transcript cannot be understood as readily as the oral language that it represents, then it is failing in its primary task of communicating what was communicated in the primary context” (2003: 30). On the other hand, transcripts should be verbatim accounts of the spoken discourse; the transcript should, therefore, be an accurate reflection of the language used in the discourse being transcribed. These two goals are often incompatible where readable transcripts are not also accurate and accurate transcripts are not also readable. Consider the following.  (2.1) EE: But, yeah you’re gonna play an active roll if, if the defence team got this  one that you sent John Allen in particular where you say you were, you did search down there and that, that you got… that you had two dogs in there and I want to say, dogs at the location knee deep in water. Well, that’s what location were they knee deep in water? Were they right there where the body was recovered? Those are the things I need to…  JJ: No, it… it was clearly within… like I mean I can… the reason.. the, the  way I can, I didn’t um, (inaudible)… for you. Is, um, I don’t know if when you guys Did your um, crime scene search and found a pink baby blanket and a Cooler sitting right there. That was the furtherest we went. I mean that was Like the furtherest point we went, which was, I want to say between five and Ten feet in.  (JJ1.11:10-22)22 While the speech transcribed in (2.1) may be accurate (and, indeed, when spoken aloud, may even be intelligible), it is not readable. Unlike in written speech, in spontaneous spoken speech, people “frequently change their mind half-way through a sentence ... add little asides under their breath … [and] convey part of their message through gesture” (Fraser, 2003: 217). Accordingly, an accurate written representation of spoken spontaneous speech will                                                  22 Following the referencing system I explain in section 2.2.2 below, this excerpt occurred at lines 10-22 on the 11th page of the transcript representing the first interview of the witness JJ.  49 often be unreadable, as exemplified in (2.1). Unreadable (but accurate) transcripts put greater demands on the reader than readable (but inaccurate) ones; for this reason, Gibbons (2003) notes that, when transcripts are created for the purpose of legal proceedings, the goal of readability generally trumps the goal of accuracy.   A transcript is a text that recreates a speech event, it is not the event itself. The language represented is constructed by the transcriber for a particular purpose (Green et al., 1997: 172). The act of creating a readable transcript requires the transcriber to actively interpret the language used by the discourse participants. In creating transcripts, transcribers are influenced by their own social evaluations of speech which in turn influence how they choose to record it (Roberts, 1997: 168). Moreover, the transcriber’s knowledge of the subject matter influences how they perceive speech and, by extension, how they transcribe speech; as Fraser (2003: 218) notes, in transcriptions of her own talks phonological has been transcribed as psychological, an error she attributes to listener knowledge rather than the quality of the recording. On the basis of these limitations, Fraser (2003) and Haworth (2010; 2013) have challenged the appropriateness of relying on the accuracy of transcripts when they are treated as evidence in a legal proceeding.  The transcripts that form the basis of this study are not being evaluated for their quality or appropriateness as evidence themselves, but rather as a record of the interview they purport to represent. Nonetheless, the transcripts that form the empirical foundation for this study are subject to the criticisms noted above, in particular since nothing is publicly known about the individuals who transcribed the discourse in the interviews in the Caylee Anthony investigation. The transcripts do not disclose the transcriber’s background knowledge about the investigation (which may affect how they perceive the speech they  50 transcribe), their native variety of spoken English, or even the quality of the recordings from which the transcripts were created. Nonetheless, the transcripts are the best available tool, since observing the police interviews themselves is not an option.  The flaws inherent in using transcript records of speech apply equally to other studies which make empirical claims about speech on the basis of transcripts. While much of the early literature on actually has focused on its use in the context of written English (Halliday & Hasan, 1976; Halliday, 1985; Chafe, 1986; Biber & Finegan, 1988; Fraser, 1988; Martin, 1992; Fraser & Malamud-Makowski, 1996), others have specifically considered actually’s use in either spoken language (Tognini-Bonelli, 1993; Lenk, 1998; Smith & Jucker, 2000) or some combination of spoken and written language use (Biber & Finegan, 1988; Tognini-Bonelli, 1993; Oh, 2000). Chafe (1986) parenthetically observes that actually is a feature of conversational English (1986: 270). Indeed, studies have established that actually is much more frequent in American English than in British English and, for both, more frequent in speech than in writing (Taglicht, 2001; Oh, 2000; Aijmer, 1986). The ratio of the use of actually in spoken versus written corpora is 10.3:1 for British corpora and 3.4:1 for American corpora (cf. Taglicht, 2001: 1). The use of actually is more prevalent in spoken British English than American English (Oh, 2000; Taglicht, 2001).  Because actually is largely a spoken phenomenon, I investigate transcripts of speech rather than written corpora. That said, all corpora in which spoken language is recorded in transcript form suffer the same potential flaws outlined above. Nonetheless, such transcripts will often be the only corpus data available to consider how lexical items like actually are used in spoken form. This is particularly true with police interview data, which is rarely made publicly available, let alone in its audio/video form. As such, this study analyses  51 spoken actually in police interviews on the basis of the recorded transcripts of those interviews, despite the flaws noted above.   2.2.2 The interview transcripts in the Caylee Anthony investigation While I used every transcript that I was able to find from the investigation, I do not assume the corpus contains the transcript of every interview conducted during the investigation. My skepticism in this regard is based on three factors. First, for some interviews, the timing suggests earlier interviews exist but were not available to view. For example, a formal interview with Casey Anthony’s father, GA, was conducted on July 24, 2008 but the first available interview transcript for her mother, CA, is dated August 1, 2008. It is unlikely that CA, who placed the 9-1-1 call reporting her granddaughter missing, was not formally interviewed until a full two weeks after the investigation began and a full week after her husband. Second, some transcripts explicitly reference earlier discussions with a witness where no earlier transcript is available. For example, in the transcript of the August 18, 2008 interview with AR, the interviewers put to him that his information is inconsistent with the information he gave in an earlier conversation. It may be that the earlier conversation was conducted informally and no audio recording, and hence no transcript, was made. For obvious reasons, interviews for which transcripts were not created could not be included in the corpus. Finally, there are certain witnesses referenced in the interview transcripts who would presumably have been interviewed in the course of the investigation but for whom I was unable to locate transcripts. In particular, Casey Anthony’s brother had a girlfriend, Mallory, at the time that Caylee disappeared and there was some evidence proffered in the interviews with the witnesses RF and DP to suggest that Mallory believed Casey Anthony caused harm to Caylee. Given the nature of her relationship with Casey and  52 the comments attributed to her by RF and DP, it is unlikely the investigators would have failed to interview her, even though I was unable to locate any transcripts. I cite specific excerpts from the police interview transcripts throughout the dissertation. Where I do so, the original source is referenced immediately following the excerpt and aligned to the right of the page. The format of the reference will look something like the following ((10) CA1.30:1-15) and which should be understood as follows. (10) is the token number; all data involving actually, apparently, or supposedly, were assigned a token number, the process for which is described below. Excerpts not including actually, apparently or supposedly were not assigned a token number and will not be referenced with one in the dissertation. The CA1 refers to the transcript, in this case the transcript of the first interview with the witness, CA. Where there was only one interview with a particular witness, the interview will be referenced by witness initials alone. The complete list of witness interviews is attached as Appendix A. The ‘30’ refers to the page number in the transcript and the ‘1-15’ refers to the line numbers where the excerpt exists in the original transcript. Where initials are intended to refer to the witness themself, they will occur in plain text (i.e., CA). Where initials refer to the transcript of the interview with that particular witness, they will be underlined (i.e., CA).  There are a number of typographical errors in the transcripts, some of which are expressly marked in the original transcript with ‘sic’ and some of which are not. Where the original transcriber or reviewer notes the error, it is marked in the transcript with “… [sic]…”. Where the error is not marked in the original transcript but where clarification is necessary, I flag the error by marking it “… [JG: sic] …” to distinguish my own corrections  53 from those in the original. Where the error has no potential effect on the point under discussion, I leave it unmarked.  2.2.3 The interviews  In total, 103 witness interviews conducted in the course of the investigation were reviewed for the purposes of this study. None of the witnesses interviewed invoked a right to silence; therefore, each of the interviews resulted in discourse between the interviewer and the witness.   Of the 103 witness interviews, 52 interviews were conducted in the course of the missing person investigation prior to Casey Anthony’s indictment, six were conducted in the course of the murder investigation after Casey Anthony’s indictment but prior to Caylee’s remains being discovered, and 38 were conducted after the remains were discovered. Finally, seven interviews were conducted by the Professional Standards Section as part of the internal investigation into police misconduct resulting from the mishandling of Roy Kronk’s August 2008 report. Those seven interviews were all conducted after Caylee Anthony’s remains were found.    The majority of witnesses were aware the interview was being recorded; only four of the 103 transcripts reviewed (GA3, KJ/LB, AP/RB, and AP/LB)23 reflect surreptitiously recorded interviews. Three of those conversations (KJ/LB, AP/RB, AP/LB) were controlled                                                  23 The interviews coded with two sets of initials (specifically, in JH/MH, KJ/LB, AP/RB, and AP/LB) reflect the fact that the interview involved two witnesses. Either two witnesses were interviewed together (as in JH/MH) or one witness was interviewing another witness on behalf of law enforcement (as in KJ/LB, AP/RB, and AP/LB, see fn. 22).   54 telephone calls24 between two witnesses and one (GA3) was a surreptitious recording created while transporting the witness to the jail where Casey Anthony was incarcerated.     Of the 99 interviews in which the witness knew their speech was being recorded, the majority were conducted in person: 65 were conducted at a police station, eleven were conducted at the witness’s place of work, seven were conducted at the witness’s residence, six were conducted in a vehicle, two were conducted in a lawyer’s office and the location of one interview was redacted in the transcript. Seven interviews were conducted over the telephone.  In four interviews (DoCa, AD, RK2, and JJ2), the witness was accompanied by an attorney, however, none of those attorneys participated in the interview other than to identify themselves for the record. Similarly, in one interview (KA), the witness had a third party with her; that third party did not contribute to the conversation in any meaningful way.   For the majority of the interviews (66), two interviewers were present. 32 interviews were conducted by a single interviewer. For three interviews (GA1, GA3 and SP2) three interviewers were present and for two interviews (RPl and JJ2), four interviewers were present. However, regardless of how many interviewers were present, one interviewer typically took the lead and did most if not all of the talking.   2.3 The speakers  30 different interviewers and 84 witnesses participated in the interviews reviewed in this study. Given the circumstances in which the transcripts were generated, I do not expect                                                  24 A controlled telephone call is normally one in which a witness calls a suspect for the purpose of obtaining incriminating statements from the suspect at law enforcement’s request. In these cases, however, the controlled telephone calls were arranged in order to make contact with a witness, LB, and her lawyer, RB, neither of whom were suspects in the disappearance or murder of Caylee Anthony.   55 there to be significant issues of sociolinguistic variation in terms of how the various speakers express evidentiality in the data. For example, with the exception of some of the Texas EquuSearch volunteers, the majority of witnesses are from or, at least, were living in Florida at the time of the interviews; thus, I do not expect differences in regional variation to skew the data. Moreover, as I show below, a significant number of witnesses are from the same speech community. Although I am unable to control for established sociolinguistic variables like age, gender, education, etc., to the extent they are discernable from the transcripts, there is no indication that evidential use varies along these sociolinguistic factors. This is consistent with Waters’ (2011) finding that, at least in Canadian English, speakers use actually robustly regardless of age, sex, or education. Nonetheless, to the extent these sociolinguistic variables are determinable in the transcripts, I record them in the following sections.   2.3.1 The interviewers  The investigation was led by a detective with the Orange County Sheriff’s Office, Corporal Yuri Melich (YM) who, by virtue of his lead role, was present at more interviews (45) than any other interviewer. YM was assisted in the investigation primarily by his then-supervisor, Sergeant John Allen (JA), who participated in 31 interviews, and a colleague, Corporal William “Eric” Edwards (EE), who participated in 41 interviews. At least one of the three main investigators was present at 75 of the 103 interviews, and at least two of them were present at 51 interviews.   56  Besides YM, JA, and EE,25 19 other detectives from the Orange County Sheriff’s Office were present at one or more of the interviews reviewed. Three Special Agents from the Federal Bureau of Investigation (SB, SMc, and NS) were involved in the investigation; the FBI conducted two interviews independently (CA3 and GA4) and were represented at an additional six interviews conducted by members of the Orange County Sheriff’s Office. The Assistant States Attorney (LDB) who ultimately led the prosecution at trial was present at two interviews, one as an observer (AD) and one as the lead interviewer (JJ2). Two representatives of the Professional Standards Section were involved in the investigation into police misconduct: a corporal (MR) conducted all seven interviews into those allegations and was assisted by a sergeant (RH) in four of those seven interviews.   For three of the interviews (KJ/LB, AP/LB and AP/RB), the interviewer was not a member of a law enforcement agency at all. One of the Texas EquuSearch volunteers, LB, made certain claims to the media that she had searched the area where Caylee’s remains were ultimately discovered and that, at the time of the search, the body had not been there. The investigators attempted to interview LB with respect to these statements but were unable to make contact with her. As a result, the investigators arranged for two other Texas EquuSearch volunteers, AP and KJ, to individually contact LB by telephone and to surreptitiously record the resulting calls. Even though investigators were listening to the conversation, it was the witness who questioned LB. A similar tactic was used to interview LB’s attorney, RB; the investigators surreptitiously recorded AP’s conversation with RB when she returned his telephone call. Therefore, AP and KJ also acted as ‘interviewers’ for                                                  25 The three main investigators are referred to in the transcripts either by their initials or by their rank and last name. So, Corporal Yuri Melich is either “YM” or “CM”, Sergeant John Allen is either “JA” or “SA” and Corporal Eric Edwards is either “EE” or “CE.”  57 the purpose of three of the transcripts despite not being law enforcement officers.  With respect to speaker demographic information, 22 interviewers were officers with the Orange County Sheriff’s Office, three were FBI Special Agents, two were members of the Professional Standards Section, one was the Assistant States Attorney, and two were Texas EquuSearch volunteers whose regular occupation, if any, is unknown.  Birthdates were not recorded in the transcripts for any of the interviewers so their age at the time of the interview cannot be determined. Assuming gender can be determined by the interviewer’s name, 24 interviewers were male and six were female.26     2.3.2  The witnesses  The 103 transcripts reviewed represent the interviews of 84 witnesses. 65 of the witnesses were interviewed a single time (JH/MH were interviewed together, generating a single interview transcript), 14 were interviewed twice, one was interviewed three times and two were interviewed on four separate occasions. For two of the witnesses (AP and KJ), they were ‘interviewed’ prior to and after their controlled telephone calls with LB and RB, which also formed part of the transcript: AP made two controlled telephone calls to LB and one to RB. KJ made a single controlled telephone call to LB.   The witnesses fall broadly into two categories: those who knew Casey Anthony before her daughter’s disappearance and those who became involved only after Caylee Anthony’s disappearance. Members of the former category formed part of Casey Anthony’s social network while members of the latter did not. Two witnesses were interviewed who had no relation to the case whatsoever; in JH/MH, a couple was interviewed because Casey                                                  26 Although the interviewer, KJ, arguably has a unisex name, he is referred to as a ‘good guy’ at KJ/LB5:16 and is therefore assumed to be male.  58 Anthony had made a telephone call to their house. As a result of the interview, it was determined that the phone call had been in error, that the husband had the same name as a friend of Casey Anthony’s and that she had dialed his number by mistake. They had no prior involvement with Casey Anthony and no further involvement in the investigation.  The transcripts disclose certain demographic information about the witnesses; where the interviewer asks for birthdates and occupation, that information forms part of the record although in a number of cases, the birthdate is subsequently redacted. The only sociolinguistic factor for which information is robustly available is gender; of the 84 witnesses interviewed, 53 were male and 31 were female. Other sociolinguistic information is more broadly available for the witnesses who formed part of Casey Anthony’s social network than for those who did not.   2.3.2.1 Social network witnesses  Of the transcripts reviewed, 40 involved witnesses who had a personal relationship with Casey Anthony prior to the investigation and formed part of her social network to varying degrees. Each witness was questioned with respect to the nature of their relationship with Casey Anthony and their answer forms part of the record. Moreover, the transcripts show the relationship between the witness and Casey Anthony as well as the witness’s relationship to the other people being interviewed; based on the transcripts, I was able to discern how each witness was connected in Casey Anthony’s social network and how strong the connections were. I relied primarily on how witnesses self-reported the nature of their relationship with Casey Anthony. For Casey Anthony’s family members, determining the nature of the relationship was noncontroversial. Friendship-based relationships, however, were more difficult to discern. Cheshire (1982) and Eckert (2000) suggest reciprocal naming  59 as a starting point for determining social connections. However, because I did not review the transcripts of Casey Anthony’s interviews,27 I was not able to determine whether the self-reported friend relationships were reciprocal. For example, while RP reported that he was one of Casey Anthony’s best friends, I could not determine whether Casey Anthony reciprocally included him in the list of her best friends. Therefore, I constructed the social network by relying on the witness’s own self-report as well as the nature and density of the network ties, as I describe below.  At the core of Casey Anthony’s social network are her family members, five of whom were interviewed in the course of the investigation: her mother Cindy Anthony (CA), her father George Anthony (GA), her brother Lee Anthony (LA), her maternal grandmother Shirley Plesea (SP), and her maternal uncle Rick Plesea (RPl). I consider these witnesses to be core to Casey Anthony’s social network both because of the density of the network and the nature of the ties within that network. Each of the witnesses has ties not only to Casey Anthony but also to each other, thereby creating a dense network. Moreover, these witnesses were not only related to each other but, in the case of CA and GA, shared a home with Casey Anthony at the time of Caylee’s disappearance; as such, they had multiplex ties.28 CA, GA, and LA all lived in Orlando, Florida while SP lived in Mount Dora, Florida. RPl’s residence is redacted in the record. Casey Anthony’s mother, CA, is a nurse who at the time of the interviews worked for a health services company. Her father, GA, worked in security after giving up careers in law enforcement and car sales. Specific ages are not given for any of                                                  27 As discussed in chapter 1, the institutional goals of suspect interviews versus witness interviews are significantly different. For that reason, I chose not to consider the publicly available interviews with Casey Anthony as part of the police interview transcripts.   28 It is unclear from the transcripts whether Casey Anthony’s brother, LA, also lived at their parents’ home at the time of Caylee Anthony’s disappearance.   60 Casey Anthony’s family members.  Also at the core of Casey Anthony’s social network are her friends. For the purpose of constructing the social network, I define ‘friend’ on the basis of the strength and density of the connection. I considered a romantic relationship to be a strong connection and, as such, included Casey Anthony’s current and former boyfriends as ‘friends’ and, therefore, as core members of her social network. With respect to non-romantic relationships, I included as a ‘friend’ those witnesses who (a) self-reported as a friend and (b) had a connection to another witness who also self-reported as a friend. Applying these criteria, 21 friends were interviewed in the course of the investigation.   Of all of the witness types, sociolinguistic demographic information is most accessible for the ‘friends’ group, particularly with respect to gender, age, and occupation. Six of these witnesses were female and 15 were male. 13 of the 21 witnesses were asked to provide or volunteered their ages; for three of those witnesses, their answer was redacted in the record. The remaining 10 ranged in age from 20 to 24. Of the 11 of the friends whose ages are unknown, five knew Casey Anthony from high school; presumably, those five witnesses were also in their early 20s at the time the interviews were recorded.29 Four of the witnesses (JG, AL, AR, and WW) for whom ages are not recorded were boyfriends or former boyfriends of Casey Anthony’s and another was boyfriend AL’s roommate (NL). Again, I assume those five witnesses were within a similar age demographic at the material time. The final witness whose age is unknown, ID, is connected to the network through Casey Anthony’s high school friends JD, SD, AD, and DS; I assume he is in the same age range as his cohort. Therefore, the record establishes that 10 of the 21 friend witnesses were in their                                                  29 Casey Anthony was 22 years old when the investigation commenced.  61 early 20s, ranging from 20 to 24, at the time they were interviewed and the remaining 11 friend witnesses were likely in the same age demographic.  16 of the 21 friend witnesses were questioned about or volunteered information with respect to their employment status.  Of those, six reported being students, two were servers, two were tattoo artists, two worked at an auto dealership, one was a travel agent, one a labourer, one in the army, and one a deputy with the Orange County Sheriff’s Office. BS volunteered that she was a single mother who works nights, but did not report the nature of her occupation. A summary of the speaker demographic information for Casey Anthony’s friend cohort is given in the table below. Speaker Gender Age  Relationship Occupation AD F  High school friend  AH F 23/2430 Friend Server/stagehand AL M  Boyfriend Student AR M  Former boyfriend Police officer BS F 22 High School friend  CCa M 21 Friend Student CS M 20 Former boyfriend Student DS M Redacted High school friend Car dealership ID M Redacted  Friend Army JD M  High school friend Tattoo artist JG M  Former boyfriend Server JeH M 23 Middle school friend Student KMC F 21 Middle school friend Travel agent LG F  High school friend  MCa F 22 Childhood friend  NL M  Friend  Student RM M 24 Former boyfriend Car dealership RP M 23 Childhood friend Student SD M  High school friend Tattoo artist TB M 24 Friend  WW M Redacted Boyfriend Labourer TABLE 2.1: Speaker demographic information for Casey Anthony’s friends  In addition to those I have defined as friends, 14 witnesses had some relationship                                                  30 AH was 23 at the time of her first interview, and 24 at the time of her second interview.   62 with Casey Anthony prior to the investigation. I refer to these witnesses as ‘acquaintances’ because they were on the periphery of her social network. Of these 14 acquaintances, six are female and eight are male. Three acquaintances are colleagues of Casey Anthony’s mother, three are neighbours or former neighbours, three are friends of friends, one is Casey Anthony’s former boss, one is a former coworker, one is a nurse at Casey Anthony’s grandfather’s care home, one is Casey Anthony’s ex-boyfriend’s father, and one is the tattoo artist who employs Casey Anthony’s friends JD and SD.  Specific age information is given for only four of the acquaintances.  Demographic information on these witnesses is summarized in the table below. Because the witness CA and the suspect Casey Anthony have the same initials, I refer to the witness CA as ‘Cindy’ in the table below for the purposes of clarity. Speaker Gender Age  Relationship Occupation BB M  Neighbour  CCr M  Cindy’s boss Gentiva manager DB F 47 Cindy’s coworker Health care/nurse DaCo M  JD and SD’s boss Tattoo artist DePo F 56 Cindy’s supervisor Gentiva employee JeanC M Redacted Neighbour  JR F Redacted Coworker server KA F Redacted Employee at grandfather’s nursing home Nurse MaKi F  Boyfriend AL’s roommate’s girlfriend  MC M  High school friend Leasing agent MiKo M  Former boss Kodak manager MM F 25 Brother’s friend  RG M  Ex-boyfriend JG’s father Minister TMF M 49 Former neighbour  TABLE 2.2: Speaker demographic information for Casey Anthony’s acquaintances 2.3.2.2 Non-social network witnesses  63  44 witnesses were interviewed as a result of their participation in the search for Caylee or their participation in the investigation into her disappearance and murder; none of these witnesses formed part of Casey Anthony’s social network. This witness group is comprised primarily of various citizens with information about the events surrounding Caylee’s disappearance, the Texas EquuSearch search team, and various private investigators, lawyers, and law enforcement officers who participated in the investigation.   13 witnesses were interviewed with respect to information about the circumstances occurring around the time Caylee was reported missing and the eventual discovery of her remains, including: RK, who reported finding what turned out to be Caylee’s remains, along with his colleagues and supervisors (DD, CG, ARo, and CSp); three employees from the auto-wrecker operation that towed and stored Casey Anthony’s car shortly before Caylee’s disappearance was reported (SB, GR, and NL); and five additional witnesses who inserted themselves into the investigation in various capacities (RH, KW, DW, DP, and RF).   On November 8 and 9, 2009, an intensive search was undertaken by Texas EquuSearch, a search and rescue organization dedicated to searching for missing persons, but that search was unsuccessful in locating Caylee’s body. 18 members of the Texas EquuSearch team, including its director, TM, were subsequently interviewed as part of the investigation. Speaker demographic information is extremely limited for this group of witnesses. Texas EquuSearch itself is based out of Dickison, Texas, but people from various geographic regions volunteered in the search for Caylee Anthony. Moreover, the interviewers asked only one volunteer searcher, SC, for her date of birth and her answer was redacted in the transcript.    Two private investigators (DoCa and JH) participated in the search for Caylee at the  64 behest of the Anthony family and both were interviewed as part of the investigation. Similarly, two lawyers representing parties in the investigation (RB, who represented Texas EquuSearch volunteer LB, and BC, who represented Casey Anthony’s parents) were interviewed.   Nine of the witnesses interviewed were active law enforcement officers: five (AA, RC, KC, TU, and JW) were interviewed with respect to their involvement in the investigation and four (JA, EC, YM, and ET) were interviewed as part of the internal investigation into police misconduct. As a result of the internal investigation, three of the investigators, JW, YM, and JA, participated as both interviewer and witness in the investigation.   2.4 The timeline of interviews  The investigation continued from July 15, 2008, the date Caylee Anthony was first reported missing, until the trial began in May 2011. While interviews were conducted throughout that period, all of the witnesses who were members of Casey Anthony’s social network, however peripherally, were interviewed within the first six months. That is, for the witnesses who were members of a specific speech community and a connected social network, their speech was collected within a relatively short period of time. Witnesses who were not part of Casey Anthony’s social network, and therefore who may not have formed part of a unique speech community, were interviewed over the course of the investigation and primarily after Texas EquuSearch’s involvement.   2.5 The data  65  A total of 3,065 pages of transcript were reviewed, the average transcript length being 29.75 pages, the shortest transcripts being 3 pages (TMF and JW) and the longest being 361 pages (CA1). Each page of transcript contained 25 lines of text and averaged approximately 225 words. Therefore, the police interview transcripts reviewed for this study contained approximately 690,000 words.   2.5.1 Data coding  The discourse markers flagged during review of the police interview transcripts were actually, apparently, supposedly, in fact, actual, must, have to, evidently, and allegedly, although the analysis is limited to actually, apparently, and supposedly.31 Each transcript was read in its entirety with the relevant markers assigned a token number and flagged for future analysis. The goal in reading the entire transcript was threefold. First, I was able to note and record instances of partial word transcriptions. Because the transcriber attempts to be as accurate as possible, partial words will often be transcribed if that is how they were produced.   (2.2)  JH: … because Caitlyn obviously goes there a lot.  And uh, she was     recommended uh, the crab puffs for uhm, for appetizer.   JA: Okay. JH: And uh, she ordered the uh, green toma… like I said, there was green tomatoes.  I think they had something else too, but.  .... (JamesH17:8-12)  Relying on a ‘search and find’ approach alone could potentially miss partial words, depending on how they were transcribed.                                                   31 As described in Section 2.5 below, I limited my analysis to only those discourse markers which occurred more than 50 times in the police interview data. Although there were fewer than 50 tokens of must, I consider its use for the purpose of comparing it with the analysis in von Fintel & Gillies (2010).    66  Second, a ‘search and find’ approach would fail to capture any of the target words that were misspelled in the transcript. Because the interviews were transcribed by police transcribers, they are subject to human error both in terms of mistakes and spelling conventions. For example, consider the extract in (2.3) below.  (2.3) CA: That this is looking more like a pre-trial or you know becoming to  look like a trial.  SB: You mean in the media? CA: No actually in the courtroom that day for her bond hearing.  It looked like a pre-trial.  It looked like you now a court preceding for a trial and not a bond hearing.  ((80) CA2.211:9-13) The transcriber transcribed now when the context suggests the correct spelling would have been know. In addition, the transcriber refers to a court proceeding as a court preceding. A ‘search and find’ approach would fail to capture any of the target words if they were subject to an error in transcription.   Third, and most importantly, by reviewing the transcript in its entirety, I was better able to situate the relevant discourse marker within the broader discourse. For example, in transcript KW1, the witness gives an account of how he came to show a police officer where a particular bag was located within a particular geographic area. In (2.4), KW’s evidence was that the officer in question didn’t really walk completely in the area, suggesting that the police officer may have entered the area in question but did not examine it in its entirety.   (2.4) MR: How close were you, when you were pointing out the bags to him? Where  were you standing, compared to where he was standing? Like, like, in feet or distance. KW: Where was I? Well, at first, when I was still trying to show him the area, I tried to walk over there, he didn’t really; he walked kind of back but he didn’t    really walk completely in the area and then… ((618) KW1.4a:23-25)   67  However, in (2.5), KW revises his evidence to say the police officer “didn’t actually walk to the area” KW was trying to show him.   (2.5) MR: Okay.  Did, at any time, he walk into the wooded area and, and start going  through bags in the woods? KW: No. He never actually went and picked up any, he just went back there and looked.  Not even looked, looked, I mean he just walked back in that little trail thing and just kind of looked, but didn’t actually walk to the area I was really trying to show ‘em. ((619) KW1.5:9-13)   The two actuallys in (2.5) can only be fully understood if one has read the preceding excerpt and knows that this is a reference to an earlier part of the discourse.   Similarly, the witness AH told the interviewers that Casey stole money from her at page 21 of AH1 (in (2.6)); however, she does not explain how she came to know that until later in the interview, at page 22 (as shown in (2.7)).  (2.6) JW: They dropped you off? AH: Yea after after they dropped me off um because she had actually stolen four hundred dollars from me the week before I left.  Um (over talking) I didn’t know she stole it, I knew it was missing … ((359) AH1.21:1-4)   (2.7)  AH: We only had a limited amount of time to talk um and so she’s telling me  all this like Casey had been stealing money from her for years so she stole money from her eighty year old grandmother by using a routing number on a birthday check and like all this stuff and I was like … so that’s where my four hundred dollars went… (AH1.22:18-22) Thus, the train of discourse is important over the course of the interview because the witness may give evidence in one part, but explain how they came to know that evidence in another. This is particularly important given that part of what I am investigating is the type of evidence the speaker has for a proposition when they use various discourse markers to  68 express that proposition.   The transcripts show that speakers maintained a train of discourse not only within single interviews but also across interviews, as in the following example from the witness CA. The witness gives an account of how she came to understand that Casey had been in Jacksonville, Florida on or about July 3, 2008 in both CA2 (which interview was conducted on September 25, 2008) in (2.8) and in CA3 (which was conducted on April 21, 2009) in (2.9).   (2.8) CA: Friday was the 4th of July, we spoke a couple times um, as to whether or  not we were gonna take Caylee to see fireworks.  And Caysee [sic] wasn’t sure if they were coming back and then…  EE: From where? CA: She was supposed to be still in Jacksonville.  Cause on Thursday is when she told me she’d actually been in Jacksonville all week.  That’s what she said….  ((122) CA2.17:19-25)  (2.9) CA: July 3rd, Casey told me that the baby sitter had Caylee at the um, like  character breakfasts and things like that at Universal park while she was working.  And I told her I was gonna come up and meet her and pick up and Caylee up [sic].  ‘Cause I was off work that week I said, you know what, let’s relieve Jennifer of her duties I want to watch her.  I actually drove up to um, Guest Services.  Didn’t go in to Guest Services just called her and asked her I said, well, I’m at Guest Services where can you meet me at?  And she says, well mom I’m not there I’m in Jacksonville… (CA3.184:9-16) Finally, the evidence source of certain critical investigation details was referenced across interviews with different witnesses; that is, I tracked how police interviewers referred to evidence obtained by one witness when interviewing another witness. In their first interview with AH on July 23, 2008, police investigators learned that Casey had sent her a text message on June 28.   69 (2.10) EE:  She texted you about gas cans?  AH: Yes.  EE: When?  AH: Uhm, the same, the 27th.  I think it was, it may, actually no, it was the 28th.  I could find it.  It would just take a while.  Uhm…  EE: Well, that’s fine.  AH: There’s a lot of text messages to go through.  EE: Okay.  AH: Uhm, but it’s, it should be in the stuff that you guys have now.   DR: Uh-hum (affirmative). AH: Uhm, uhm, because I remember looking at it and I think it was the 28th that she asked me.  It was either the 28th or… it was before she showed up, but it was not the same day as her telling me about the animal or that she had run out of gas. Uhm, so I was like, “Why don’t we, we’ll just go to target. We’ll buy a gas can.” (AH1.18:25-19:13)  AH’s evidence was that she received the text message about gas cans on June 27 or 28, 2008 and that she had given copies of her text messages to the police. On July 29, 2008, the police investigators refer to AH’s evidence and their own review of her text messages in their interview with LA.  (2.11) EE: … Fast forward.  27th.  You know that the 27th your sister runs out of gas  again at the AmScot?  LA: Right. EE: Now how do you know that it was the uh, the 27th that she ran out of gas? What we’re trying to establish… LA: Just what’s been conveyed to me by my, by my mother and father. EE: Okay.  Because the car gets towed on the 30th. LA: Correct.  I’ve been told, and I’ve heard that, that AmScot waited three days before they towed it. EE: Okay.  Yeah (affirmative), and I told you Amy actually got a text message from her on that day. ((179) LA62:8-18)  Thus, the interviewers reference evidence obtained in the interview with AH in the interview with LA.  The tracking of evidence across witnesses would not be possible without reviewing the interviews in their entirety.  Once all of the transcripts were read in their entirety and the relevant discourse  70 markers flagged for analysis, I performed a ‘search and find’ scan of each document to ensure no tokens were missed. Any missed tokens were then similarly flagged.   2.5.2 The tokens  Once the transcripts were coded, the total number of tokens occurring in the corpus were counted and the results are as follows.   Marker # of tokens actually 712 supposedly 76 apparently 55 had to (have) be(en) 50 actual 24 in fact 18 must 9 allegedly 1 evidently  1 TABLE 2.3: Token count by marker As is clear from Table 2.3, significantly more tokens of actually were produced than any of the other discourse markers being investigated. Accordingly, this study focuses primarily on actually; although I will have less to say about them, apparently and supposedly will also be analysed.   2.5.2.1 The use of actually  Of the 103 transcripts reviewed, 93 transcripts contained one or more tokens of actually.       (2.12) Tokens in the corpus        N  (%)  71  Transcripts containing actually  93  (90.3)  Transcripts not containing actually  10  (9.7)  Total number of transcripts   103  (100)  712 tokens of actually were produced, which is significantly more than would be expected, given Oh’s (2000) observations about the frequency of occurrence of actually in spoken American English. Based on a comparison of the Brown Corpus for written data and the Switchboard Corpus for spoken discourse, Oh (2000) observed that actually occurs in spoken American English more frequently than in written American English; the frequency of occurrence of actually in the written corpus was 0.016% and 0.054% in the spoken corpus. The frequency of occurrence in the interview corpus constructed for this study is 0.102%, almost double the frequency found in the Switchboard spoken corpus.      Brown  Switchboard  Interview Transcripts  Total words  1,014,312 2,400,355  690,000  Actually  166 (0.016%) 1,293 (0.054%) 712 (0.103%)  TABLE 2.4:  Frequency of occurrence in spoken English versus police interview     transcripts   Each individual token and its surrounding dialogue were extracted to a separate document where they could be analyzed for the following: speaker type (interviewer or witness), utterance type (question, negation, or statement), position within the utterance (parenthetical or clause-internal) and the speaker’s evidence source for the prejacent, where it could be determined. I coded for utterance type in order to determine whether, like other evidentials, actually interacted with different clause types. I coded for position in order to determine whether actually functioned differently, depending on where it occurred in the  72 utterance, as previous research has suggested (see, for example, Taglicht (2001); Tognini-Bonelli, 1993; Lenk (1998); Aijmer (1986)).   Of the 712 tokens of actually analyzed, 218 were produced by an interviewer and 494 were produced by the interviewee; that is, 69% of the tokens were produced by a witness. In 54 of the interviews in which at least one token was produced (58%), the witness produced more tokens of actually than the interviewer. In 31 interviews (33%), the interviewer produced more tokens of actually than the witness and in 8 interviews (9%), the number of tokens produced by the witness and the interviewer were equal.  124 tokens of actually from the corpus involved questions, five of which were negative questions.  All 124 questions were asked by interviewers. 73 tokens involve negation, not counting the five negative questions. Although I initially coded for negation and counted these tokens, they did not ultimately turn out to be significant and, as a result, I did not perform a separate analysis of actually in negative clauses as part of this study. The remaining 515 tokens are in positive assertions.  The vast majority of tokens (566) occurred clause-internally; 125 occurred parenthetically, with 113 occurring sentence-initially and 12 occurring sentence-finally. For 21 tokens, the utterance position could not be determined from the transcript. For example, in (2.13), it is unclear how the utterance was intended to be parsed and, as a result, it is unclear whether actually is initial, medial or final.  (2.13) SA: They interviewed you or what?  JJ: Yeah.  SA: Okay. CE: Did you share your opinion about the body being moved on the radio interviews? JJ: Um no actually not at all no. ((673) JJ2.71:24-72:3)  73   (2.14) a. Initial:  Um no. [Actually, not at all no.]  b. Medial: Um [no actually not at all] no.  c. Final:  [Um no actually.] Not at all no. As shown in (2.14), depending on how the utterance is parsed, the token could occupy sentence initial, medial or final position. 21 tokens were similarly inconclusive.  To summarize, the vast majority of actually tokens were produced in declarative statements by witnesses and occurred clause-internally. In chapter 3, I show that speakers use actually to indicate that they have the best possible evidence for the information they proffer in the course of police interviews. I argue that, given the institutional goal of the witness interview, it is unsurprising the majority of tokens are produced by witnesses in affirmative statements and that all of the actually-questions are produced by interviewers.  2.5.2.2 The use of apparently and supposedly  Far fewer tokens of apparently and supposedly were attested in the police interview transcripts than were attested for actually. Only 55 tokens of apparently were attested and 76 tokens of supposedly. Given the paucity of data for apparently and supposedly, less can be definitely said about their discourse use than can be said for actually. Nonetheless, I observe that, like with actually, the vast majority of apparently and supposedly tokens were produced by witnesses (43 for apparently, 74 for supposedly); far fewer were produced by interviewers. As with actually, the majority of supposedly tokens occur clause-internally (62 tokens) while far fewer occur parenthetically, either sentence-initially (nine tokens) or sentence-finally (one token). On the other hand, apparently is attested almost as often parenthetically (22 tokens) as clause-internally (26 tokens). There are three attested tokens of  74 sentence-final apparently. Neither apparently nor supposedly is attested in questions or negations.   2.6  Summary  I have set out the method by which I acquired the data which forms the basis for this study and how I identified individual tokens for analysis. While initially I intended to analyse a number of discourse markers, including actually, apparently, supposedly, in fact, actual, must, had to have been, allegedly, and evidently, it became clear that some discourse markers occurred less frequently than I might have expected. For that reason, I arbitrarily chose 50 tokens as my cut-off. I chose to analyse discourse markers with more than 50 attested tokens and to leave the others to future research. As such, my analysis is limited to actually, apparently, and supposedly. As noted earlier, while I could not explicitly control for sociolinguistic variation, speakers’ use of actually, apparently, and supposedly appeared to be consistent across speakers regardless of their age, gender, or social class (to the extent these could be determined). In this chapter, I set out some broad observations about where those tokens occur, both syntactically and with respect to who introduces them into the discourse. In chapter 3, I present empirical evidence to show how speakers use those discourse markers and what they are intended to mean.       75 “We’re gonna let the facts take us where they take us” EE (CA3.44:11)      3. THE SEMANTICS OF ENGLISH EVIDENTIALS   In this chapter, I set out in detail the empirical results from my review of the police interview transcripts and set out my semantic analyses of actually, apparently, and supposedly. The data in the police interview transcripts show that English speakers use actually to mark that they have the best possible evidence for their propositions. That is, English speakers use actually in the same evidential contexts that Faller (2002) shows Cuzco Quechua speakers use -mi. Moreover, actually is the evidentiary opposite of must, which von Fintel & Gillies (2010) and Matthewson (2015a, 2015b) show is only felicitous where the speaker’s evidence is not the best possible. I further show that speakers use apparently when they have indirect evidence for their propositions and supposedly where they have reported information that they specifically distrust.  In section 3.2, I formalize my analysis of actually as a marker of best evidence, following Matthewson (2015a, 2015b). I show that both Faller’s analysis of -mi and von Fintel & Gillies’ analysis of must contain additional formal components neither of which my data require me to adopt for actually; as such, I propose an analysis of actually as a marker of direct evidence strength in the sense of Matthewson (2015a, 2015b) but which nonetheless draws heavily from both Faller and von Fintel & Gillies. I argue that apparently makes an evidential contribution on the evidence type dimension of meaning, encoding indirect evidence. Supposedly encodes both indirect evidence type, specifically, reported evidence, and evidence strength, marking that the speaker’s evidence is ‘not best’ in the circumstances.   76  This chapter is organized as follows. I set out my empirical findings in section 3.1. In section 3.2, I set out Faller’s and von Fintel & Gillies’ analyses in greater detail and show why modifications are necessary to account for actually. I then propose my analysis of actually as a marker of evidence strength. Because an evidence strength analysis relies in large part on speaker intuitions about reliability, I show that such an analysis predicts that there will be speaker variation, particularly with those evidence types that require some reasoning. I show that this prediction is borne out in the data. In sections 3.3 and 3.4, respectively, I propose an analysis of apparently in a more traditional evidence type framework and analyse supposedly as marking both evidence type and evidence strength. In section 3.5, I speculate about what additional evidentials may exist in English. I show in section 3.6 that the evidential contexts in which speakers use actually largely mirror the evidential contexts where a court can admit evidence as reliable. Finally, I end the chapter by considering how the evidential analyses I propose for actually, apparently, and supposedly permit parties to align with specific participant roles in police interviews.   3.1 The evidence that licenses actually, apparently, and supposedly In this section, I show that speakers regularly use actually where they have ‘best’ evidence for the information contained in their propositions. Specifically, I show that speakers regularly use actually p where the source of their information for p is speaker performance, sensory observation, trustworthy reports, or general knowledge. I show that, to this extent, the use of actually largely parallels the evidential marker -mi in Cuzco Quechua, as described in Faller (2002) and discussed in chapter 1. Because actually was more prevalent in the police interview transcripts than apparently and supposedly, this chapter is primarily devoted to describing its context of use, with comments on apparently and  77 supposedly following. Specifically, I show that speakers use supposedly to introduce reported information that is not reliable, and apparently for information they know indirectly and where the speaker is agnostic about the reliability of the information.  3.1.1 Where actually occurs Like Faller (2002) shows for Cuzco Quechua -mi, and as discussed in Matthewson & Glougie (2015), actually p is felicitous where the speaker’s source of information for p is one of the following (repeated from chapter 1):  (3.1) i. sensory observation of the event itself   ii. trustworthy reports  iii. general knowledge  I further show that, unlike -mi, actually p is also felicitous where the speaker’s evidence for p is speaker performance; actually p is felicitous where the speaker performed the act described by p. I show this is consistent with the analysis of actually as marking best evidence. Finally, I show that actually is felicitous in precisely the contexts in which von Fintel & Gillies (2010) and Matthewson (2015a, 2015b) show must is infelicitous.  3.1.1.1 Sensory observation The police interview transcripts show that witnesses use actually to refer to events they observed. That is, actually is felicitous where a speaker has any type of sensory information for the event.    (3.2) VISUAL   CA: She [Eric’s widow] got in touch with Casey and told her um right after  Eric died Annie knows the exact date when Eric passed away.  …  78 EE: Tell me about the widow how how how the widow got in touch with uh Casey? CA: Casey um and um… she said she called um the widow called Casey and told her and then Casey actually had an obituary and believe it or not I remember seeing the obituary and never paid a bit of attention to the kids name or anything else …  ((95) CA2.46:15-47:3)   p: Casey had an obituary  EV: speaker saw p  (3.3) AUDITORY  LA: … And as soon as that started happening my mother realized that I’m  whispering to my sister, she gets you know, smart to the situation and realizes my sister is saying something and she busts into the room.  And again, the door wasn’t shut or anything, but she comes into the room, sees that my sister was crying, and she said, and actually the first thing that she asked her, she says “What have you done?”  EE: Really? LA: That’s the first question that my mom asks her. ((161) LA12:11-18)   p: CA asked Casey “What have you done?”  EV: speaker heard p (auditory)   (3.4) OLFACTORY  SP: You know, I just, my power went out.  …  SP: ... We'll go in here because I got a thing, I just got, smell the house? It’s  that thing, that big, my kitchen light just went phew.  … SP: … Want to come in and try some lights or something?  UM: Well yeah (affirmative), go ahead and turn your lights.  YM: Well, it’s not flashing anymore.  SP: It smelled like they burned up actually.  Wait a minute (inaudible).  UM:  (Inaudible) There ain’t no telling with electricity.  YM: Huh.  SP: No. ((566) SP2.2:4, 2:19-20 and 46:23-47:4)   p: the lightbulbs burned up  EV: speaker smelled p (olfactory)  79  While only three of the five senses were attested in the corpus, it is clear that actually can be used to indicate any of the five senses in the appropriate circumstances, as shown in (3.5).   (3.5) A: Mary actually burned the toast this morning.   B: How do you know that? (a) I saw that it was all black when she brought me some. (b) I heard the smoke alarm go off while she was cooking. (c) The kitchen smelled like burnt toast after she was cooking. (d) I felt it crumble in my hand when I picked up a piece. (e) It tasted burnt when I ate some.  The fact that actually p can be used whenever the speaker has any appropriate type of sensory evidence for p is consistent with the intuition that led the interviewer in (3.6) to ask specifically about various types of sensory information.   (3.6) JW: Okay.  Had you ever met these people before?  NL: No. No.  JW: Uhm, uh, had you ever seen the car before?  NL: No.  JW: Did you actually see, touch or feel the car, look at it? NL: I didn’t until after they came to pick it up when I went back out.  Then I watched them drive the vehicle.  Get in the vehicle and drive out.  ((482) NiL11-12:25-6)  Actually is felicitous where the speaker has any type of sensory evidence for the information they proffer.  3.1.1.2 Trustworthy Reports In addition to when they have sensory evidence, speakers in the corpus use actually p where their source of information for p is a report the speaker considers to be trustworthy.     80 (3.7) CA: They looked at Casey’s phone records all up until this point too from I  think the end of May the first couple weeks of June.  He [Lee] said there was a… big change in pattern.  You guys will need to look at those phone records.  Big change in pattern from the 14th from about… SB: June 14th? CA: June 14th, in the afternoon.  And I’m not sure if it starts around 4 or something in the afternoon.  But it’s like frantically phoning people.  You know calling boom… boom… boom calling all these people.  As if she were looking for something or trying to reach somebody that knew where someone was.  That there was actually a change in Casey’s pattern on the phone.   ((40) CA3.78.:11-20)   p: there was a change in Casey’s cell phone use pattern  EV: speaker was told by LA that p    (3.8) CA: Yeah Jeff was IT person a computer geek and he worked on things like at  Nickelodeon studios and Universal Studios.  And Heather that person that called me the other day that worked with Casey does remember a Jeff from the IT um, thing that Casey… and she actually met [him?]…  ((62) CA3.173.5-8)   p: Heather met the IT person, Jeff  EV: speaker was told by Heather that p   Actually p is attested in the data for both second-hand reports, as in (3.7) and (3.8) above, and for third-hand reports, as in (3.9) where the witness CA relies on information reported by her daughter; specifically, CA’s information was reported by Jeff’s mother to Casey Anthony who then reported it to her own mother, CA.  (3.9) SB: But Jeff’s mom had been sick.  CA: That’s what she said.  SB: Okay. CA: That she was back for cancer treatments.  And she was, she was doing okay, but she was back for um, for some cancer treatments.  And she had been up there since the beginning of the year.  But she was doing fine.  SB: Uh-huh (affirmative) CA: And that she was actually probably going to stay there because um, she had met someone or something up there.  Casey said that her mom Jeff’s mom had    81  met someone up there and was probably gonna stay.  ((72) CA3.186:17-187:3)   p: Jeff’s mom was going to stay (in Jacksonville)  EV: speaker was told by Casey who was told by Jeff’s mom that p  While speakers use actually p to share reported information, the speaker’s use of actually to introduce reported information marks that the speaker believes the source of the report to be trustworthy. In part, I argue on the basis of native speaker intuition that actually encodes that the speaker believes the report to be ‘best’. In addition to my own native speaker intuition, however, I rely on the fact that speakers appear to mark untrustworthy reports explicitly, by using apparently or supposedly; where the speaker does not trust the reliability of the report, the use of actually is unattested but speakers can and do use supposedly.  (3.10) YM: Is there anything else about um, about Casey in particular about Cindy and  Casey’s relationship or even Caylee and Casey’s relationship that you feel is important that I haven’t asked you about that you wanted to share with me?   CC: Um, I’ve often wondered about the story about, and it’s always concerned me.  Is about Caylee’s father you know.  Um, of course you know I know the whole story about first it was supposedly Jesse’s and then it wasn’t.  Um, but then we were told it was a guy from Tennessee.  And later on he was involved in an accident according to… YM: Who told you that? CC: Cindy. (CCr16:22)   p: Jesse was Caylee’s father  EV: speaker was told by Cindy that p   All of CC’s evidence about the identity of Caylee’s father was reported information from Cindy Anthony, and that reported information changed over time. First, Cindy Anthony reported that Jesse was Caylee’s father, but it was subsequently determined that was untrue. She then reported that a guy from Tennessee was Caylee’s father and, because that man died in a car accident, Casey Anthony’s paternity claim against him could not be resolved. While  82 both of the alleged fathers’ identities were reported by Cindy, CC knew the report that Jesse was the father was untrue; he introduces that proposition with supposedly.  In (3.10), the witness knew the prejacent was untrue. However, supposedly p may also signal that the speaker believes their evidence for p is unreliable. Consider (3.11).  (3.11) PR: Do you know of her uhm, a boyfriend of hers or anything like that? DC: Just in rumors.  Yeah (affirmative), I was talking to Jonathan last night and apparently from what he was making it Casey had pulled herself out of their crowd for about the last month, month and a half.  I’m trying to remember the guy’s name.   PR: Pulled away from Jonathan?  DC: That whole crowd.  PR: Just the whole crowd? DC: Yeah (affirmative).  She’s been hanging out with a different crowd.  And uh, she had said supposedly that she was dating a guy named I believe it was Rodrigo. (DaCo9:8-17)   p: Casey was dating a guy possibly named Rodrigo  EV: speaker was told by Casey that p   The witness DC goes on to explain his skepticism about his source of information for p, that is, Casey, as follows.  (3.12) DC: Yeah (affirmative).  Because I asked Jonathan, “Have you ever met the  guy? Is there any way we can get ahold of him?” And Jonathan’s like, “No.  Well she never brought the guy around.”  You know but then when, you know we, we were playing detective last night and uh..  PR: Oh yeah (affirmative) DC: … it was like, at that point as much lying as she’s doing, who did uh, who’s to know if any of it’s true. (DaCo11:1-7)  Conversely, a speaker’s use of actually p where it is clear from the context that they are skeptical of p is unattested in the corpus. I argue this is because speakers can only use actually p where they believe the reported information to be reliable. This accounts for the infelicity of actually in (3.13).  83   (3.13) Context: Belinda, Bob’s wife, told the speaker that Bob is home. Belinda is     a notorious liar. The speaker now tells someone else:   a.     # Bob is actually home. b. Bob is supposedly home.  One might observe that, if supposedly p encodes a speaker’s skepticism about the reliability of their source of information for p then actually p may simply encode the absence of skepticism rather than an expression of reliability per se. However, the data suggest that where a speaker is agnostic about the reliability of the reported information, they use apparently rather than actually as in (3.14) below.   (3.14) DB: Okay.  I was not there when it happened, but I was told she came in um,  passed right by everyone to go sit at her mother’s cubicle.  She had a big coat pulled over her stomach and she went and sat and waited for her mom to get out of a meeting and didn’t speak with anyone.  When her mom came out of the meeting I was told she um, went out to her, the car with her mom.  I guess it was time for her mom to go home.  And apparently from what I heard that was when her daughter told her that she was pregnant um, at that time.  And then about two days later Cindy announced she was gonna be a grandmother.  So she verified her daughter was pregnant. (DB2:19-3:2)   p: Casey told Cindy she was pregnant in a car on the way home from work  EV: speaker was told by Cindy that p   As discussed in chapter 2, each token was classified, among other things, according to the speaker’s evidence for the proposition introduced by actually, apparently, or supposedly. While speakers’ use of apparently p is attested in the police interview transcripts where the speakers appear to be agnostic about the reliability of their source of information for p, speakers’ use of actually p is not. This is consistent with the intuition that, where a speaker cannot vouch for the reliability of the source of their information for p, actually p is infelicitous, as in (3.15).  84 (3.15) Context: Belinda, Bob’s wife, told the speaker that Bob is home. The speaker has  never met Belinda but has heard that she is trustworthy from one friend and a liar from another friend. The speaker now tells someone else:   a.     Bob is apparently home. b.     # Bob is actually home. c.     #  Bob is supposedly home.  Apparently and supposedly will be discussed in more detail in sections 3.3, 3.4 and chapter 5; for now, however, I simply observe that while actually, apparently, and supposedly can be used where the speaker’s information source is reported, actually signals that the speaker believes the information to be trustworthy, apparently signals that the speaker is agnostic about the trustworthiness of the information, and supposedly signals that the speaker believes the information to be untrustworthy.    3.1.1.3 General and encyclopedic knowledge Faller (2002: 19) observes that -mi is licensed for information over which the speaker considers themself to have authority, either because they have a particular expertise or because that information was obtained through cultural immersion and is “known by everyone”. She argues that a speaker may only use -mi with encyclopedic knowledge over which they claim authority. That is, a speaker can only use -mi where they have (i) obtained the information from an authority and (ii) the speaker themself has authority over the information (at p. 135). A speaker will have authority over the information where they are in a position to “respond to challenges” and to “expand on the topic” (at pp. 135-36). Where the speaker is able to claim authority over the information, they are in a position to assimilate the information to their “network of related beliefs” and can introduce that information with -mi  85 (at p. 136). A speaker can use -mi p, then, to indicate they have “incorporated … [p] into his or her belief system” (at p. 138).32  Actually p is attested where the speaker’s information source is general or encyclopedic knowledge in the police interview transcripts.  (3.16) EE: … Um, and all that is, is that’s ah, that’s a photograph taken by our  forensics’ lab.  It’s a close up of that same imagine [sic] that we’re just looking at there that attached photo if you will.  JD: Yes sir. EE: And it’s, and it’s actually a close up of Hope Springs Drive and Suburban Drive.  Um, and that close up imagine [sic] catches some handwritten notes in that same area or close to that same area, where you say that you had parked… ((280) JuD8:11-18)  p: the photo is a close up of Hope Springs Drive & Suburban Drive EV: p is common regional knowledge    (3.17) DI: Um, cadaver’s all about environment.  An, A, how long has it been.  B,  water… my cadaver dog will… it not water certified [JG: sic].  Okay, let me be clear on that.  He’s land certified.  But submerged in water however deep I don’t care if it’s six inches or what the dog will show some reaction.  Water or moisture actually magnifies the dogs’ ability to smell even more.  ((384) DI11:12-16)   p: water magnifies a cadaver dog’s ability to smell  EV: speaker learned that p from an authority   In the data noted above, the speaker uses actually to introduce general knowledge (in (3.16)) and encyclopedic knowledge (in (3.17)). In both cases, the speaker is in a position to claim authority over the information; in (3.16), the speaker was a Texas EquusSearch volunteer who was familiar with both the style of map under discussion and the geographic                                                  32 Faller argues that the notion of assimilation is relevant not only for general/encyclopedic knowledge but also for personal information (2002: 140). This will be discussed in greater detail in section 3.2.   86 area that map purported to represent, while in (3.17), the speaker is a professional dog-handler. Thus, both speakers are able to use actually to felicitously introduce the information.   3.1.1.4 Speaker performance  So far, I have shown that actually occurs in the same evidential contexts as Cuzco Quechua -mi. However, actually differs from -mi in that actually is also licensed where the speaker’s evidence for p is their own personal performance of p. Faller (2002) observes that this is not the case for -mi. Nonetheless, the felicity of actually p where the speaker’s evidence for p is speaker performance is consistent with the analysis of actually as marking best evidence.   Speaker performance evidentials are independently attested cross-linguistically. For example, Oswalt (1986) shows that, in Kashaya, the evidentials -wêla and -mela indicate that the speaker’s evidence for p is the speaker’s own performance of p.  (3.18) Kashaya evidentials a.  qowá-qala (Performative imperfect: -wêla)  ‘I am packing (a suitcase).’  b. qowáhmela (Performative perfect: -mela)  ‘I just packed.’  c. qowá-qh (Factual Imperative: -ŵă)  ‘(I see) she is packing.’33  d. qowahy (Visual Perfective: -yă)  ‘(I just saw) he packed, I just saw him pack.’  e. mod.dun (Auditory: -V nnă)  ‘I hear/heard someone running along.’                                                   33 It is unclear on its face how (3.18c) is an imperative however nothing specifically turns on this for my analysis.   87 f. mu cohtochqh (Inferential: -qă) ‘He must have left, he has left’ (Said on discovering that the person is no longer present; the leaving itself was not seen […] nor heard)  g. mul =ído-. hayu cáhno-w (Quotative: -do)  ‘Then, they say, the dog barked.’ (adapted from Faller, 2002: (32))  Based on the above distinctions, Kashaya has been analysed as having five evidentials: Performative, Factual-Visual, Auditory, Inferential, and Quotative (Oswalt, 1986; Faller, 2002).  In terms of classifying these evidentials into evidence types, Faller notes, the speaker performance may “be said to indicate a subtype of direct evidence, in fact it might be said to be the most direct evidence possible, since the speaker knows something because (s)he is doing it” (2002: 46). As a result, Faller proposes the following cline of personal evidence.  (3.19) The personal evidence cline: PERFORMANCE > VISUAL > AUDITORY > OTHER SENSORY > REASONING > ASSUMPTION  (Faller, 2002: (34)) The personal evidence cline represents a continuum of evidence types with incremental differences in terms of the amount of reasoning required to make a statement. Faller (2002: 50) argues that performance and visual evidence require little or no reasoning. She further argues that only those evidentials that cover a continuous area on this cline are possible evidentials. In her analysis, Quechua -mi covers the following personal evidence types:34  (3.20) Quechua -mi: PERFORMATIVE > VISUAL > AUDITORY > OTHER SENSORY > REASONING > ASSUMPTION                                                   34 Faller proposes another cline, the Mediated Evidence Cline, which functions alongside but independently from the Personal Evidence Cline, to account for reported evidence and hearsay/folklore. This will be discussed in greater detail in section 3.2.2.1.    88 Unlike -mi, actually p is felicitous where the speaker’s evidence is their own performance of p; speakers can use actually p to indicate they did p. This is demonstrated in (3.21) through (3.23).  (3.21)  EE: Now, can you describe the attached map that you’re looking at there? JD: It’s a large overview of the area um, of Suburban… It shows the area that we actually were able to search. ((277) JuD5:6-9)   p: the searchers were able to search a particular area  EV: speaker performed p  (3.22) JA: I see.  So um now wow, gosh, so she sent you (inaudible), cause seem like  she sent you a bunch and I guess, do you remember sending her more than one or did you only send her the one? RM: I sent her more than one.  Actually I sent her, yeah, I sent, yeah I have sent more than that. ((516) RM2.9:10-14)  p: speaker sent more than one email  EV: speaker performed p    (3.23) CA: … And last night I received a phone call at 4 o’clock somebody from  Seattle, Washington called me with a thought you know.  Waking me up at 4 o’clock in the morning.  Finally I’d actually fallen asleep for a night where I could’ve gotten more than 2 or 3 hours together.  So I, I asked her if it was something pertinent and she said she thought it was… ((10) CA3.18:19-24)   p: speaker had fallen asleep  EV: speaker performed p  Given the fact that actually p is felicitous where the speaker performed p, I propose that English actually differs from Quechua -mi by including an additional contiguous evidence type (performative) in the area of the personal evidence cline it covers, as shown below.   (3.24) English actually: PERFORMANCE > VISUAL > AUDITORY > OTHER SENSORY > REASONING > ASSUMPTION  89 3.1.2  Actually does not occur where must is felicitous  As noted in Section 3.1.2 above, actually p is unattested in the police interview transcripts where the speaker’s evidence for p is a source they consider to be unreliable or untrustworthy or where the speaker is agnostic about the reliability of the reported evidence; in those cases, supposedly and apparently are attested (respectively) and actually is not, as shown in section 3.1.1.2. Similarly, actually p is generally not used in contexts which would license must p. Rather, where the speaker’s evidence is inference or reasoning, speakers use must p as opposed to actually p. 35   Actually is unattested in discourse contexts where von Fintel & Gillies (2010) and Matthewson (2015a, 2015b) show that must is licensed; namely, where the speaker’s information is inference, reasoning, or an untrustworthy report.36 Consider the following examples.  (3.25) LA: Oh no, I take that back.  Uhm, my mother and father told me that my dad  actually drove the car back because my mother was talking about how she didn’t know how my dad survived because the smell was so bad. (laughs) EE: Did she follow him in the other car or? LA: That must have been how they did it. I haven’t really asked all those intricacies of how that went down. ((180) LA64:19-24)  p: mom followed dad home in the other car  EV: speaker reasoned that p  In (3.25), the speaker, LA, knows his father drove the car home from the towing company lot because he was told so by his parents. His knowledge source is a trustworthy report and                                                  35 This is an oversimplification. There are certain contexts in which actually is used to introduce information the speaker knows by way of reasoning. Specifically, where speakers have assimilated the information into their belief system, they can introduce that information with actually, as I will discuss in section 3.2.2.2.   36 Although see section 5.3 for discussion of contexts where must and actually co-occur.   90 actually is felicitous in order to introduce that piece of information. However, with respect to the proposition ‘LA’s mother followed his father home in another car,’ his information is based on reasoning; it is clear from the following comment (I haven’t really asked all those intricacies of how that went down) that p was not discussed in his conversations with his parents. Rather, he knows p through reasoning and introduces it with must.  Similarly, in (3.26), the interviewer is reviewing Casey Anthony’s cell phone records which show a series of telephone calls between her and the witness, AL. AL indicates that he has no recollection of what was said, if anything, during the telephone calls over the course of a particular evening. The interviewer notes that a particular phone call, at 0419, lasted 58 seconds. He reasons that AL and Casey Anthony either spoken briefly or she left him a message, given the length of the phone call.   (3.26) EE: You get a call, she calls you at 0418…  AL: Uh-hum (affirmative)  EE: … with no connection.  And then she gets ahold of you at 0419 with 58 seconds.  Like you guys talked for a minute, or she must have left you a message.   AL: Okay.  (AL3-26:4-8)   p: Casey left AL a message  EV: speaker reasoned that p  In (3.27), the speaker is reporting a conversation she had with a restaurant hostess in which the speaker is trying to determine the identity of a particular waitress who worked at the restaurant and who served the speaker on an earlier occasion.  (3.27) CM: Okay, and the waitress’ name. Did you get the waitress’ name there? RF: When I called back, I couldn’t remember her name, but when I called back over and I talked to a hostess that was there and I described her to a T.  And I remember during the conversation because we were talking about children.  And, and she was saying she couldn’t even fathom that a mother would do  91 that to her children, that she had two young boys.  She was in her mid-twenties.  So, when I described that back to the hostess.  She said, “Oh, the [JG: sic] must be Heidi.”  CM: You jump…  RF: So, I’m assuming Heidi is her name. ((S5) RF3:6-4:5)   p: the waitress RF is referring to is Heidi  EV: speaker (the hostess) reasoned that p  The hostess’s information source for p is reasoning; that is, she reached her conclusion on the basis of the description RF gave her. Interestingly, RF’s explanation above was offered in response to the interviewer’s question about the waitress’s name. RF explains she did not get the name directly and learned it only through the hostess’s reasoning.  When the interviewer asks her about the waitress’s name, RF appears compelled to explain how she learned the information and concludes with the qualified statement So, I’m assuming Heidi is her name. That is, RF does not treat the hostess’s information as a trustworthy report, introducing the statement “Heidi is her name” neither with a bare assertion nor with actually.37  In (3.25) through (3.27), the source of the speaker’s information for p is reasoning and in each case, the speaker introduces p with must. This is consistent with von Fintel & Gillies’ (2010) analysis of must as marking information indirectly obtained. However, von Fintel & Gillies show that must will be felicitous as long as the speaker’s knowledge of p requires any reasoning. For that reason, must can be felicitous with speaker performance even though speaker performance is generally considered to be ‘strong’ evidence (in an evidence strength sense) or ‘direct’ evidence (under an evidence type analysis).  Von Fintel & Gillies (2010) argue that, in order for must p to be felicitous, the speaker must not have evidence that settles p directly. I argued above that speaker                                                  37 Reported information will be discussed in greater detail in sections 3.3, 3.4, and chapter 5.   92 performance is a type of direct evidence that licenses actually. However, it is clear from von Fintel & Gilllies’ discussion that must is only felicitous with speaker performance where the speaker requires some reasoning for p, despite having performed p; for example, where the speaker forgot that they performed p or was incoherent at the time they performed p. In these circumstances, the speaker’s evidence for the fact that they performed p requires some reasoning. By way of example, they give the following.   (3.28) Context: Suppose Billy is at the zoo standing in front of the big gazelle-plus-antelope enclosure. She can see the animals off in the distance towards the other end. Later, she says:  I must have seen gazelles. (adapted from von Fintel & Gillies, 2010: 36)  p: I saw gazelles  EV: speaker reasoned that p Although von Fintel & Gillies do not discuss it necessarily in these terms, they suggest that must p is felicitous in the context above because, while the speaker performed p, the speaker’s best evidence for p is indirect; that is, the speaker saw gazelles without being directly aware of the fact she was doing so. As shown in (3.29), a similar use of must is attested in the police interview transcripts; specifically, where the speaker knows they did p but only from indirect information, the data show they can use must p.   (3.29) BC: Uhm, let me see if I got Bill, Bill Fitzgerald.  I might still have his  number.  I believe Bill is still involved with Texas EquuSearch.  Uh, I don’t see it here anymore.  I must have deleted.  (BrC16:17-19)   p: I deleted Bill’s number  EV: speaker reasoned that p  93 It is clear in the example in (3.29) that the speaker does not remember deleting the phone number; rather, his evidence for the fact that he did so is (a) that the phone number is no longer listed in his phone and (b) the only explanation for the missing number is that he deleted it. Therefore, the source of his information for p is indirect (reasoning), despite the fact that he performed p. Thus, the data in (3.29) is consistent with von Fintel & Gillies’ analysis of must as marking indirect evidence.   On the other hand, a speaker whose evidence for p includes any reasoning generally will not use actually to introduce that information; actually p is unattested in this context in the police interview data. However, as I show in section 3.2.2.2, a speaker may use actually, even if their source of information initially requires reasoning, if that information has been assimilated into their belief system; that is, they can use actually p in (3.28) and (3.29) if some time has passed since they reasoned that they performed p and if no contradictory information has come to light in that period. Assimilation is discussed in more detail in section 3.2.2.2 but for now, it suffices to say that a speaker who has not assimilated p into their belief system will use must and not actually to introduce p where their evidence for the fact that they performed p includes reasoning.    Must (and not actually) is felicitous any time the speaker’s knowledge for p includes some reasoning. This is true not only where the speaker performed p, as shown above, but also where the speaker has sensory evidence for p. Although sensory evidence is categorized as ‘strong’ in an evidence-type analysis, a speaker can use must if the sensory evidence they have for p does not, on its own, directly settle p. In those cases, some reasoning is still required in order for the speaker to know p. Thus, must will be licensed in the circumstances, as shown in (3.30).  94  (3.30) Context: A speaker sees water drops on the window and knows that means that either  the sprinkler is on or it is raining. The speaker turns off the sprinkler and the window continues to be covered in water drops.   It must be raining.38  The speaker can say It must be raining felicitously because they know p (it is raining) by a combination of visual evidence and reasoning, and crucially, not visual evidence alone.  Must is felicitous where the speaker’s evidence for p includes some reasoning, while actually is felicitous where no reasoning is required. Therefore, despite at first appearing to be contradictory, the examples in (3.28) through (3.30) are consistent with both von Fintel & Gillies’ analysis of must and my analysis of actually. In section 3.2, I set out my analysis of actually and show how it both captures and correctly predicts the empirically observed uses.   3.2 The semantics of actually  Given the empirical similarities between Cuzco Quechua -mi and English actually, Faller’s (2002) analysis of -mi is an appropriate jumping off point for a formal analysis of actually. As noted in chapter 1, Faller analyses -mi as marking best possible grounds. She defines best possible grounds as the best possible evidence available to the speaker in the context. Where the context is appropriate, best possible grounds can include sensory observation, trustworthy reports, or general/encyclopedic knowledge. Thus, she defines -mi as encoding the evidential value that the speaker possesses the best possible source for the type of information conveyed by the utterance (2002: (90)). I adopt Faller’s idea of best possible grounds in my analysis of actually although I use Matthewson’s terminology to describe it.                                                  38 Thank you to Hotze Rullmann (p.c.) for drawing this example to my attention.    95  Faller’s analysis of -mi contains an additional formal component which I have refrained from including for actually. Specifically, Faller analyses -mi as an illocutionary operator which modifies the sincerity conditions of a bare assertion. It is widely accepted in the speech act literature that, in order to make a bare assertion, the speaker must believe the assertion to be true (Vanderveken, 1990). Therefore, a bare assertion has the illocutionary force of asserting that the speaker believes the assertion. Faller argues that, by adding -mi, the speaker modifies the sincerity condition of a bare assertion by adding the condition that the speaker has the best possible grounds for the assertion. Her definition of -mi, then, is as follows.  (3.31)   ASSERT(p)   ASSERT(p)       -mi:    ⟼    Sinc = {Bel(s,p)}  Sinc = {Bel(s,p), Bpg(s,Bel(s,p))}  (adapted from Faller, 2002: (130))  Thus, for Faller, -mi not only encodes that the speaker has the best possible grounds for asserting p but also functions to modify the sincerity conditions on a bare assertion to explicitly mark that the speaker believes p and that they have the best possible grounds to do so.   I analyse actually as encoding direct evidence strength (that is, that actually p encodes that the speaker has best evidence for p) but refrain from analysing actually as an illocutionary operator. Because evidence for the illocutionary analysis is not surface-obvious, it is not possible to perform the relevant tests on data obtained from a corpus like the police interview data. Therefore, I have not attempted to perform such tests. I prefer to remain  96 agnostic as to whether the formal definition of actually requires reference to its illocutionary force and leave that issue to future research.39   Von Fintel & Gillies (2010) analyse must not as an illocutionary operator but rather as an epistemic modal. Thus, their analysis of must contains two components: must is an epistemic necessity modal and must requires that the speaker’s evidence for p be indirect. With respect to its evidential contribution, von Fintel & Gillies’ analysis is premised on the idea of ‘privileged information’ which they define as information that the speaker treats as being both direct and trustworthy (2010: 369). Where the speaker’s privileged information (alternatively referred to as the ‘kernel’) establishes p, then it “directly settles” p and must is infelicitous. Where the information itself does not establish p, it “indirectly settles” p and must is felicitous.  Importantly, the question is not whether or not p is true; von Fintel & Gillies convincingly show that must p entails p, therefore p is always true. They show:  (3.32) If Carl is at the party, then Lenny must be at the party.  Carl is at the party.  So: Lenny is at the party.   If , must    (adapted from von Fintel & Gillies (2010): (14) and (15))  Since must p entails p, the privileged information in their analysis is that which proves p. The question is whether that information proves p directly or indirectly. They show that, where the privileged information proves p indirectly, must is felicitous. von Fintel & Gillies use the following examples by way of illustration.                                                  39 See Portner (2006) for further discussion and criticism of Faller’s appeal to speech act theory in her analysis of -mi.  97  (3.33) [Seeing the pouring rain]  a.  It’s raining  b. ??It must be raining (von Fintel & Gillies (2010: (6))  (3.34) [Seeing wet rain gear and knowing rain is the only possible cause]  a.  It’s raining  b. It must be raining (von Fintel & Gillies (2010: (7))  In both of the examples, p (it’s raining) is true; what is at issue is how the speaker knows that it is raining. In (3.33), the speaker’s information is sensory observation of the event itself; it directly settles p in the kernel. Because must is only felicitous where the kernel does not directly settle p, must is infelicitous in (3.33). On the other hand, in (3.34), the speaker knows p by sensory observation of some result of the event itself (wet rain gear) and reasoning (rain is the only possible cause of wet rain gear). The kernel does not contain information that directly settles p, therefore, must is felicitous.   von Fintel & Gillies define must, therefore, as having two lexical components:   (3.35)  Fix a c-relevant kernel, K: i. [[ must ϕ ]]c,w is defined only if K does not directly settle [[ϕ]]c  ii. If defined, [[ must ϕ ]]c,w = 1 iff BK    [[ϕ]]c  (adapted from von Fintel & Gillies, 2010: 372)  The lexical component in (i) reflects must’s evidential contribution by requiring that the kernel not directly settle . The component in (ii) reflects must’s modality by providing the universal quantification.40                                                  40 Specifically, von Fintel & Gillies (2010: 371) define kernels and bases as follows:  (Kernels and bases) K is a kernel for BK, BK is determined by the kernel K, only if: i. K is a set of propositions (if P  K then P  K) ii. BK =  K  98  I suspect, as Faller does for -mi, that actually is not an epistemic modal, however, I remain agnostic in this regard. Like with the tests for illocutionary force, the tests to determine whether or not something is a modal (see, for example, Papafragou, 2000, 2006; Faller, 2002; Matthewson et al., 2007) are not surface obvious and cannot be conducted on corpus data. Since I advance no argument one way or the other with respect to epistemic modality, I simply note that my analysis of actually does not include a modal analysis equivalent to von Fintel & Gillies’ (3.35(b)).   3.2.1 Actually encodes ‘best’ evidence  I analyse actually solely with respect to its evidential contribution and remain agnostic as to whether it is an illocutionary operator (as Faller argued for -mi) or an epistemic modal (as von Fintel & Gillies argued for must). I argue that actually encodes a direct value on the evidence strength dimension of meaning following the categorization system developed in Matthewson (2015a, 2015b). Matthewson argued that, regardless of the terminology used in Faller (2002) and von Fintel & Gillies (2010), both were really advocating that -mi and must, respectively, encode evidence strength rather than evidence type. She argues that, by marking best possible grounds, -mi encodes a direct value for evidence strength and, by marking that p is not directly settled by the kernel, must encodes an indirect value. I follow Matthewson and refer to direct evidence strength as ‘best’ evidence and indirect evidence strength as ‘not best’ evidence. I argue that actually encodes that a speaker has best evidence for the proposition it introduces.  In order to make sense of this, we must consider what makes evidence ‘best,’ as opposed to ‘not best.’ In other words, we must consider what evidential contexts are encoded by evidence strength. Clearly evidence strength involves more than simply a consideration of  99 the type of evidence available to the speaker; if evidence type were the deciding factor, we would not need to distinguish type and strength as separate dimensions of meaning. Matthewson’s (2015a, 2015b) cross-linguistic data show that evidence type and evidence strength are distinct and should be analysed differently.  I argue that the difference between evidence type and evidence strength is the role played by context. For evidence type, the only relevant factor in determining which evidential marker is appropriate is the kind of information the speaker has for the prejacent: the direct, sensory evidential marker is appropriate where the speaker observed the event, the reportative marker is appropriate where the speaker was told about the event, etc. The discourse context is less relevant in an evidence type analysis.41 In an evidence strength analysis, on the other hand, evidentials mark whether or not the speaker’s evidence for p is the best available evidence in a particular context; that is, evidence strength requires a consideration of both evidence type and context. Whether a particular type of evidence is the most reliable available evidence in a particular context requires an exercise of judgment on the part of the speaker.  With this in mind, I define actually as follows.  (3.36) i.   [[Actually φ]]c,w is felicitous iff the speaker in c has best evidence for [[φ]]c,w, in c.  ii.  If [[actually φ]] c,w is felicitous in c, then [[Actually φ]] c,w = 1 iff [[φ]] c,w = 1.  Of course, the definition in (3.36) does not formalize what constitutes the best evidence type in a particular context; rather, an evidence strength analysis relies heavily on speaker judgment.                                                  41 Context may nonetheless be relevant in evidential selection, depending on the speaker’s interactional goals in the discourse. This will be discussed in greater detail in chapter 4.   100 Both Faller’s and von Fintel & Gillies’ analyses similarly rely on speaker judgment, although neither formalizes that concept. Indeed, it is unclear how one could formalize the idea of speaker judgment, since judgment will, by definition, be subjective. Faller observes that what constitutes best possible grounds is context-dependent and will vary on a case-by-case basis. Although she did not explicitly say so, it is clearly the speaker who decides whether their evidence constitutes best possible grounds so as to licence -mi. While the addressee may disagree with the speaker’s assessment of the evidence, the speaker’s judgment about evidence strength is expressed by their use of the best possible ground evidential. To that extent, the idea of speaker judgment is implicitly built into Faller’s analysis of best possible grounds. Further, she explicitly builds speaker judgment into her analysis in her treatment of trustworthy reports. Recall that, for Faller, trustworthy reports constitute best possible grounds in appropriate contexts. Where they do, speaker judgment operates on two levels; if, in the speaker’s judgment, reported information is the best possible type of evidence in the circumstances and if, in the speaker’s judgment, a report is trustworthy, it can license -mi. Therefore, speaker judgment is the basis for determining whether certain evidence constitutes best possible grounds in a particular context in Faller’s analysis.  Von Fintel & Gillies’ (2010) analysis similarly relies on speaker judgment about whether or not privileged information directly settles p and, as a result, licenses must. Specifically, the kernel contains trustworthy evidence, and the speaker decides what counts as trustworthy. Von Fintel & Gillies (2010: 369-70) observe that an analysis, like theirs, which depends in part on speaker judgment will involve some context dependency and vagueness. They observe:  101  [j]ust about everyone will treat the direct visual observation of pouring rain as a piece of privileged directly evidenced information. But a professional epistemologist, even when on vacation in Seattle, might be tempted to say: “Well, I am getting the kind of visual input that is only consistent with rain, so it must be raining.” It is only us simpler folks that can do without the epistemic must and might at least once in a while. (page 370)  By leaving the assessment of evidence strength to speaker judgment, we resign ourselves to the conclusion that there will be context-dependent variation between speakers as to whether a particular evidence type in a particular context constitutes best evidence. I predict that speakers may have different judgments about what types of evidence will qualify as best in a particular context. In the next section, I show that this prediction is borne out in the data.   3.2.2 Evidence strength and speaker variation   Following Matthewson (2015a, 2015b), I argue that evidence strength refers to the strength of the speaker’s evidence for the proposition in a particular context; by contributing a direct value for evidence strength, actually p signals that the speaker has ‘best’ evidence for p and by contributing an indirect value, must p signals the speaker has ‘not best’ evidence for p. This analysis predicts that speakers will vary in terms of how they assess evidence strength in two crucial ways. First, different speakers can have differing intuitions about the reliability of certain information sources; some speakers may value certain sources of evidence more highly than other speakers do. Second, an individual speaker’s intuition about the reliability of information may change over time. These predictions are borne out empirically for English in the police interview transcript data.   3.2.2.1 Inter-speaker variation and the assessment of ‘best’ evidence  102  If evidence strength marks that the speaker has best evidence for their proposition in a particular context, then we predict that different speakers may have differing judgments about the same evidence and that this difference will be reflected in their linguistic choices. Specifically, where speakers differ with respect to whether they view certain information sources as reliable, they will differ in terms of whether evidence gleaned from those sources can license actually. This prediction is borne out in the data.   This type of variation is attested in particular with respect to the interviews of Casey Anthony’s parents, Cindy Anthony and George Anthony. It is clear from the transcripts that Cindy Anthony and George Anthony have the same evidence type for much of their information; that is, for much of their respective interviews, both are reporting what was reported to them by their daughter. However, it is equally clear from the transcripts that George Anthony is significantly more skeptical of his daughter’s reliability than Cindy Anthony is. The data in (3.37) and (3.38) show that this difference is reflected in how they give their information to the interviewer.   In the course of the investigation, both Cindy Anthony and George Anthony provided interviewers with information about Casey Anthony’s relationship with a man named Jeff Hopkins, who had a son named Zachary. Neither George Anthony nor Cindy Anthony had ever met Jeff Hopkins or his son; indeed, they knew nothing about them except what they heard from Casey. The Anthonys gave their information about Jeff and Zachary in very different ways, indicating a stark difference in terms of how they viewed the reliability of their information source; Cindy Anthony reports that Jeff has a son named Zachary as a bare assertion (in (3.37)) while George Anthony uses supposedly to report the same information (in (3.38)).    103  (3.37) a. CA: And Casey and Jeff were friends, Jeff had a um… a child um a son that  was almost a year older than Caylee and he was dating this person Zany at the time … (CA1.68:12-14)         b. CA: … Jeff is the one that has Zachary. (CA1.67:22)   p =  ‘Jeff has a son named Zachary.’  EV =  speaker was told p (by Casey Anthony)   (3.38) GA: Jeff’s… Jeff’s son supposedly is…  SB: Zack?  GA: Yeah …  (GA4.52:18-20)   p =  ‘Jeff has a son named Zachary.’  EV =  speaker was told p (by Casey Anthony) Supposedly will be discussed in greater detail in section 3.4 and chapter 5. For present purposes, it is enough to observe that, intuitively, it is clear from (3.37) and (3.38) that Cindy Anthony and George Anthony had different instincts about the reliability of Casey Anthony’s report and this difference resulted in different linguistic choices in terms of how they presented their information to the interviewer.   While the distinction between the evidence given in (3.37) and (3.38) involved a contrast between a bare assertion and supposedly, the same distinction is observed with actually and supposedly. The data in (3.39) through (3.41) involve witnesses reporting what they know about Casey Anthony’s job. At one point, Casey was employed by Kodak to take pictures at Universal Studios. It becomes clear through the course of the interviews that Casey allowed people to continue to believe she worked for Kodak at Universal Studios long after her employment ended. Cindy Anthony reports her evidence with actually (in (3.39))  104 marking that her information about Casey’s job at Kodak, which she knows from Casey, is reliable.     (3.39) CA: That was through Kodak she worked…  SB: Right. CA: … directly through them.  It’s not a Universal employee even though they went to Universal had Universal badges.  She was actually employee of Kodak. ((52) CA3.122:13-16)  p =  ‘Casey worked at Universal but she was an employee of Kodak.’  EV =  speaker was told p (by Casey Anthony) Alternatively, the witnesses MM (in (3.40)) and RP (in (3.41)) report the same information with supposedly, indicating that their intuition about the reliability of that information is different than Cindy Anthony’s, despite having the same source.  (3.40) YM: Did um, did you know at the time what Casey did for a living, what, what  her job was? MM: Uh as far as, supposedly she worked for the um Universal... (MM3:14-16)   p =  ‘Casey worked at Universal.’  EV =  speaker was told p (by Casey Anthony)  (3.41) YM: Can you tell me a little bit about where she used to work uhm, while, I  guess we’ll start with last year because I don’t want to go too, too far back.  RP: Uhm, well I mean uh, that whole time supposedly she was working at Universal. And that’s the only thing she ever told me…   (RP12:8-13:2)   p =  ‘Casey worked at Universal.’  EV =  speaker was told p (by Casey Anthony) The data show that speakers can vary in terms of their intuitions about the reliability of their information sources. In each of (3.39) through (3.41), the speaker only knows what Casey Anthony has reported to them. Where the speaker believes Casey Anthony to be a reliable  105 source of information, they use actually (in (3.39)) and where the speaker does not, they use supposedly (in (3.40) and (3.41)).  Importantly, all of the examples given above involve reported information. I argue that speakers will vary most with respect to their intuitions when their information sources are reports or reasoning. Recall Faller’s (2002) personal evidence cline from section 3.1.   (3.42) Faller’s personal evidence cline          The personal evidence cline: PERFORMANCE > VISUAL > AUDITORY > OTHER SENSORY > REASONING > ASSUMPTION  In the personal evidence cline, Faller sets out the evidence types that require no reasoning or assumption (on the left) to those requiring purely reasoning and assumption (on the right), with varying degress of reasoning required for those types in the middle. Similarly, she proposes the cline in (3.43) to account for reported evidence, with the most directly reported evidence on the left and the least on the right. Like with the personal evidence cline, Faller argues the evidence on the leftmost edge of the mediated cline is more direct than that on the right.  (3.43) Faller’s mediated evidence cline          The mediated evidence cline: DIRECT > SECONDHAND > THIRDHAND > HEARSAY/FOLKLORE  (adapted from Faller, 2002: (58)) Faller’s evidence clines reflect Willett’s (1988: 86) observation that attested evidence is ranked as the most reliable evidence source and assumptions as the least reliable evidence source, with reasoning somewhere in the middle. To paraphrase von Fintel & Gillies (2010: 370), just about everyone will treat the direct visual observation of pouring rain as reliable  106 evidence. As Karttunen (1972: 13) observed, ‘direct’ knowledge that requires no reasoning is generally valued as most reliable.42 Where the speaker’s information for p is on the left edge of the personal evidence cline, little if any reasoning is required in order for the speaker to know p. The further the evidence type is to the right of the cline, the more reasoning is required. This is similarly reflected in the mediated evidence cline; while the evidence types to the left of the cline are more ‘direct’ than those on the right of the cline, all mediated evidence requires the speaker to make a judgment about the reliability of the information source.  I predict that evidence types which require little to no reasoning will be uniformly treated as ‘best’ across speakers and that assumption will be uniformly treated as ‘not best;’ that is, I predict little variation between speakers with respect to the left edge of the clines (which I predict will unanimously be treated as ‘best’) and the right edge of the clines (which I predict will unanimously be treated as ‘not best’). I further predict that most speaker variation will occur where some reasoning, but not reasoning alone, is required on the part of the speaker. That is, I predict that speakers will vary primarily with respect to the types of evidence that fall in the middle of both the personal evidence cline and the mediated evidence cline. This prediction is consistent with the data in the police interview transcripts.43                                                  42 While direct sensory observation is widely believed to be reliable, eyewitness evidence has been consistently shown to be subject to witness bias, police influence, and the classic forgetting curve (see Kassin, 2006, and studies cited therein; see also Watt, 2010 and Fraser et al., 2011 for similar challenges to ear-witness testimony). It nonetheless stubbornly remains the benchmark for reliability by laypeople and at law.  43 This claim is perhaps bolder than the data justify. The data are consistent with, but do not prove, the prediction. Because evidentials are not obligatory in English, speakers regularly make statements without actually even where they have best evidence for p.   107  There are 530 tokens of actually attested in positive statements in the police interview transcripts and for which the speaker’s source of evidence can be determined. The total breakdown of actually tokens used for evidence sources is given in Table 3.1.   Evidence source Tokens Total tokens  Performative 255 255 (48%) Sensory                 Visual Auditory Olfactory Unspecified44 100 178 (34%) 37 1 27          Reported            73 73 (14%) Knowledge             General Enclyclopedic Regional 11 24 (4%) 9 5 TABLE 3.1; Tokens of actually by evidence type (where source of information       can be determined)  Table 3.1 shows that the use of actually to indicate that the speaker’s evidence is performative is the most widely attested in the data. Of course, the numbers are merely suggestive. To show the "cline" conclusively, I would have to also supply statistics on how frequently statements about speaker performance, visual information, auditory information, etc. occur in the corpus overall (including sentences without actually). The fact that more tokens of actually occur in examples of speaker performance may simply result from the fact that speakers talk about what they did more often than they talk about what they observed. With that caveat in mind, Table 3.1 nonetheless suggests that speakers use actually robustly where their evidence is sensory. Speakers use actually less often where the speaker’s evidence is reported or general knowledge. It is interesting to observe that speaker use of                                                                                                                                                         44 Where the speaker directly observes an event, they may witness it on multiple sensory levels (they may see it and hear it at the same time). Where it is unclear which sense was predominant in the circumstances, I have marked the token as ‘unspecified’.    108 actually in the police interview transcripts appears to mirror Faller’s (2002) personal and mediated evidence clines; that is, the further to the left the evidence source is on the evidence clines, the more tokens of actually occur for that evidence source in the data.  Conversely, I predict speakers will consider information sources at the right edge of the personal evidence cline, where information is known purely by assumption, to be ‘not best.’ This also is consistent with the police interview transcript data; there are no attested examples in which a speaker uses actually p where their information source for p is assumption.45 It is in particular with respect to the evidence types that require some inference (but not entirely inference) where I predict speakers will vary the most in terms of their intuitions about reliability. This is consistent with the English data in the police interview transcripts, which showed that speakers varied with respect to their assessment of the reliability of reported information.   3.2.2.2 Intra-speaker variation and time of utterance My analysis predicts a second type of speaker variation; namely, that a single speaker may have different intuitions about the strength of their evidence at the time it is initially received as opposed to the time at which they are reporting it (the time of utterance). A speaker may reassess their initial intuition about the strength of their information once they have had an opportunity to increase their factual knowledge (Hintikka, 1962: 8) or after a period of time in which no contrary evidence comes to light (Faller, 2002: 139). Explicit discussions of speaker judgment about information sources over time are rare in the literature. This is likely because most work on evidentials has focused on the traditional split                                                  45 Though few, there are examples of speakers using actually with information they know by reasoning. These will be discussed in sections 3.2.2.2, 4.1.3.3 and 5.3.   109 between direct and indirect evidence type; under this type of analysis, reported information is still reported information, regardless of whether the speaker heard the reported information the day before or the year prior.46 To account in part for the change in speaker perspective about the quality of their evidence over time, Waldie (2012) argues for the concept of ‘origo’ as a central component of evidentiality. He defines the origo as a pair consisting of the judge (i.e., the person who “has the evidence”) and the situation (at p. 115). He argues that evidential meaning is determined in part based on the identity of the origo. Different origos exist where there are different judges in a particular situation or where the same judge is evaluating the same evidence in different situations. That is, he argues that “as an origo learns about the world, his or her perspective will change, adding or removing … propositions” to their belief system (at p 122). Thus, in his analysis, the evidence-holder is associated with one origo at the time they acquired the information and another at the time of utterance, where their opinion about their evidence has changed. Similarly, Faller (2002) observes that, for Cuzco Quechua, information initially licensed by a reportative can at a later point in time be licensed by -mi if the speaker has assimilated the information into their belief system. This would be unexpected if the evidential system cared only about evidence type and not a speaker’s assessment of reliability.   Faller initially discusses assimilation in the context of encyclopedic information. She argues that a speaker need not have direct evidence to use -mi to convey encyclopedic information, as long as the speaker has authority over the information they proffer (2002: 135). Minimally, to have authority over encyclopedic information, a speaker must (i) have obtained the information from an authority and (ii) be able to respond to challenges and to                                                  46 I suspect that, even with evidence-type evidentials, speakers who assimilate the reported knowledge will no longer be required to use a reportative. More will be said on this below.   110 expand on the topic if necessary (2002: 135-36). It is the second requirement she refers to as ‘assimilation;’ a speaker cannot claim authority over encyclopedic information unless that information is connectable to a network of the speaker’s related beliefs (2002: 136). The use of -mi to introduce encyclopedic knowledge, then, signifies that the speaker has integrated that fact into their belief system (2002: 138).     Faller (2002: 139) observes that the concept of assimilation is relevant for personal information as well as encyclopedic information. As an example, she gives the following.  (3.44) Context: My sister tells me on the phone that my brother was sent to Italy on  assignment for a week.   (a) Tura-y-qa Italia-pi-s llank’a-sha-n  kay  semana-pi   brother-1-TOP Italy-LOC-REP work-PROG-3 this week-LOC      p = ‘My brother is working in Italy this week.’   EV = speaker was told (=by speaker’s sister) that p   (b) Tura-y-qa Italia-pi-n llank’a-sha-n  kay  semana-pi   brother-1-TOP Italy-LOC-mi work-PROG-3 this week-LOC      p = ‘My brother is working in Italy this week.’   EV = speaker has best possible grounds for p  (adapted from Faller, 2002: 139) Faller observes that, if the speaker wishes to convey the information about her brother immediately after receiving the news, she has to use the form in (3.44a); however, by the following day, she can use the form in (3.44b). She argues that this is because -mi can be used after a certain time has elapsed after acquiring the information from a reliable source and if no evidence to the contrary has emerged in that time (p. 139). Faller (2002: 140-41) argues that -mi not only indicates that the speaker has the best possible source of information but that this information has been assimilated into the speaker’s belief system; having the  111 best source of information and having assimilated that information into a network of beliefs together constitute best possible grounds (2002: 141).   This analysis, in which -mi is licensed where the speaker (a) has the best source of information and (b) has assimilated that information into their network of beliefs, is problematic for a number of reasons. First, the dual requirement for licensing -mi is inconsistent with Faller’s analysis of best possible grounds, which she specifically argues includes trustworthy reports. If information learned from trustworthy reports can independently constitute best possible grounds, then it should not need to be assimilated in order to be licensed by -mi. That is, if the speaker’s sister is a reliable source, then her information should not require assimilation in order to constitute best possible grounds. Recall from (1.13c) (repeated below) that Faller argued on empirical grounds that reliable reports were independently licensed by -mi.  (3.45) Lima-ta-n viaja-n.    Lima-ACC-mi travel-3  p = ‘She travelled to Lima.’    EV = speaker was told by her (= speaker’s sister) that p (Faller, 2002: (10(c)) If trustworthy reports constitute best possible grounds (in appropriate contexts) for the purposes of licensing -mi, and the speaker’s sister is a trustworthy source, then the use of the reportative in (3.44a) is unexpected.   This problem cannot be resolved by assuming the speaker’s sister is not a trustworthy source. While this would explain (3.44a), it would render (3.44b) problematic. In Faller’s analysis, only information from a reliable source is capable of assimilation (2002: 139). Therefore, if the source is not trustworthy, then the reported information should neither be capable of assimilation nor constitute best possible grounds. This is a problem for the data in  112 (3.44); if the speaker’s sister is trustworthy, then (3.44a) is unexpected, whereas if she is not trustworthy, (3.44b) should not be possible.    Alternatively, if Faller’s second requirement for best possible grounds is correct, that even information from a trustworthy source must be assimilated in order to be licensed by  -mi, then there is no independent motivation for categorizing trustworthy reports as best possible grounds. If reported information must be assimilated, then arguably it is the act of assimilating the reported information that allows it to be licensed by -mi and not that trustworthy reports constitute best possible grounds, per se. In that case, there is no independent motivation to say that the -mi in (3.45) is licensed by the trustworthiness of the report. Rather, -mi may just as well be licensed because the speaker has assimilated the information into their belief system. This is problematic for Faller since assimilation is not itself a source of information (2002: 140).   The claim that assimilated information is reliable (as opposed to Faller’s claim that reliable information can be assimilated) avoids the empirical problem in the Quechua data in (3.44) and (3.45) above. Speakers presumably do not incorporate unreliable information into their network of beliefs; therefore, if information has been assimilated, we expect that speakers will judge it to be sufficiently reliable to license -mi. This predicts, however, that even initially unreliable information can be licensed by actually where the speaker assimilates that information into their belief system. That is, even ‘not best’ evidence can become ‘best’ evidence over time, in appropriate circumstances. This prediction is borne out empirically in the corpus. Consider the following example, first introduced in chapter 1.  (3.46) JW: They dropped you off?  113 AH: Yea after after they dropped me off um because she had actually stolen four hundred dollars from me the week before I left.  Um (over talking) I didn’t know she stole it, I knew it was missing … ((359) AH1.21:1-4)   p: Casey Anthony stole four hundred dollars from AH.  EV: speaker reasoned that p  In (3.46), AH uses actually to say Casey Anthony had stolen four hundred dollars from her. She explains later in the interview how she came to know that Casey Anthony had stolen money from her.  (3.47) AH: We only had a limited amount of time to talk um and so she’s telling me  all this like Casey had been stealing money from her for years so she stole money from her eighty year old grandmother by using a routing number on a birthday check and like all this stuff and I was like … so that’s where my four hundred dollars went… (AH1.22:18-22)  AH’s information for her claim that Casey Anthony stole her money has two sources. First, she knew she was missing four hundred dollars. Second, Cindy Anthony told her that Casey Anthony had been stealing money from her (Cindy) and Casey’s grandmother. Based on these two pieces of information, AH reasoned that her four hundred dollars was missing because Casey stole it. That is, AH’s evidence is ‘not best’ (reasoning), yet she still uses actually in the circumstances.   AH uses actually to reflect that her information constitutes ‘best’ evidence at the time of utterance. Whatever she may have suspected at the time her four hundred dollars initially went missing, it was not until after she learned that Casey Anthony had been stealing money from the people close to her that she began to suspect the theft. An additional period of some weeks passed between the conversation between AH and Cindy Anthony, on the basis of which she reasoned her missing money had really been stolen, and the time of utterance (i.e.,  114 the time at which AH reports her information to the interviewer with actually). By the time of utterance, she has assimilated p into her belief system, despite the fact her initial source of information for p was reasoning on the basis of (1) the fact of the missing money and (2) Cindy Anthony’s report about Casey Anthony’s behaviour in general. Her use of actually in the circumstances may have been intended less to reflect the strength of her evidence but rather, motivated by a desire to convince the interviewers of Casey Anthony’s character.47 In any event, regardless of AH’s discourse goal, data like (3.46) and (3.47) show that speakers are able to use actually p where their initial source of information for p is not reliable but where enough time has passed for that information to be assimilated into their belief system.48 This is so regardless of their intuition about the reliability of the evidence at the time it was initially received. The idea that speaker intuition can change over time is particularly interesting in the context of police interviews. All of the information received by the investigators in the course of an interview is reported information. The purpose of the interview is for the witness to report their evidence to the interviewer.49 Part of the interviewer’s role is to make assessments of witness credibility and to accept for the purposes of the investigation only credible and reliable information. An interviewer’s initial determination about a witness’s                                                  47 The use of actually as a tool of persuasion will be discussed in section 4.1.3.3.  48 This is not to say that speakers will always assimilate information into their belief system, even when that information otherwise meets the requirements for assimilation. Of course, speakers may choose not to assimilate information into their network of beliefs. As Faller (2002: 138, fn 8) observes, a speaker may also be unprepared for the information they receive and forget it before they are able to assimilate it. Where the speaker has not assimilated the information, they will not use actually to introduce it.  49 Some of the reported information given in an interview may be immediately verified, if the witness brings appropriate physical evidence (e.g., documents, cell phones, etc.) with them. However, the interviews are primarily a source of reported information.   115 credibility may be inconclusive, but may change depending on whether the witness’s information is subsequently confirmed or contradicted by independent physical evidence or because it is corroborated (or not contradicted) by other witnesses. Where an interviewer ultimately decides a witness’s information is credible, they adopt that information as reliable for the purposes of the investigation.  I predict that the act of adopting information for the purposes of an investigation may be mirrored in the interviewer’s language; that is, interviewers can use actually to introduce information that has been adopted as credible even where they have no direct evidence for the proposition themselves. Indeed, interviewers regularly use actually p where they have learned p in the course of a witness interview and have adopted p for the purposes of the investigation. For example, in (3.48), repeated from chapter 1, the witness AH explains to the interviewer that she received a text message from Casey Anthony about gas cans on June 30th.   (3.48) AH: 30th, phone rings at about nine forty-five (0945) in the morning.  Wakes  me up.  Says “this is Casey.” And she’s like, “Come open the door.” And I said, “Are you outside of it?” And she said, “Yes.” So I opened the door and she was pretty much with me until I went to work that night.  Uhm, again, asked me about the, having gas cans.  She actually texted me and asking me if I had gas cans and all my stuff is in storage right now.  Uhm, so we were talking about that and I was like, “Well, why don’t we got to Target?”  EE: She texted you about gas cans?  AH: Yes. ((375) AH2.18:18-19:1)   p: Casey Anthony texted Amy on June 30th.  EV: sensory (visual) – speaker saw the text message It becomes clear later in the interview that the witness provided copies of her text message history to the police for the purposes of the investigation. Armed with the reported  116 information from AH, as presumably corroborated by the text message records, the interviewer puts p to another witness with actually.   (3.49) EE: Okay.  Because the car gets towed on the 30th. LA: Correct.  I’ve been told, and I’ve heard that, that AmScot waited three days before they towed it. EE: Okay.  Yeah (affirmative), and I told you Amy actually got a text message from her on that day. ((179) LA62:14-18)   p: Casey Anthony texted Amy on June 30th.  EV: speaker was told p (by Amy) The interviewer’s use of actually p indicates the interviewer has adopted the information for the purposes of the investigation and assimilated it into his belief system. A speaker’s intuition about the strength of their evidence is only relevant for the purpose of licensing actually at the time of utterance. That is, actually p is licensed if, in the speaker’s judgment, they have ‘best’ evidence for p at the time of utterance regardless of their initial judgment about the information. Conversely, actually p is not attested where the speaker believes their information to be ‘not best’ at the time of utterance, even if they believed it to be ‘best’ when they first received it. Rather, where the speaker initially believed their information to be reliable and subsequently learned it was not, actually p is only licensed where it is qualified, often by adding think.  For example, in (3.50), the witness JJ is explaining that he thought he and others believed they found Caylee Anthony’s remains at a different location than where they were ultimately found. As a result, he wrote in an email that the remains must have been moved to the location after his search had taken place. In (3.50), however, he explains that he wrote the email because, at the time, he thought they had actually found something at the Oviedo location.   117  (3.50) SA: Alright.  Alright are you still… I mean you seemed fairly convinced at the  time you wrote this email that this body was moved ah, over there at… sometime later.  Is that… JJ: I… I don’t… I mean I… I don't know whether the body was moved or not.  Based on… I was basing it on I thought that we actually found something …  ((668) JJ2.55:19-24)  However, he makes it clear in (3.51) that, by the time of utterance, he no longer believes that he and the searchers found Caylee Anthony’s remains at the Oveido location.  (3.51) SA: Okay. Do you still believe today that the body was moved from this Oviedo  location to over there off Suburban Drive? JJ: Based on the fact that I don’t think that, no I don't think so.  …  (JJ2.56:25-57:3)  In (3.50), JJ makes it clear that, at the time of the search, he believed p (that he and the other searchers found something which may have been Caylee Anthony’s remains). Thus, he can use actually to describe his evidence at the time of the search. However, because he no longer believes p at the time he is giving his information in the context of the interview (as shown in (3.51)), actually must be qualified by thought to make clear that he no longer believes p to be true. Indeed, actually p is unattested in the corpus without such qualification where it is clear the speaker no longer believes p to be true.  Actually p is similarly unattested in the police interview transcripts where it is clear the speaker believed their source of information to be ‘not best’ at both the time the information was received and at the time of utterance. In those cases, the speaker uses apparently p or supposedly p. Speakers use both apparently p and supposedly p where their source of information for p is indirect. While apparently marks indirect evidence generally, including reported information, supposedly marks only reported information and that the  118 reported information is ‘not best’. I argue below that, unlike actually and must, apparently p contributes an indirect value of the evidence type dimension of meaning and supposedly p contributes meaning on both the evidence type and evidence strength dimensions.  3.3 The semantics of apparently  I argue that apparently contributes an indirect value on the evidence type dimension of meaning. That is, I argue that apparently marks that the type of evidence a speaker has for the information they introduce is indirect. Recall Willett’s (1988) taxonomy repeated from above.  (3.52) Willett’s (1988) classification of evidence   Direct           Indirect        |         /          \   Attested  Reporting  Inferring        |        |        |   Visual   Second-hand  Results   Auditory  Third-hand  Reasoning   Other sensory  Folklore    The police interview transcripts show that speakers can use apparently to introduce both reported information and information known by inference, both of which are classified as ‘indirect evidence types’ in the taxonomy. The fact that English has evidentials that make contributions on different dimensions of meaning is not problematic; Matthewson (2015b) specifically shows that such systems are attested cross-linguistically.   3.3.1 Apparently marks indirect evidence   Apparently is regularly used to introduce reported information in the police interview transcripts. Where a speaker is sharing information they were told by someone else, they can mark that information with apparently. For example, in (3.53), DB explains how Casey  119 Anthony told Cindy Anthony she was pregnant. In the course of her explanation, DB makes it clear she is only repeating a story that was told to her by others.  (3.53) DB: Okay.  I was not there when it happened, but I was told she came in um,  passed right by everyone to go sit at her mother’s cubicle.  She had a big coat pulled over her stomach and she went and sat and waited for her mom to get out of a meeting and didn’t speak with anyone.  When her mom came out of the meeting I was told she um, went out to her, the car with her mom.  I guess it was time for her mom to go home.  And apparently from what I heard that was when her daughter told her that she was pregnant um, at that time.  And then about two days later Cindy announced she was gonna be a grandmother.  So she verified her daughter was pregnant. (DB2:19-3:2)   p: Casey Anthony told Cindy Anthony she was pregnant in the car after work.  EV: speaker was told p  In (3.54), JJ was explaining some comments he had made about a particular woman who had confronted the director of Texas EquuSearch and one of the lead investigators on camera and who subsequently became friends with the Anthonys. However, he makes clear in his explanation that he is only repeating what he was told by someone else.  (3.54) CE: Or, or you mentioned to me a lady who had been baker acted. 50 What  was… I, I don’t have her name. JJ: I… I don’t know what her name… My understanding was when, when they first found the body there was a lady that confronted Tim on camera and confronted John Allen on camera. CE: Okay. JJ: Um, I got um, I … I … I don’t know what her name was, but she apparently was a friend of the Anthony’s after that and she even admitted she was baker acted five times.  But I don’t know much about her, I heard that third party.  So I don’t know who she is but um… CE: Who’d you hear that from? JJ: You know… Ah, who did I hear that from. I want to say it was … um… I want to say it was Jennifer, Jennifer Cregg. ((A41) JJ1.19:16-20:5)                                                  50 The Baker Act provides for non-voluntary institutionalization for mental illness. To be “baker acted” is to have been institutionalized non-voluntarily under the auspices of the Baker Act.   120   p: The lady (who had previously been “baker acted”) and that confronted Tim    and John Allen on camera became friends with the Anthonys  EV: speaker was told p (by Jennifer Cregg) In (3.53) and (3.54), apparently is functioning as a reportative; the speaker uses it to introduce evidence that they know indirectly through the report of another.   In addition to being used as a reportative, apparently can also be used where the speaker’s information source is inference. In (3.55), for example, LA explains that, during a video visitation with Casey Anthony during her incarceration, she indicated that the family was not to trust her ex-boyfriend, Jesse, but refused to explain why. When asked why she would warn them off Jesse, LA explains as follows.  (3.55) LA: Uh, I do think that’s somebody that she could confide in.  She has said and  conveyed to us that we’re not to trust him.  EE: Why is that?  LA: Well…  EE: Not to trust Jesse? LA: She doesn’t, apparently she doesn’t feel comfortable enough through video    uhm, visitation to convey that to us. ((A4) LA32:10-16)   p: Casey doesn’t feel comfortable enough through video visitation to explain    why they should not trust Jesse.  EV: because Casey didn’t explain why they should not trust Jesse (inference) LA’s evidence for the information that Casey did not feel comfortable explaining why Jesse should not be trusted was the fact that she refused to offer such an explanation. In other words, the fact she refused to explain was evidence that she did not feel comfortable doing so. His evidence for the proposition, therefore, is his own inference about her actions.   Similarly, in (3.56), the witness DB is explaining about the circumstances in which Casey Anthony advised her parents she was pregnant with Caylee. DB recalled that the  121 pregnancy was announced around the time of Casey’s uncle Rick’s wedding and DB’s assumption was that it occurred shortly afterwards because the wedding went smoothly.  (3.56)  YM: I mentioned a wedding that she sent [sic] to Rick’s wedding, which is  Cindy’s brother.  You said you believe…  DB: I know her daughter went.  YM: … it [Casey announcing she was pregnant] was after that? DB: Yeah I just don't remember um, I don’t remember if that happened before or after the wed… I want to say that it happened after the wedding.  ‘Cause the wedding apparently went smooth.  Her and her daughter and the husband went and everything was good.  They had a good time.  So, I don’t remember any um, turmoil or any problems at that wedding. ((A6) DB3:7-15)   p: Rick’s wedding went smoothly.  EV: Cindy said they had a good time and never said there was any turmoil    (inference) DB’s evidence that the wedding went smoothly was based on her interpretation of Cindy Anthony’s report of the wedding afterwards. In particular, DB could not recall Cindy saying there had been any turmoil or problem. DB inferred from those reports that the wedding had gone smoothly.   In (3.57), the source of SC’s information that someone from Texas EquuSearch flagged a Gap bag was that someone had gone back to search more closely in that area and determined that the Gap bag held men’s clothing.  (3.57) YM: Okay.  Do you remember ever seeing a plastic bag, a duffle bag, that, any  type of container, a laundry bag at all out here, while you guys… ? SC: I remember a GAP bag that we found out here in the vines.  There was a GAP bag and apparently we flagged it, and they said that there was umm, they came out and checked and said there was men’s clothes in it.  ((A21) SC6:7-12)   p: a volunteer flagged a GAP bag during the search.  EV: someone went back and searched the GAP bag and found men’s clothing    (inference)  122 SC inferred that, had one of the volunteers not flagged the GAP bag earlier, no one would have known to go search it in more detail later. This was clearly done and its contents recorded; therefore, she inferred that the bag had been flagged earlier.   The police interview transcripts show that apparently is used to mark information that is known by report and by inference; of the 47 tokens of apparently that occur in declarative statements for which the speaker’s information source can be determined, 37 (79%) of those tokens occur with reported information and ten (21%) occur where the information is reasoning or inference. The use of apparently to mark direct, sensory data is unattested in the data. Thus, the types of evidence that license apparently coincide neatly with the ‘indirect evidence’ given in Willett’s (1986) taxonomy. On this basis, I argue that apparently marks that the speaker’s evidence for their proposition is of an indirect type.  (3.58) [[Apparently ]]c,w is felicitous iff the speaker in c has indirect evidence for [[]]c,w in  c.   Unlike with actually, I am only analysing apparently’s evidential contribution and not making any claim about what is asserted with apparently.   Both apparently and must mark that the speaker’s evidence is indirect, but on different dimensions of meaning; specifically, apparently marks indirect evidence type and must marks indirect evidence strength. In the next section, I show how this distinction plays out empirically.   3.3.2 Apparently versus must; evidence type versus evidence strength   Apparently marks that the speaker has indirect evidence for the information they share. Recall that von Fintel & Gillies (2010) show that must also contributes an indirect  123 evidential meaning. However, I argue that, while both apparently and must contribute an indirect value, they operate on different dimensions of meaning. Specifically, while I follow Matthewson (2015a, 2015b) in arguing that must marks indirect strength, I argue that apparently marks indirect type. Because both apparently and must mark indirectness, although on different dimensions, I predict there will be significant overlap with respect to the type of evidence that licenses each of them. I show that these predictions are borne out in the police interview transcript data.    I show above that apparently is felicitous where the speaker’s evidence for the information they proffer is inference or reasoning. I argue that this is because inference is properly classified as indirect evidence in Willett’s taxonomy of evidence types. Von Fintel & Gillies (2010) show that must is also felicitous where the speaker knows p by way of inference. Matthewson (2015a, 2015b) argues that inference is encompassed by an indirect value on evidence strength, just as it is in an evidence type framework. Therefore, where the speaker’s evidence is inference, I predict that both apparently and must will be felicitous.  This prediction does, indeed, appear to be borne out empirically. While the data from the police interview transcripts in (3.55) through (3.57) above all involve apparently, (3.59) shows that must is equally felicitous in those contexts.  (3.59) a. … apparently she doesn’t feel comfortable enough …   … she must not feel comfortable enough …    b. … ‘Cause the wedding apparently went smooth …     … ‘Cause the wedding must have gone smooth ….     c. … There was a GAP bag and apparently we flagged it…   … There was a GAP bag and we must have flagged it…  124 The felicity of both apparently and must results from the fact that inference counts as indirect evidence on both the evidence type and the evidence strength dimensions of meaning.   Unlike inference, reported evidence does not necessarily count as both indirect evidence type and weak evidence strength. Under Willett’s taxonomy, reported evidence is always indirect on the evidence type dimension. However, as discussed in chapter 1 and section 3.1, both Faller (2002) and Matthewson (2015a, 2015b) show that, depending on the context, reported evidence may be direct on the evidence strength dimension. I repeat the relevant examples from both English (repeated from 1.10) and Cuzco Quechua (repeated from 1.13c) below. The reported information must count as direct on the strength dimension because it is felicitous with -mi (3.60b) but renders must infelicitous (3.60a).  (3.60) a. ENGLISH   Context: Belinda, Bob’s wife, told the speaker that Bob is home. Belinda is  a very reliable source. The speaker now tells someone else:    # Bob must be home.  (Matthewson, 2015a: (17)) b. CUZCO QUECHUA  Lima-ta-n viaja-n.     Lima-ACC-mi travel-3    p =  ‘She travelled to Lima.’     EV =  speaker was told by her (= speaker’s sister) that p (Faller, 2002: (10(c)) I predict that apparently will always be felicitous where the source of the speaker’s information is a report. Von Fintel & Gillies show, however, that must is not felicitous as a reportative. I argue below, however, that must is felicitous with certain reported information.  125 Specifically, I argue that must p is felicitous where the reported information does not directly settle p.  Von Fintel & Gillies (2010) observed that must is infelicitous as a reportative generally, giving the following as an example.  (3.61) Context: After reading the newspaper report about last night’s Knicks game  #The Knicks must have lost again.    The Knicks apparently lost again.  (adapted from von Fintel & Gillies (2010), fn. 3) Where the speaker’s sole source of information is a report, must is infelicitous but apparently is perfectly fine.   The example from von Fintel & Gillies in (3.61) involves a report where the information necessary to directly settle the prejacent, the result of the Knicks game, is expressly stated. Because the report in a newspaper is generally considered to be trustworthy, the report constitutes direct evidence which disallows must. However, where the newspaper does not expressly report the information necessary to directly settle the prejacent, both must and apparently are felicitous.  Consider a slightly different example, in (3.62).  (3.62) Context: You know the Knicks played last night. You watched the first half, after  which the Knicks were winning. You didn’t watch the end of the game and don’t know who won.  The newspaper reports “Robin Lopez injured in Knicks’ loss.” You say     The Knicks must have blown the lead.   The Knicks apparently blew the lead. The newspaper report does not expressly state that the Knicks had blown their halftime lead, yet the speaker (correctly) infers that this must have happened in order for them to have lost the game. Must is felicitous in this context because the speaker knows the proposition (the  126 Knicks blew their halftime lead) to be true on the basis of indirect evidence; the reported information plus the inference it prompted. On the other hand, apparently is felicitous with reported information, regardless of whether the reported information contains p itself because apparently marks indirect evidence type. A report is a report regardless of whether or not it contains the information necessary to directly settle the prejacent and is, therefore, still indirect evidence from the perspective of evidence type. Moreover, the fact the speaker also uses reasoning as information for the proposition is inconsequential for apparently because reasoning is also an indirect type of evidence.  The data show that, where the speaker’s evidence is of an indirect type and where that evidence does not constitute the best evidence in the circumstances, both must and apparently will be felicitous. This was the case in (3.62) above. However, the circumstances which license apparently will not always license must. Where the evidence in question is of an indirect evidence type, but constitutes best possible grounds in the circumstances, apparently will be felicitous but must will not. This was the case in (3.61). The newspaper report of the Knicks loss was the best possible grounds of evidence for a speaker who did not watch the game itself. Therefore, must is infelicitous. However, because on the evidence type dimension of meaning, reported evidence is ‘indirect’, apparently is perfectly felicitous in the circumstances.51                                                  51 Note that, if the speaker had watched the Knicks game the previous night, then must would be infelicitous and apparently would be awkward.  (i) Context: The speaker watched the Knicks game and saw that they lost. After  reading the next day’s newspaper report about the Knicks game.     # The Knicks must have lost.   ? The Knicks apparently lost.   127  Apparently differs from must in another crucial respect. As discussed earlier, von Fintel & Gillies (2010) convincingly show that must entails its prejacent. The same is not true for apparently. While must commits the speaker to the truth of the prejacent, apparently is felicitous regardless of whether the prejacent is true or false. Because it does not entail the prejacent and because it is not marked for evidence strength, speakers can use apparently to indicate that they are agnostic about the reliability of the prejacent.52 This is consistent with the data in section 3.1.2, where I showed that speakers in the police interview transcripts used apparently where they were agnostic about the reliability of their evidence and supposedly where they expressly distrusted the information they were sharing.   Because apparently marks evidence type, not evidence strength, I predict that it should co-occur with both must and actually, both of which mark evidence strength, not type. This prediction is born out in the police interview data for apparently and actually.  (3.63) JW: Okay, how do you know uh… Casey Anthony? AH: Uh Casey and I apparently actually met a little over a year ago at a party that I had my house, I don’t recall meeting her it was a big party…   ((347) AH1.2:12-14)  While there are no attested examples of must co-occurring with apparently in the police interview data, my native speaker intuition is that they can co-occur, as in the example below.                                                                                                                                                          Must is infelicitous because the speaker has reliable evidence for the Knicks loss; they watched the game itself. Apparently is awkward because it violates Grice’s cooperative principle of conversation; specifically, by (falsely) implicating that the speaker has indirect evidence for p.  The details of this require further working out, which I leave to future research.   52 See chapter 5 for further discussion.   128 (3.64) Q: Okay, how do you know uh… Casey Anthony? A: Uh Casey and I apparently must have met a little over a year ago at a party that I had my house, I don’t recall meeting her it was a big party…  I discuss multiple evidential markers in greater detail in section 5.3. Suffice it to observe, at this point, that apparently can, indeed, co-occur with actually and must.  3.4 The semantics of supposedly Like apparently, speakers use supposedly where the source of their information is reported. Supposedly differs from apparently, however, insofar as it only occurs with reported information. While apparently’s ability to function as a reportative marker falls out from its function as a marker of indirect evidence generally, supposedly is a true reportative marker and is unattested with any other type of evidence source. Moreover, I argue that supposedly also makes a contribution on the evidence strength dimension of meaning, encoding that the speaker’s evidence is ‘not best’ in the circumstances.  The police interview transcripts clearly show that speakers use supposedly to refer to reported information. In (3.65), GA explicitly notes that his information about who Casey called after running out of gas is based on what he was told by others.   (3.65) GA: [Casey ran out of gas at Goldenrod and] 50, at the Amscot.  The first  person she called supposedly, I don’t know, I don’t, I, I’m just going by what I’ve been told by a couple friends that she mentioned (unintelligible) that Jesse said. ((S3) GA1.18:15-17)    p: The first person Casey called after running out of gas was …  EV: GA was told p (by a couple of friends)    129 Similarly, in (3.66), the witness MM makes explicit that her information about Cindy taking Casey away from a restaurant after finding out about her second pregnancy was reported to her by Lee.   (3.66)  YM: Let me go, let me go backwards real quick uh as far as, when Cindy found  out [about the second pregnancy] did she like go grab her from out of a restaurant or something? MM: Supposedly I heard she took her out of a TGI Fridays. YM: But you weren’t there? MM: No I was not there… YM: You just heard about it. MM: … that, that time, no. YM: Who’d you hear it from, Lee? MM: Uh Lee. ((54) MM12:9-17)   p: Cindy pulled Casey out of a TGI Fridays.  EV: MM was told p (by Lee)  Of the 74 tokens of supposedly in declarative statements in the police interview data, all of them refer to reported information; there are no attested instances of supposedly in the police interview transcripts where the source of information is something other than a report.   In addition to marking reported information, supposedly marks that the speaker does not believe the prejacent, either because they know it to be false or because they believe the source of the report to be untrustworthy. In (3.67), the witness JG is explaining that he began to put a calendar together to track days where Casey Anthony said that Caylee was either with the nanny or with Cindy Anthony. All of these days were in the month between Caylee’s disappearance and the date on which the disappearance was reported; in other words, JG knows that, despite Casey’s report about Caylee’s whereabouts, Caylee could not have been with Cindy Anthony at the material time.   (3.67) JG: You know, things of that nature.  Because otherwise you know, because it   130 was my first thing.  That’s why I put that calendar book together because I figured okay, if the days that we can find out that she was supposedly with her nanny, or supposedly with mom, and she wasn’t with mom, and Casey was alone, where was she? She had to have been with someone…  ((S40) JG2-9:18-22)   p: Caylee was with her nanny, Caylee was with Cindy.  EV: JG was told p (by Casey)  Similarly, in (3.68), the witness RP reports what Casey Anthony told him about her job with supposedly. Later in the discourse, he explains that everything she told him about her job had been a lie.  (3.68) YM: Can you tell me a little bit about where she used to work uhm, while, I  guess we’ll start with last year because I don’t want to go too, too far back.  RP: Uhm, well I mean uh, that whole time supposedly she was working at Universal.  And that’s the only thing she ever told me.  I think the biggest problem with her is that, well with me especially, is that she had to, she this issue with uh, you know, trying to, you know, be one-uppy? And I guess she kind of felt her life wasn’t necessarily going in the direction that she wanted it to go.  So every time she would talk to me and maybe that something was going good for me like I finished up my final exams, or I’m getting this much closer to graduating, she would always come up with something to kind of make the, the playing field level between the two of us somehow.   YM: Uh-hum (affirmative)  RP: I guess she felt like it was a competition.  YM: Right. RP: So when it came to the job situation, you know, I again, I guess it was a  lie… ((S57) RP12:8-13:2)  p: Casey was working at Universal.  EV: RP was told p (by Casey)  In both (3.67) and (3.68), the witness uses supposedly to indicate reported information which they know at the time of utterance is untrue.  In (3.69) and (3.70), however, the witness does not indicate whether the prejacent is true or false; their skepticism arises because they do not believe the source of their  131 information to be trustworthy. In (3.69), the witness GA expresses his doubt about Casey Anthony’s claim that she had been held up at gunpoint at work. As a former law enforcement officer, GA explained that Casey’s story did not correspond to what he would have expected, if she really had been held up at gun point.   (3.69) GA: And, and I (unintelligible) and this, well and I even asked this Ryan, God  what’s Ryan’s last name? Hastings? Yeah (affirmative).  I asked Ryan, I said “Ryan, did you work with Casey?” “No, she doesn’t work here.” I said, “Well then how in the hell did she make.” Oh, did we tell you about the report supposedly that she was at gunpoint…  YM: Well…  GA: … of a four thousand dollar uh…  YM: Yeah (affirmative), actually uh…  GA: … (inaudible)  YM: … Lee told me about that when we were sitting right here.  GA: Right.  YM: Just us four (inaudible) GA: Because I asked her about that night.  I mean she puts on a pretty good act I guess.  I don’t know if she’s playing actress or what, but I said “Where’s, where’s the officer’s card and I want to see it.  I want to know what’s going on with this.  Just if we have to place you someplace else because this guy’s got your information I want to get you the hell out of this.”  That’s exactly what I said.  YM: Right. GA: “Dad, I’ll be fine.  I just need a moment.”  And my wife says, “Oh, we’re going to work with this through her and all.”  And I said, “This doesn’t, this shit doesn’t” excuse me “this stuff does not add up.”  ((125) GA1.30:13-31:9)    p: Casey was held up at gunpoint.  EV: GA was told p (by Casey) Similarly, recall the following example from section 3.1.2.  (3.70) PR: Do you know of her uhm, a boyfriend of hers or anything like that? DC: Just in rumors.  Yeah (affirmative), I was talking to Jonathan last night and apparently from what he was making it Casey had pulled herself out of their crowd for about the last month, month and a half.  I’m trying to remember the guy’s name.   PR: Pulled away from Jonathan?  132  DC: That whole crowd.  PR: Just the whole crowd? DC: Yeah (affirmative).  She’s been hanging out with a different crowd.  And uh, she had said supposedly that she was dating a guy named I believe it was Rodrigo. (DaCo9:8-17)   p: Casey was dating a guy possibly named Rodrigo  EV: speaker was told by Casey that p   Recall further that the witness DC explained his use of supposedly in (3.71) as follows.  (3.71) DC: Yeah (affirmative).  Because I asked Jonathan, “Have you ever met the  guy? Is there any way we can get ahold of him?” And Jonathan’s like, “No.  Well she never brought the guy around.”  You know but then when, you know we, we were playing detective last night and uh..  PR: Oh yeah (affirmative) DC: … it was like, at that point as much lying as she’s doing, who did uh, who’s to know if any of it’s true.  (DaCo11:1-7)  DC explains that he does not necessarily disbelieve that Casey Anthony was dating a guy, possibly named Rodrigo, but rather that his source of that information (namely, Casey Anthony) was not reliable or trustworthy.  Like apparently, supposedly marks that the speaker’s evidence source for the prejacent is of an indirect evidence type; for apparently, the evidence is any sort of indirect type whereas supposedly functions as a true reportative. Supposedly has an additional lexical component in that it marks that the speaker does not believe the prejacent, either because they know it is false or because they distrust the source of their information.   (3.72) [[Supposedly ]]c,w is felicitous iff the speaker in c  i. has reported evidence for [[]]c,w in c, and  ii. does not have best evidence for [[]]c,w in c   3.5 Summary of the semantics of English evidentials, so far  133  I argue that actually, must, and supposedly make evidential contributions on the evidence strength dimension of meaning and apparently and supposedly make contributions on the evidence type dimension. Specifically, I argue that actually encodes ‘best’ evidence (i.e., a direct value for evidence strength) and must and supposedly encode ‘not best’ evidence (i.e, the corresponding indirect value). Both apparently and supposedly mark indirect evidence type; apparently encodes indirect evidence generally and supposedly encodes reported evidence. These evidentials can be organized as in the following table.   DIRECT INDIRECT EVIDENCE TYPE   apparently supposedly  EVIDENCE STRENGTH actually must supposedly TABLE 3.2: English evidential specifications  This evidential breakdown is somewhat unusual in that there are specific evidentials to mark indirect evidence type, but no corresponding markers for direct evidence type. While actually can be used where the speaker’s evidence is of a direct type, that results because of the overlap between direct evidence type and direct evidence strength; as discussed above, direct evidence type will generally be considered ‘best.’ However, I have not identified a dedicated marker for direct evidence type that is distinct from actually.   Following Ward et al. (2003), epistemic would is a promising candidate as a marker of direct evidence on the type dimension of meaning in English.53 They show that epistemic would is felicious where the speaker has “conclusive objective evidence for the truth of the proposition conveyed” (Ward et al., 2003: 78), giving the following as examples.                                                   53 Thanks to Jeffrey Kaplan (p.c.) for drawing epistemic would to my attention as a marker of evidentiality.    134  (3.73) a. … I do have some answers for you. You asked about one person declaring all  of the income on one property and one person taking all of the expenses. The answer to that would be no.  b. B:  Are you the Meredith that was listed in the Graduate Student News?  M:  Yeah, that would be me.  c. (A is holding a plastic bag with a fish in it)   A:  I don’t know, maybe it’s in shock.  B:  Looks dead to me.  C: That would be one dead fish.  (Ward et al. (2003): (4))  Although it remains to be tested empirically, epistemic would warrants consideration as a marker of direct evidence type in English that might fill the typological gap noted in Table 3.2. Similarly, while supposedly marks reported information, I have not identified a reportative marker that does not also mark ‘not best’ evidence. Such reportative markers are clearly attested cross-linguistically.54 Therefore, I make no claim that the evidentials I discuss in this dissertation form a complete system; rather, I leave to future research whether other English discourse markers also contribute evidential meaning and on what dimension(s).55    3.6 Legal rules of evidence                                                  54 The German reportative sollen is a possible instance of this. Thanks to Lisa Matthewson (p.c.) for pointing out this example.   55 An analysis of evidentials as making contributions on one or more of the three possible dimensions of meaning predicts that there will be some evidential markers that contribute a direct value on one dimension of meaning and an indirect value on another. While this may sound intuitively odd, such evidentials appear to be attested cross-linguistically. For example, the St’át’imcets evidential lákw7a appears to mark direct evidence type but indirect evidence location. See Matthewson (2015b) for discussion.    135 By encoding ‘best’ evidence, English and Cuzco Quechua evidential systems reflect the same standards of reliability that are used to determine the admissibility of evidence in the Canadian and other legal systems. Evidence will only be admissible at court if it counts as the best available evidence in the circumstances. As I alluded to in chapter 1, the types of evidence that are admissible and the ways in which that evidence may be properly admitted in a legal proceeding are highly constrained. The rules respecting the admissibility of evidence have developed in such a way as to largely permit the types of evidence that license actually and -mi and to exclude the types of evidence that do not. Actually and -mi encode a similar concept of evidential reliability; the information that may constitute ‘best’ evidence for the purpose of licensing actually and -mi also tend to be admissible in a legal proceeding, either generally or as a specific exception to the Rule Against Hearsay.  3.6.1 Knowledge, observation and experience Recall from chapter 1 that, in the context of deciding a legal dispute, the trier of fact is only entitled to rely on information properly put into evidence before them during the course of the trial. In that context, the role of the witness is to testify to the information necessary to prove or disprove the issue in dispute. In doing so, the witness is limited to testifying only to “facts within their knowledge, observation and experience” and not opinions based on those facts (Paciocco & Stuesser, 2015: 195).56 A witness is generally only entitled to give p as evidence where their source of information for p is either speaker                                                  56 Paciocco & Stuesser rightly observe that ‘fact’ and ‘opinion’ are often difficult to distinguish in practice. For example, a witness who testifies that a wound is “life-threatening” is offering impermissible opinion evidence whereas a witness’s testimony that the assailant “struck quickly” is likely to be admissible, even though that, too, is an opinion (2015: 195), although one based on sensory and, therefore, “direct” information.     136 performance (the witness performed p) or secured through direct sensory observation. Witnesses are not entitled to give p as evidence where their information for p is reasoning or inference. With respect to information known through report, general knowledge, or encyclopedic knowledge, this type of evidence is presumptively inadmissible on the basis that it violates the Rule Against Hearsay. However, this is precisely the sort of the evidence a court may choose to admit, provided that it meets the procedural safeguards set out in the principled exception to the Rule Against Hearsay as described in R. v. Khelawon.57  3.6.2 The Rule Against Hearsay  Hearsay evidence is generally inadmissible unless it falls within a specific exception to the Rule Against Hearsay. In chapter 1, I observed that the Rule Against Hearsay will prevent a witness from giving in evidence an out-of-court statement as proof of the content of that statement. This is because the court generally considers reported evidence to be less reliable than evidence of experience or observation. Consider again the example from chapter 1.  (3.73) Context: Jane sees Sue break a window. Later, Jane tells John that Sue  broke the window.    (a) Jane: “Sue broke the window”   p =  ‘Sue broke the window’  EV =  Jane saw p (sensory visual)   (b) John: “Sue broke the window”    p =  ‘Sue broke the window’  EV =  John was told p (by Jane) (report)                                                  57 2006 SCC 57.   137 The Rule Against Hearsay dictates that the party who intends to prove that Sue broke the window must call Jane to swear to that information in evidence.58 This is so even if John has good reason to believe Jane’s report of p is trustworthy.  Indeed, it is the notion of reliability that gives rise to the Rule Against Hearsay. The basic rationale for the Rule Against Hearsay is that the source of the reported information (the ‘declarant’) should be present and subject to cross-examination so that the trier of fact can assess their reliability and trustworthiness for itself, rather than rely on someone else’s assessment of trustworthiness. As Mr. Justice Fish, writing for the majority of the Supreme Court of Canada, held in R. v. Baldree,59   First, the declarant may have misperceived the facts to which the hearsay statement relates; second, even if correctly perceived, the relevant facts may have been wrongly remembered; third the declarant may have narrated the relevant facts in an unintentionally misleading manner; and finally, the declarant may have knowingly made a false assertion. The opportunity to fully probe these potential sources of error arises only if the declarant is present in court and subject to cross-examination.  (Paciocco & Stuesser, 2015: 114. Emphasis in original)  In other words, the trier of fact is generally unwilling to give credence to John’s belief about Jane’s trustworthiness; rather, it prefers to make its own assessment of her reliability.60                                                   58 Note of course that the Rule Against Hearsay only operates against John’s evidence for the proposition that Sue broke the window. If p is Jane said Sue broke the window, then John’s evidence is no longer hearsay because his evidence for p is direct sensory (auditory). That is, if the purpose of John’s evidence is to prove the fact of Jane’s report (and not the contents of her report), the Rule Against Hearsay does not apply.    59 2013 SCC 35, at paragraph 32.   60 See section 5.3 for further discussion about the difficulty in relying on another speaker’s assessment of reliability.    138  However, let us assume that for some reason Jane is no longer able to give evidence based on her own observation of p. Her inability may arise for any number of reasons. Perhaps she has developed a medical condition which prevents her from either remembering the event or talking about it. Perhaps she has died or moved abroad and is no longer available to give evidence. In such cases, John may be entitled to give p as evidence, even though it is hearsay, subject to the two general guiding principles of necessity and reliability. Like in the speaker’s assessment of ‘best’ evidence for the purpose of licensing -mi and actually, the question of whether hearsay evidence is necessary and reliable will generally be determined on a case-by-case basis.61 ‘Necessity’ is concerned with obtaining “the relevant direct evidence from a particular witness” (Paciocco & Stuesser, 2015: 133); in other words, a court will not admit hearsay evidence unless that evidence is necessary to prove the case. If the hearsay evidence does not go to an essential component of the case, it will not be admissible regardless of whether it is reliable. In our example, the necessity requirement will arguably be satisfied if p is a necessary piece of information to prove the offence.   Having established necessity, John’s evidence for p will be admitted as an exception to the Rule Against Hearsay where it meets the test for threshold reliability, that is, where the hearsay statement exhibits sufficient indicia of reliability so as to afford the trier of fact a satisfactory basis for evaluating its truth (R. v. Hawkins;62 Paciocco & Stuesser, 2015). Such indicia may include any corroborating or conflicting evidence otherwise in the record (Paciocco & Stuesser, 2015). Although John’s reported evidence is not inherently reliable in                                                  61 R. v. Khelawon, 2006 SCC 57; R. v. Khan, [1990] 2 S.C.R. 531; R. v. Smith, [1992] 2 S.C.R. 915.  62 (1996), 111 C.C.C. (3d) 129 (S.C.C.) at 75.   139 the court’s eyes, it may become so if it is corroborated by other evidence in the proceeding. Where the trier of law is satisfied that the hearsay statement, otherwise determined to be necessary, meets the test for threshold reliability, it can be admitted as evidence.63   ‘Best’ evidence markers like Cuzco Quechua -mi and English actually will be licensed where the speaker has the best evidence for p available to them in the circumstances. What constitutes best evidence is context-dependent and must be evaluated on a case-by-case basis. Reports will constitute best evidence where the speaker was not in a reasonable position to observe p directly and only where the source of the report is trustworthy. In other words, reports will constitute best evidence in the same contexts where hearsay evidence is arguably admissible at court: where (1) there is no opportunity to obtain the information through direct observation (or through a witness who directly observed the event) and (2) where the report is determined to be reliable.         3.6.3 General and encyclopedic knowledge   The courts have also recognized an exception to the Rule Against Hearsay which permits it to receive generally known or encyclopedic information. The need for such an exception is explained in R. v. Zundel, where the court held:                                                  63 The ‘threshold reliability’ test is used to determine admissibility. However, the hearsay statement will still be scrutinized against a standard of ‘ultimate reliability’ by the trier of fact, who will decide how much weight (if any) to afford it in determining whether the offence has been proved.    140  … the events of general history may be proved by accepted historical treatises on the basis that they represent community opinion or reputation with respect to an historical event of general interest.  The historical event must be one to which it would be unlikely that living witnesses could be obtained, and in addition, the matter must be one of general interest, so that it can be said that there is a high probability that the matter underwent general scrutiny as the reputation, evidenced by the historical treatises, was formed: see Wigmore On Evidence, 3rd ed., vol.5, p. 462; McCormick on Evidence, 34rd ed., p. 906.64   The general/encyclopedic knowledge exception to the Rule again Hearsay plays out in one of two ways. With respect to general knowledge, the court can take judicial notice of facts that are considered ‘common knowledge’. With respect to encyclopedic knowledge, the court can hear opinion evidence from experts in that particular field.  The court has the option of taking judicial notice of general or encyclopedic knowledge in appropriate circumstances. Where a court takes ‘judicial notice’ of p, it means that the court accepts p as a fact in evidence without either party having to prove it through witness testimony. A court may take judicial notice of “any fact or matter which is so generally known and accepted that it cannot reasonably be questioned, or any fact or matter which can readily be determined or verified by resort to sources whose accuracy cannot reasonably be questioned:” R. v. Potts.65 For example, if the case involved an injury resulting from being hit by a falling object, the court can take judicial notice of the existence of gravity, rather than require the prosecutor to prove it. Having taken judicial notice of a fact, the court is entitled to rely on it in the absence of witness testimony in the decision making process.                                                  64 [1987] O.J. No. 52, at paragraph 130. 65 (1982) 36 O.R. (2d) 195, at page 201.   141  The court has also recognized a second and related exception to the Rule Against Hearsay which pertains specifically to expert knowledge. This exception is described in R. v. Zundel as follows.   The second exception which is relevant in this case is that an expert witness may give evidence based on material of a general nature which is widely used and acknowledged as reliable by experts in that field.  This exception, however, has hitherto been confirmed to a few narrow classes of cases such as, for example, mortality tables and a standard pharmaceutical guide: see Pattenden, “Expert Opinion Evidence Based on Hearsay,” [1982] Crim. L.Rev. 85 at pp. 90-1; Cross on Evidence, 6th ed. (1985) at p. 452.66  Thus, expert witnesses are entitled to give evidence on the basis of what they know by virtue of being an expert in their particular field; they are not limited to giving evidence only on the basis of what they experience or observe. However, before this exception can be invoked, the court must be satisfied that special knowledge or experience is necessary to assist the trier of fact and that the proffered witness has the necessary qualifications and expertise to offer that evidence. Recall that Faller (2002: 135) argues encyclopedic knowledge will license –mi where two conditions are met. First, the speaker must have learned the information from a source of authority and second, the speaker must consider themself (and be considered by others) to be an authority on the subject. In other words, an authority on a particular subject is entitled to use -mi to introduce information they know as a result of that expertise. As shown above, actually is similarly licensed in this context. Arguably, this is precisely the sort of information an expert would be entitled to give as evidence in a legal proceeding.   3.6.4 ‘Best’ evidence and the legal rules of admissibility                                                  66 [1987] O.J. No. 52, at paragraph 141.   142  The legal rules of evidence admissibility have been crafted in a way so as to ensure that only the most direct, authentic evidence is admissible. A witness is only permitted to testify to information they know on the basis of reliable sources. Those reliable sources include: speaker performance, sensory observation, trustworthy hearsay, and general or expert knowledge. That is, the sources of information that may meet the threshold test of reliability at court are the same sources of information which license actually. While speaker performance and observed information will generally be admissible as a matter of course, the court must sometimes admit evidence that is not of the most direct, authentic type as a matter of public policy. As the court held in R. v. Khelawon and R. v. Baldree, the admission of evidence must first and foremost be guided by society’s interest in getting at the truth.67 Moreover, whether the evidence is sufficiently reliable to be admitted at court or to license actually will be determined on a case-by-case basis. The type of evidence that is arguably admissible as an exception to the Rule Against Hearsay largely mirrors the type of evidence that may be licensed by actually and -mi. The fact that a determination of threshold reliability at law largely parallels how speakers use actually suggests that the linguistic encoding of ‘best’ evidence aligns with some more general cognitive understanding of reliability that the courts also invoke in order to justify decisions about the admissibility of evidence.    3.7 Actually, apparently, and supposedly and the alignment of participant roles   In this chapter, I motivated evidential analyses of actually, apparently, and supposedly. I argued that actually contributes a direct value on the evidence strength dimension of meaning. To this end, actually contrasts with Matthewson’s (2015a, 2015b)                                                  67 R. v. Khelawon, 2006 SCC 57, at para. 49; R. v. Baldree, 2013 SCC 35, at para. 109.   143 analysis of must as marking an indirect value for evidence strength. This means that actually will be felicitous where the speaker has best evidence for the prejacent and must will be felicitous where the speaker’s evidence is not best. I further argued that both apparently and supposedly mark evidence type and that both contribute an indirect value. While apparently marks indirect evidence on the evidence type dimension, supposedly is a reportative that also makes a contribution on the evidence strength dimension of meaning, marking ‘not best’. In this section, I consider how, given these semantics, speakers can use actually, apparently, and supposedly to align themselves with certain participant roles in the investigative interview.   Recall from chapter 1 that Heydon (2005) argued that participants in interrogations align with different participant roles depending on the institutional goal of the relevant phase of the interview. The goal of the information-gathering phase of the interrogation was to have the suspect voluntarily confess to the crime. That goal is best achieved when the suspect aligns themself as the author, animator, and principal of the information. Heydon adopted Goffman’s participant roles, which she explained as follows.  Goffman proposes that participants in talk be identified as having four possible roles: principal, animator, figure and author. Goffman (1974) describes the principal role as ‘the party who is held responsible for having willfully taken up the position to which the meaning of the utterance attests (p. 517) and the animator role as producer of an utterance, ‘current, actual sounding box from which the transmission of articulated sound comes” (p. 517). The role of figure can be occupied by anyone or anything that can be spoken of or animated by another, while the role of author is one who creates or ‘writes’ an utterance. Goffman proposes that when people interact, they assume one or more of these roles forming a network of participant roles… (at p. 21)  Even though the institutional goals of witness interviews are different than suspect interviews, the goal remains to have the witness voluntarily proffer the information necessary  144 to advance the investigation. That goal is best achieved when the witness, like the suspect, aligns as the author, animator, and principal of the information proffered.     With a bare assertion, the witness is at the very least aligning themself as the animator of the information; they are the “actual sounding box from which the transmission of articulated sound comes” (Heydon, 2005). However, it is not clear from a bare assertion whether the witness is also aligning as an author and/or a principal of that information. For example, if a witness gives their information through a bare assertion, with nothing to indicate how they know the information, it is unclear whether they are merely the animator of the information (who is merely passing on what they were told by someone else), whether they are the author of the information (the creator of the information) or whether they are the principal of the information (who willfully adopts the information and attests to its reliability). A bare assertion is not sufficient to achieve the institutional goal of aligning the witness to the animator, author and principal roles with respect to the information they proffer. By marking that the speaker has reliable evidence for the prejacent, actually commits the speaker to a more substantial role with respect to the information itself. Specifically, actually commits the speaker to being the principal and/or author of the information.  Looking ahead, I show in chapter 4 that speakers in the police interview transcripts use actually to expressly align with the author and/or principal roles. On the other hand, by using apparently and supposedly, the witness expressly limits their role to that of animator; that is, the speaker can use apparently and supposedly to explicitly reject the principal and author roles. Thus, the witness avoids taking responsibility for the information by expressly aligning as only the animator of the information. I show how speakers use apparently and supposedly to do this, in chapter 5. Finally, in chapter 6, I show how investigators use  145 actually in questions to force the witnesses to align themselves with participant roles when they fail to do so voluntarily.           146 “…I need to know why you’re saying these things” YM (CA2.74:2-3)      4. THE PRAGMATICS OF ACTUALLY  Chapter 3 was devoted to presenting a semantic analysis of actually, apparently, and supposedly in English. Specifically, I argued that actually makes an evidential contribution on the evidence strength dimension of meaning, contributing a direct value (i.e., ‘best’ evidence). To this end, actually is in complementary distribution with must, which Matthewson (2015a, 2015b) argues contributes an indirect evidence strength value (i.e., ‘not best’ evidence). Apparently, on the other hand, contributes evidential meaning on the evidence type dimension, encoding that the speaker’s evidence is indirect (reported, reasoned or inferred). Supposedly makes contributions on both the evidence type and evidence strength dimensions of meaning, marking reported evidence that is ‘not best’ in the circumstances. In this chapter and the next, I focus not on what English speakers mean when they use these evidentials, but rather on what English speakers do with these evidentials in the course of social interaction. I show that, since evidentials are not grammatically obligatory in English, speakers use an evidential-marked assertion as opposed to a bare assertion where circumstances are such that a bare assertion will not suffice. I follow Fox (2001) and Kim (2006), and argue that a speaker’s choice of evidential is determined in large part by the ongoing discourse in which the speaker is a participant. I argue that speakers use evidential-marked assertions to negotiate the common ground and that their choice of evidential is determined by their discourse objective in the circumstances.   147 I argue that speakers use evidentials to negotiate the common ground where a bare assertion will not suffice. I show that, if a bare assertion constitutes a proposal that its propositional content be included in the common ground, then speakers use actually not only to propose information for inclusion in the common ground but also to advocate for its inclusion on the basis of best evidence. Speakers do this with actually by explicitly aligning as the information’s author and/or principal. In chapter 5, I argue that speakers use apparently and supposedly to put information into the discourse without proposing it for inclusion in the common ground by only aligning as the animator of information.  I show that, while evidential marking is not obligatory in English, the speakers in the police interview data use actually for two broad discourse purposes: to advocate for the inclusion of their information in the common ground by emphasizing its reliability and to control the flow of the discourse. In section 4.1, I discuss four contexts which require the witness to explicitly mark their evidence as ‘best’: where the speaker believes that the proffering of their information has not been met with an appropriate response, where the speaker is required to self-correct an error, where the speaker is compelled to challenge the accuracy of information provided by their discourse partner, and where the speaker is required to persuade their discourse partner to accept their evidence.  I discuss the well-documented contrastive reading of actually in section 4.1 and argue that the contrastive function observed in the literature is not an inherent part of actually’s semantics. I show that speakers use actually both to emphasize their position when their initial bare assertion has not received the expected response in the discourse and, in questions, to commit the witness to the information they previously proffered with a bare assertion. In neither of these contexts is actually properly interpreted as a marker of contrast.  148 These felicitous contexts of use would be unaccounted for if contrast were an inherent part of actually’s semantics. Moreover, since the contrastive interpretation of actually falls out from the semantic analysis of actually set out in section 3.2, specific reference to contrast in its lexical entry is unnecessary.  Even though contrast is not part of actually’s lexical semantics, speakers regularly use actually to mark contrast or counter-expectation. I argue that, in the context of the police interview transcripts, the contrastive interpretation falls out from actually’s evidential contribution as set out in section 3.2 and the institutional goal of the witness interview more generally. Specifically, the goal of the witness interview is to establish an accurate evidentiary record on which to advance the investigation. If the speaker, in particular the witness, knows that the information proffered for inclusion in the common ground is inaccurate, the witness’s institutional role requires them to correct or challenge that information. By marking that the speaker has best evidence for their information (which happens to be different from that proffered), actually allows the speaker to challenge the inaccurate information in a way that is consistent with the institutional discourse. In the same way, a speaker can use actually to advocate for the inclusion of their information in the common ground when they believe that information has been or will be dismissed; that is, the speaker can use actually to attempt to persuade their discourse partners that their information should be accepted as part of the common ground. I show that, to these discourse ends, actually functions like evidentials in Korean and Wanka Quechua. In addition to using actually to advocate for the inclusion of propositional content in the common ground, speakers also use actually for a broader discourse purpose: namely, to control the discourse topic. In section 4.2, I show that speakers use actually to control the  149 discourse by introducing or changing topics, resisting topic change (where they have more to say on the issue) and embarking on a discourse-relevant digression. To this end, I show that English functions like Korean and Mandarin.  Up to this point, I have provided a unified analysis of actually rather than arguing for an analysis based on the two distinct grammatical objects discussed in section 1.3. Recall that Greenbaum (1969), Taglicht (2001) and others have posited two types of actually on the basis of examples like the following (repeated from (1.16) and (1.17)).  (4.1) A: How was the bus ride? – B: Actually, we went by train.  (4.2) This isn’t (just) hearsay, you know; I actually saw it with my own eyes.  (from Taglicht, 2001: (1) and (2)) I argue that, regardless of its syntactic position, actually’s evidential contribution remains the same; namely, that the speaker has the best evidence for their proposition. In section 4.2, however, I show that parenthetical actually (in (4.1)) behaves differently from clause-internal actually (in (4.2)). Namely, while both actuallys are used to achieve the informational goals set out in section 4.1 (emphasizing, self-correcting, challenging, persuading), only parenthetical actually performs the topic control functions shown in section 4.2. I nonetheless argue that a unified analysis is still appropriate; I argue that the distinction between clause-internal actually and parenthetical actually is better explained by appealing to syntax, rather than by adopting two separate lexical entries. I provide some insights into what such a syntactic analysis might look like, but leave the specifics of that analysis to future research.      150 4.1 Actually and the negotiation of the common ground  I argue that the exchange of information in the course of a police interview requires negotiation in order to ensure that the witness’s institutional goal, to provide accurate information on which to advance the investigation, is achieved. I show that speakers use actually to negotiate the common ground and, in particular, to advocate for the inclusion of evidence in the common ground.  Stalnaker (1978) observes that discourse unfolds against an ever-changing background made up of propositions already confirmed by the discourse participants which forms the common ground of the discourse (Stalnaker, 1978: 185; Farkas & Bruce, 2010: 81). For Stalnaker, information is added to the common ground through the act of assertion; unless there are objections from other discourse participants, the effect of the assertion is to add the information expressed in the assertion to the common ground. Others have convincingly argued that an assertion is more properly characterized as a proposal to add information to the common ground, rather than a direct addition of information (see, for example, Farkas & Bruce, 2010, and others cited therein). Farkas & Bruce argue that characterizing assertions as proposing additions to the common ground, rather than actually changing the common ground, is “necessary to make room for the large variety of conversational moves that react to assertions signalling, for instance, confirmation, rejection or the need for further discussion” (2010: 82).  In her discussion of how evidentials in particular are used to update the common ground, Murray (2009) distinguishes at-issue content, which is negotiable, from not-at-issue content, which is not negotiable. She acknowledges that assertions containing evidentials make both an evidential and propositional contribution. She argues that the evidential  151 contribution is not directly challengeable or up for negotiation. The propositional contribution (the ‘main point’ of the sentence), however, is directly challengeable and is up for negotiation (2010: 324). In this chapter, I show that speakers use actually’s not-at-issue evidential contribution to advocate for the inclusion of the at-issue propositional contribution in the common ground. That is, I argue that, by asserting their information with actually, English speakers both (a) propose the propositional content of the assertion for inclusion in the common ground and (b) advocate for its inclusion by marking that the speaker has best evidence for that propositional content. In this way, speakers use actually to negotiate the common ground.  The concept of ‘common ground’ has two potential applications in the context of a police interview; the common ground of discourse, as that term is generally understood in the literature, and the ‘common ground’ of the investigation. The goal of proffering information in a witness interview is to have that information included in the common ground of the investigation itself, not just of the discourse. That is, the witness’s goal is to convince the interviewer that their reliable information should not simply be included in the common ground for the purpose of the discourse, but should be accepted as evidence for the purpose of the investigation. For the purposes of this study, I observe that negotiating the inclusion of an assertion into the common ground of discourse is a precursor to including it into the common ground of the investigation; an interviewer is unlikely to include an assertion as evidence for the investigation without first having accepted it into the common ground of the discourse. Nonetheless, I raise this issue specifically in order to flag that there may be more at stake in the negotiation of the common ground during a police interview than in non-institutional discourse and that this difference may be reflected in the speakers’ linguistic  152 choices in the circumstances. In the sections below, I show that speakers use actually to negotiate the common ground of the investigation, in particular where a bare assertion will not suffice.   4.1.1  Where speakers use actually versus bare assertions  Evidentials like actually are non-obligatory in English; in most circumstances, a bare assertion will suffice for the purpose of proffering information. At issue, then, is a determination of the discourse contexts in which speakers tend to use actually. I argue, following Pomerantz (1984), that speakers use actually in those discourse contexts which require them to expressly commit to the proposition they make. In terms of participant roles as discussed in Heydon (2004, 2005), a speaker uses actually to commit to being the author or principal of the information. Such express commitment is not always, or even often, required in normal discourse. Indeed, as Schiffrin (1987: 18) observes,  … [a]lthough a key part to a position is an idea, i.e., descriptive information about situations, states, events, and actions in the world, another important part is speaker commitment to that idea. The simplest display of commitment is through an assertion, i.e., a claim to the truth of a proposition. In more complex displays, speakers indicate their confidence in that truth, e.g. by hedging or intensifying what they say.   If Schiffrin’s observation is correct, then a bare assertion in English should be sufficient to claim the truth of the proposition and to claim that the speaker has the evidence necessary to claim the truth of the proposition. Indeed, it is widely accepted in speech act theory that, in order to make an assertion, the speaker must believe the assertion to be true (Vanderveken, 1990). This is what gives rise to Moore’s Paradox, which shows that a speaker cannot perform an illocutionary act while at the same time denying (one of) its sincerity conditions (Vaderveken, 1990: 118; Faller, 2002: 159-60).   153  (4.3) Moore’s Paradox  # It’s raining but I don’t believe that it’s raining.  There is nothing logically contradictory about a proposition of the form “p and NOT(I believe that p)”, and yet sentences of that form are infelicitous, as shown in (4.3). Faller argues that the sincerity condition associated with making a bare assertion requires that the speaker believe the proposition expressed in the bare assertion (2002: 124). Schiffrin would say that, by uttering the bare proposition, p, the speaker commits themself to p; therefore, NOT(I believe that p) is pragmatically inconsistent with asserting p. The commitment a speaker makes in a bare assertion is strong enough to generate the infelicity observed in Moore’s Paradox. Therefore, for the purposes of committing the speaker to the information given, a bare assertion will usually suffice in most discourse contexts.  I argue that speakers use actually to advocate for the inclusion of their evidence in the common ground in contexts where a bare assertion will not suffice. Two such contexts are consistently attested in the police interview transcripts. First, I show that speakers use actually where their initial proffering of information with a bare assertion failed to elicit an appropriate response from their discourse partners; that is, where the bare assertion did not suffice in the sense that it failed to elicit the desired response, the speaker can reformulate the utterance with actually to emphasize their information. Second, I show that speakers use actually where something about the discourse context has explicitly or implicitly challenged the speaker’s information; these contexts require the speaker to engage in delicate discourse actions including self-correction, challenge, and persuasion. Because speakers generally use actually to emphasize the quality of their evidence where a bare assertion has proven not to suffice, actually is generally unattested when introducing information novel to the discourse.   154 4.1.2  Actually as a means of eliciting appropriate response In the first type of context, the speaker initially presents their evidence with a bare assertion and then later repeats it with actually. For example, in (4.4), the interviewer SB asks CA where she (CA), Casey Anthony, and AH were, relative to each other, in CA’s car in the ride home after CA confronted Casey Anthony about Caylee’s whereabouts. CA initially indicates that AH (Amy) was ‘behind me.’ Then, five turns later, she repeats that information with actually (‘Amy actually went behind me’).   (4.4) SB: What was, where were they sitting in the car? CA: Casey was in the front seat, Amy was in the backseat.  So, Amy was behind me because the car seat in my… we all have a car seat.  SB: Car seat, yeah.  CA: Caylee’s car seat’s on the…  SB: Behind the passenger seat. CA: … passenger side behind the passenger side, which I feel is the safest place to be.  SB: Right. CA: So, that’s where it was so Amy actually went behind me and Casey was right next to me.  ((17) CA3.37:1-11)  In (4.4), the information the CA appears to highlight is Amy’s location in the vehicle. She explains that Amy sat behind the driver’s seat because Caylee’s car seat was on the passenger side. However, SB’s response (“Car seat, yeah”) focused on the fact of the car seat not where Amy was sitting. CA uses actually to repeat her information about Amy’s location.  Similarly, in (4.5), the witness RP initially gives his evidence that Cindy Anthony called him to warn him not to talk to Casey as a bare assertion and then, in the next turn, restarts his information with actually.    155 (4.5) RP: No, I found, I found this out uhm, about a month, no not, uh three months,  maybe three, three and a half weeks before Caylee went missing.  Her mom, Cindy…  YM: Uh-hum (affirmative). RP: … called me and told me not to talk to Casey because she didn’t want me to get hurt because she’s been lying about a lot of stuff, and put her in, stole money from her and her grandmother.  And she doesn't want me to get hurt you know.  Which I, you know, I was, I was “Okay.  You know, I understand.”  And she, you know, she seemed really upset.  She was just calling to tell me all, what was going on and, and also to ask about some lies she had suspected Casey of telling.  YM: Oh. RP: Uhm, so she actually phys… she just called me.  Her mom called me uhm, you know, in my summer semester, at one of my classes.  I remember the specific time and what I was doing.  And she basically told me “I don’t, I don’t think it’s a good idea that you even talk to Casey anymore because she’s a…” uh, the actually, those are the exact words that came out of her mouth was that she was a sociopath. ((525) RP14:21-15:14)  In (4.6), MC initially gives her evidence, that she and Casey Anthony talked about Casey’s ‘boyfriend being out of town’ and then, on her next turn, gives the same evidence with actually.   (4.6) RL: Okay, and did you have a conversation with Casey?  MC: I did, uh, for uh, the duration of my lunch hour.  RL: Okay, and what was that conversation in reference to if you don’t mind?  MC: Uhm, for the most part just, uh small talk.  Uh, what was the activities of  each of our Fourth of July.  Uhm, what was going on that week.  Uh, we talked about uh, her boyfriend being out of town.  Uhm, just, uh, just, just small talk.  RL: Okay. Did she say who her boyfriend was?  MC: Uh, yes, uh, she did. Uh, it’s a gentleman by the name of Anthony I  believe. Uhm, and actually uh, uh, he was out of town at the time.   ((244) MC2:10-18) In (4.7), CA provides a list of reasons why she considered JG (Jesse) to be a viable suspect in Caylee’s disappearance. All of those reasons are introduced with bare assertions, including  156 the comment that Jesse was someone Casey ‘could fear.’ She then goes on to indicate that she would be afraid of Jesse herself, introducing that information with actually.   (4.7) CA: … We’ll let the police you know, know ‘cause I’ve been forthcoming with  um, Sergeant Allen on all my concerns with Jesse.  I’ve given him information.  He’s read cards and stuff that Jesse’s written and Jesse has been infatuated with Caylee since Caylee’s birth.  And kind of overly you know…  SB: yeah. CA: … overly connected.  And overly tried to push way into Casey’s life and I’ve got some many inconsistencies talking to Jesse, over the last couple weeks.  That stuff hat [sic] he said just doesn’t make sense.  And if it was anybody close that she trusted he was probably my first thought from day one.  And that someone she could fear.  Actually I would be afraid of Jesse he’s got a hot temper.  He has connections.  He could be someone that I would think would take…  SB: When you say he has connections to what? CA: He was, he had a Orange County Sheriff’s um… I mean, Orange… no Orlando Police connections.  He was in training he actually was in training for to be an officer… ((44) CA1.84:15-85:4) In each of the examples in (4.4) through (4.7), the speaker initially proffers their information with a bare assertion before repeating that information with actually.  Kim (2006) observes that Korean speakers do the same. As noted in chapter 1, she argues that evidential use is dictated in large part by discourse; where the addressee fails to appropriately respond to the speaker’s information given with a bare assertion, the speaker can reformulate their information with an evidential marker in order to emphasize their position.       157 Speaker:         initial claim                        (zero evidential marker: the sentence-ender –e/a  Addressee:     lack of apt uptake                        (display of disagreement, challenge, doubt, etc.)  Speaker:         reconfiguration of initial claim                        (overt experiential evidential marker –telako: achieving                         entitlement, objective or detachment)  Addressee:      uptake (reconstrual of the speaker’s modified claim)    FIGURE 4.1:  The recurrent sequential format of the -telako evidential marker  (adapted from Kim, 2006, Figure 1) While a bare assertion may be initially sufficient to proffer information, where the speaker does not believe that information has been appropriately received, they reformulate their information with overt evidential marking.   The use of actually to reformulate, and thereby emphasize, information is inconsistent with an analysis of actually as a marker of contrast. That actually functions to mark contrast or counter-expectation has been well-documented in the literature (see, for example, Tognini-Bonelli, 1993; Lenk, 1998; Oh, 2002; Taglicht, 2001; Halliday & Hasan, 1976; Aijmer, 1986). Moreover, the functions of actually I discuss in the following sections (self-correction, challenge, persuasion) are examples of actually’s contrastive function. I nonetheless argue, following Martin (1992) and Waters (2011), that actually is not inherently a marker of contrast or counter-expectation. My reasons for doing so are twofold: first, to do so would fail to account for the situations like those in (4.4) through (4.7) where actually functions without a contrastive meaning and second, including contrast as part of actually’s semantics is unnecessary.  To include contrast or counter-expectation as part of actually’s inherent semantics would be problematic insofar as it would fail to account for non-contrastive uses of actually  158 like those in (4.4) through (4.7). In those examples, the speaker uses actually where they fail to achieve the appropriate response from their discourse partners even though there is nothing overt in these examples with which the actually-sentence contrasts. In addition to the data in (4.4) through (4.7), actually’s use in questions is also inconsistent with an analysis of actually as a marker of contrast or counter-expectation. While questions will be discussed in greater detail in chapter 6, for present purposes, it is sufficient to note that, if actually were indeed a contrastive marker, we would predict that it would only be felicitous in questions where there was some prior information in the discourse that required correcting or contrasting. However, that does not appear to be the case in examples like (4.8) and (4.9).  In (4.8), GA gives his evidence (that Casey Anthony told him on the 16th that she planned on spending that night at the nanny’s house).    (4.8) GA: Oh she definitely had some stuff cause she sat it on our one chair that was  right there, one lazy boy recliner that we have.  She just set it down and she says “Hey” she says “I’m a be working a little bit late uh, Caylee’s gonna be staying with the nanny, I’m gonna come back and stay there and I’ll see you and mom tomorrow afternoon”.  That’s not nothing unusual; she’s done that a few different times so.  She said that she already had, he says [sic] “I already talked to mom, mom knows I’m gonna be staying over”, I said, “Okay just be careful and I’ll see you tomorrow type deal.” I was one of those…  EE: And that’s on the 15th?  GA: No that was on the 6…  EE: Or the 16th.  GA: That was the 16th, that was Monday. (GA2.11:12-24)  EE then asks GA a question about that evidence later in the same interview, using the actually-question in (4.9).  (4.9) EE: … these two days out but um, she, so she actually told you on the  16th, I plan on staying at Zanny’s on a late event? GA: Yes most definitely. ((149) GA2.16:15-19)  159 If actually marks contrast, it is unclear what information in the bare assertion in (4.8) the actually-question in (4.9) is intended to contrast with.   One might argue that, by using a bare assertion, the speaker implicates that they are not in a position to claim they have best evidence for the information contained in the bare assertion. The actually-marked question in (4.9) would be posed in order to confirm or cancel that implicature. However, it seems counterintuitive to suggest that a speaker who provides information about a prejacent through uttering a bare assertion is, in fact, creating an implicature that they do not have reliable evidence for the prejacent. Moreover, such a suggestion is inconsistent with the speech act theory literature which, as noted above, assumes that bare assertions include a sincerity condition that the speaker believe the assertion. Faller (2002) specifically argues that bare assertions in Cuzco Quechua implicate that the speaker does have best evidence for the prejacent. Finally, such a suggestion is inconsistent with the police interview data in its entirety. If a bare assertion created an implicature that the speaker either did not have reliable evidence for the assertion or was agnostic as to its reliability, then we would predict that most if not all bare assertions would be followed up with clarification questions, particularly where the bare assertion contains a detail that is crucial to the investigation. This prediction does not appear to be borne out in the police interview transcripts.  Take (4.10), for example. The witness AL was Casey Anthony’s boyfriend at the time Caylee Anthony disappeared. Shortly prior to Caylee’s disappearance, he went to New York and joked to Casey Anthony that he might not come back.   (4.10) AL: … I was just, just joking around, saying you know “Uh, yeah (affirmative), I,  I might have to stay up in New York,” just to get a rise out of her. And uh,  yeah, (affirmative)…  160  EE: (Unintelligible)…  AL: … she got upset about it.  EE: Anything about the child? Did she worry about the child being the reason, or?  AL: Uhm…  EE: She never make any statements like that?  AL: No, nothing, nothing like that… (AL1.5:8-16) AL’s evidence is that Casey Anthony got upset about his claims that he might stay in New York but he does not explain his information for that assertion; he proffers his evidence in a bare assertion. This information is potentially important for the investigation because it speaks to a potential motive for Casey Anthony to murder her child; she might be motivated to dispose of her daughter if she believed Caylee was the reason AL was abandoning her. Therefore, if AL’s bare assertion created an implicature that created doubt about the reliability of his information, then one would predict that the investigators would seek to clarify it, particularly given its potential importance. Instead, the investigator EE asks a follow up question that suggests he accepted AL’s information without concern. EE’s follow up question is unexpected if AL’s bare assertion created an implicature that would create doubt about the reliability of AL’s evidence. Thus, I argue, the goal of following up a bare assertion with an actually-question is not to cancel an implicature created by the bare assertion itself.  If a bare assertion does not create an implicature that the speaker cannot claim that they have best evidence for their information, then it is unclear what the actually in the question in (4.9) is intended to contrast with. Indeed, I argue that it is not intended to contrast with anything; rather, I argue that contrast is not an inherent part of actually’s meaning. Moreover, since, as I show below, the contrast or counter-expectation meaning can be  161 derived from actually’s semantics as a marker of best evidence in appropriate circumstances, building contrast into actually’s semantics is unnecessary.   4.1.3 Actually and delicate discourse actions Actually marks that the speaker has best evidence for the prejacent. Therefore, it can be used to achieve delicate discourse actions like correcting, challenging, or disagreeing with discourse partners. In this way, actually functions like the Korean direct experiential marker -telako as observed by Kim (2006). By emphasizing the reliability of their own information, the speaker is able to indirectly challenge contrary information proffered by their discourse partners. Specifically, by using actually, the speaker foregrounds the reliability of their evidence which has the effect of correcting or challenging contrary information previously given in the discourse.  I observe that speakers use actually to foreground the reliability of their evidence for the purposes of negotiating the common ground in three distinct but related ways: through self-correction, challenge, and persuasion. The term ‘self-correction’ is self-explanatory; it refers to situations in which the speaker corrects their own previously given information which is either inaccurate or has been misunderstood. I use ‘challenge’ to refer to situations in which the speaker calls into question the accuracy of information proffered, directly or indirectly, by their discourse partner. I use ‘persuasion’ to refer to situations where the speaker defends their evidence in the face of a direct or indirect challenge.   4.1.3.1  Self-correction  Actually has previously been observed to function as a marker of self-correction (see, for example, Tognini-Bonelli, 1993; Taglicht, 2001). Consistent with this observation, the  162 police interview transcript data shows that speakers regularly use actually to correct themselves when they realize they have proposed inaccurate information for inclusion in the common ground. This is true both for interviewers and witnesses. In (4.11), the witness DS explicitly marks that he is making a correction to his answer to an earlier question.   (4.11) JA: Okay.  Do you know how Casey met Ricardo?  DS: No, I do not.  JA: Okay. Alright. DS: Actually, you know what? I take that back.  Yeah (affirmative) I do.  Uhm, through Lauren Copple. ((597) DS7:13-17)  In (4.12) and (4.13), neither witness explicitly marks that they are correcting the information they proffered for inclusion in the common ground earlier, but the meaning is clear; they use actually p to indicate p should replace their earlier, contradictory information. In (4.12), the witness initially uses her maiden name to identify herself for the record, then corrects herself to indicate her married name.  (4.12) JW: Okay, today the date is July 24, 2008, and it’s 11:40 a.m.  And I’m  Detective White with the Orange County Sheriff’s office, Domestic Violence Unit.  And I’m [blocked out] with?  NL: Nicole Fournier, well actually Nicole Lett now.  Sorry.  ((480) NiL1:8-11) In (4.13), the witness LA initially suggests that the laptop up for discussion belonged to his sister. He uses actually to clarify that the laptop is, in fact, his mother’s, but that his sister had been using it; that is, the evidence he has for p (that it is his mother’s laptop) is more reliable than the evidence he gave earlier.     163 (4.13) LA: Yes. Uhm, this was uhm, around midnight.  I went and picked up uhm, my  sister’s laptop.  Well it’s actually my mom’s laptop, but you know, my sister was, she’s had it for the past month or so… ((162) LA18:19-21)  Interviewers also use actually to self-correct; the data in (4.14), (4.15) and (4.16) show interviewers correcting inaccurate information they explicitly introduced into the discourse. In (4.14), YM is interviewing a police officer about her response to a particular call. He starts the interview by indicating he is investigating a particular call. He uses actually to correct himself to more accurately reflect that he is investigating a series of calls, even though he is only asking her about one in particular.   (4.14) YM: … And we had some questions to ask you specifically about a call that we  were investigating a particular call, actually we’re investigating several calls in that area.  This is one of them. ((1) AA1:20-22)   In (4.15), SB initially suggests to the witness that none of Casey Anthony’s old friends knew the nanny that she was alleged to have hired to mind Caylee. He then corrects himself to reflect that none of Casey’s friends, new or old, knew Zanny.  (4.15) SB: None of her friends outside of the new friends actually none of the  friends know who Zanny is. ((145) GA4.80:23-24)  In (4.16), the interviewer, EE, is asking the witness, WW, about the events that followed a party on July 4. Specifically, he puts it to WW that Casey Anthony left his apartment around 1:30 in the morning after the party ended. He initially indicates that was July 4, but then corrects that statement, using actually, to indicate he means July 5.   (4.16) EE: Saturday.  The 5th was a Saturday.  We’ll get…  WW: But…  164 EE: … into that.  Did you, so she left your apartment with Amy around 1:30 you said? WW: Yes.  EE: In the morning, on the Fourth?  WW: Uh-hum (affirmative).  EE: Well it would actually have been the 5th by then, the Fourth of July party?  WW: Yeah (affirmative). ((611) WW5:14-21)  In the data in (4.11) through (4.16), the speaker corrects information they explicitly proposed for inclusion in the common ground; that is, they correct their prior explicit statement by introducing the corrected answer with actually.   Speakers also use actually to self-correct information proffered by implication. Where a speaker makes a statement which might create an implicature the speaker knows to be inaccurate, they can use actually to cancel the implicature. For example, in (4.17) the witness AH indicates that Casey Anthony regularly invited her to a nightclub but that AH never wanted to go because she did not like the type of music they played. Her answer was capable of creating an implicature that she had been to the nightclub before. Indeed, the interviewer EE’s subsequent question about whether they played techno music suggests that such an implicature may indeed have been created. AH uses actually to cancel the implicature before it can be included in the common ground.   (4.17) AH: She invited me every single Friday.  I never wanted to go because I don’t  like the music they play there.  I know what kind of music it is and it’s not a place that I would have fun at.  EE: Ecstasy type of stuff? AH: Uh, I think it’s just more like rap, hip-hop music that I just don’t like.  I don’t enjoy it.  EE: Techno, anything?  AH: I don’t think so.  EE: No?  AH: Not that I’m aware of.  EE: Okay.  AH: I, I haven’t, I haven’t actually gone but…  165  EE: Huh-huh (negative) AH: … from the party that we had gone to that first night we met Tony, and knowing Tony and his friends and the kind of music that all of a sudden Casey started listening to, not music that I like. ((373) AH2.14:21-15:12) It is unclear in this example what implicature is being cancelled: the implicature AH initially created or the implicature in EE’s follow up question. For our purposes, it is unnecessary to determine which specific implicature is being cancelled, assuming such a determination is even possible. It is sufficient that AH uses actually to cancel an implicature created in the course of the discourse and that, by cancelling the implicature, AH not only corrected the information before it could be included in the common ground but also provided an explanation as to why she was unable to answer EE’s question about the type of music played there.  Similarly, in the background leading to the example in (4.18), YM explains that the investigators found a receipt that had belonged to one of AL’s roommates, NL, in the garbage bag in the trunk of Casey’s car. The investigators were attempting to determine how Casey came to be in possession of that receipt. AL suggests that she likely put it in the garbage and then took the garbage out. The interviewer YM then asks why Casey would have put the garbage in the trunk of her car as opposed to the dumpster and whether that was normal behaviour for her. That is, YM appears to interpret AL’s statement, that Casey likely put the receipt in the garbage, to include the fact that she also took the garbage out.  AL cancels this implicature using actually.  (4.18) AL: She probably got it [JG: NL’s receipt] from the garbage.  She probably threw  it in the garbage bag and then put it, or she was cleaning up.  Because she was cleaning my apartment all the time.  YM:  Okay. I’m just curious why she would put a garbage bag in the car as opposed to just taking it to the dumpster?  And she was coming from her parents  166 house.  So that would mean that she would have had to put the garbage bag in the car and then run out of gas as she’s driving back up north.  And I was just curious, do you, do you, if this rings a bell to you.  If you, you know was she normally the one that took your garbage out? I mean (inaudible)  AL: Uh, we all did.  Me, Cam, Nate.  Uhm, she…  YM: Yeah (affirmative) AL: She would clean up.  I mean I never actually saw her take garbage out.  She could have.  I mean to do, like to throw garbage out like to go, the dumpster…  YM: Uh-hum (affirmative)  AL: … at my place is all the way in the front.  ((445) AL3.35:12-36:2)  As was the case in (4.17), AL used actually to cancel the inaccurate implicature created earlier in the discourse while at the same time explaining why he was unable to answer the interviewer’s follow-up questions.   A witness who realizes they have created an inaccurate implicature can use actually to cancel that implicature, even without feedback from the interviewer. In (4.19), for example, the witness RP indicates that Casey Anthony took a cheque from her grandmother’s cheque-book and tried cashing it. His information, as stated, could create the implicature that Casey Anthony was unsuccessful in her attempt to cash the cheque. He cancels the implicature, even before the interviewer asks a follow up question, with actually.  (4.19) RP: … Well, my mom told me that uh, Casey stole money from her in a check.   But then she also somehow got into my dad’s checkbook and that, that’s a dedicated checkbook to the nursing home.  And only my mom can write checks from that and it has to be for the nursing home that he’s in.  Because that was set up, it’s a separate account set up for my dad and the nursing home.  And somehow Casey got a check out of that and she tried cashing it.  And actually she did cash it… ((554) RPl.6:19-25)   That actually can function to mark a correction to explicit information and to cancel an unintended implicature falls out from its semantic contribution as a marker of best evidence. The proposition introduced by actually is marked as one for which the speaker has  167 what they consider to be best evidence; thus actually enables a speaker to help the hearer choose between two potentially conflicting propositions by signaling that the actually-marked proposition is to be preferred. Moreover, by introducing the corrected proposition with actually, the speaker is able to make such a correction without overtly admitting to having made an error.   4.1.3.2 Challenge  Just as speakers use actually to self-correct, they also use actually to challenge information provided by other discourse participants. This, too, is a well-documented observation about actually’s discourse function (see, for example, Oh, 2000; Taglicht, 2001).  English speakers in the police interview data use actually to challenge, correct, or disagree with their discourse partners. Actually provides speakers with a linguistic mechanism for indirectly challenging information others have proposed for inclusion in the common ground that the speaker believes to be inaccurate. Watts (2003: 275) defines ‘indirect utterances’ as those whose truth-conditional content does not directly express their communicative purpose. I argue that a speaker who challenges the information provided by their discourse partner with actually is offering an ‘indirect’ challenge. Rather than explicitly expressing their communicative intent (you’re incorrect or your information is wrong), they indirectly achieve that result by foregrounding the reliability of their own evidence and allowing their discourse partner to draw their own conclusions about the speaker’s communicative intent. This discourse strategy is used by both interviewers and witnesses to clarify what they perceive to be a mistake or a potential for mistake before it is accepted in the common ground.    168  In (4.20), the interviewer, YM, is asking the witness, AR, about certain instant message conversations he had with Casey Anthony. AR indicates that he was unable to remember what he was referring to in the particular conversation at issue and suggests he may have been drunk at the time, given that he believed the conversation took place at 1:03 in the morning. YM challenges AR’s information by using actually to clarify that the conversation took place at 1:03 in the afternoon.  (4.20) YM: This is a different format…  AR: Uh-hum (affirmative).  YM: … that’s over there, mind you, but uh, since you’re NY Italiano Three…  AR: Uh-hum (affirmative)  YM: … uh, 1:03 p.m., I don’t know what date…  AR: Yeah (affirmative).  YM: … it says.  AR: Yeah (affirmative). It said…  YM: Uhm…  AR: … yeah (affirmative), I went downtown and got (inaudible). YM: Yeah (affirmative).  Do you remember what, what, what day this was when you went downtown and got (unintelligible). AR: Oh, no, I, I don’t.  I don’t.  That was probably one day that, I was probably obviously probably more towards the latter and being drunk.    YM: Okay.  AR: Because it’s one o’clock in the morning.  YM: Well, that’s one o’clock in the afternoon, actually, so.  AR: Oh, is it? Oh…  YM: Yeah (affirmative) yes.  Hopefully you weren’t getting drunk that early.  ((592) AR17:9-18:2) YM challenges AR’s evidence (that the conversation occurred at night) not by telling AR he is wrong but by highlighting the reliability of YM’s own contrary information (namely, that the conversation occurred in the afternoon).   In (4.21), CA’s evidence is that she was prohibited from giving Casey Anthony a bible and, on that basis, she concludes that Casey Anthony is not permitted to have a bible in  169 prison. SB challenges that conclusion by pointing out that the prison would provide Casey Anthony with a bible and he uses actually to do so.  (4.21) CA: Yeah she [Casey] reads the bible more then [sic] any of us.  In fact we tried to  get her bible to her and they won’t let her have it.  I gave her… in fact I was so shocked when I opened her bible and my picture was the one that was in her bible.  SB: Actually they will have one for her there [JG: in prison]. ((48) CA1.114:6-9)   The data in (4.22) come from the second of two interviews with the witness RC who was a deputy with the sheriff’s office and who responded to the initial call in which the meter reader, Roy Kronk, claimed to have found human remains. RC was interviewed twice in one day, once in the morning and again after the interviewers spoke with his back up officer and determined the information he gave in the earlier interview had not been accurate. The interviewer, YM, challenges the accuracy of the information given in the earlier interview by indicating that they want to talk to him now that he has had an opportunity to think about “what actually happened” on the day in question.   (4.22) YM: … We spoke to you this morning about a, a call that you were here on  back on the 13th of August. Uh, do you remember that conversation?  RC: Yes.  YM: Okay. And we, at the time we showed you a, and a call log, and we went  through the times. Uh, your back up officer who was?  RC: Kethlin.  YM: Kethlin. Uhm, you remember the call?  RC: Yes.  YM: Okay. Uh, it’s been a couple of hours, and I think during that time you’ve  had some chance to think about the call, and think about what actually  happened. And I understand that from the onset what you originally told  us, it differs a little bit from what you told us now because you’ve had a  little bit more time to think about it, correct?  RC: Yes, yes. ((199) RC2.1:14-2:3)  170 In (4.20) through (4.22), the interviewer challenges the witness and their evidence by introducing his own evidence with actually. In (4.20) and (4.21), he uses actually to introduce information that directly conflicts with the witness’s information. In (4.22), he uses actually to indicate that the information the witness previously provided was unreliable.68  Like interviewers, witnesses use actually to challenge information they know to be inaccurate. The interactional act of challenging is a particularly delicate one for witnesses because of the tension between their institutional purpose and their limited institutional power. As stated, the institutional role of the witness is to provide information on which to advance the investigation. Therefore, when the investigator proffers information the witness believes to be inaccurate, the witness has an obligation to correct that information for the record. However, a direct challenge would be incompatible with the witness’s limited institutional authority in the discourse and politeness norms generally. By using actually to highlight the reliability of their own information, the witness is able to indirectly challenge the information proffered by the interviewer which allows the interviewer to save face in the circumstances.  For example, in (4.23), the interviewer, AW, is explaining to a witness, KC, that he was unable to locate a driver’s license for a woman named Casey Williams. KC explains that Casey Williams did not, to the best of her knowledge, drive; therefore, she did not have a driver’s license. She uses actually to introduce this information.                                                   68 It is worth noting that none of the witnesses in (4.20) through (4.22) were suspects; none of CA, AR, or RC were believed to have been involved in the disappearance or murder of Caylee Anthony. Therefore, while it was important for the interviewer to challenge the accuracy of their information for the purposes of the investigation, there was no expectation that CA, AR, or RC were being deceitful in order to hide their involvement in the crime. Had that been the case, the interviewer may have chosen a more direct method of challenging their information, for example, by accusing them of lying.    171  (4.23) AW: I have ah… hm I got a Casey.  Casey Williams in Panama City.  Casey  Williams in Umatilla.  Casey Williams in St. Petersburg and a Casey Williams in Lauderhil.  She must not have a driver’s license from here. KC: I don't think she actually drove, to be honest with you.  She took the bus everywhere.  She took the Lynx. ((266) KMC2.6:22-7:2)  Note that, in addition to actually, the witness KC uses a number of hedges (don’t think, to be honest with you) when making her point that Casey Williams did not have a driver’s license.   One particular discourse situation in which the witness is compelled to challenge the interviewer is where the interviewer puts a leading question to the witness that is premised on an inaccurate presupposition. In such cases, witnesses regularly use actually-marked clauses to challenge the underlying presupposition. A leading question is one which presupposes some proposition that the addressee may not share. Leading questions can be either yes/no questions or Wh-questions, as shown in (4.24).  (4.24) Leading questions a. Y/N Question: Did Casey smoke pot with you almost every day?  b. Wh-Question: What kind of drugs did Casey like? (4.24a) is a leading question because it presupposes that both the addressee and Casey smoked pot. (4.24b) is a leading question because it presupposes that Casey liked drugs. From a linguistic perspective, leading questions have traditionally been studied in two specific contexts: courtroom discourse and false confessions. Less linguistic study has been devoted to leading questions in naturally occurring speech.  Leading questions have a very specific function in legal discourse; namely, they are used to commit witnesses to certain facts on the record. In court, leading questions are used  172 primarily in cross-examination;69 the opposing lawyer uses a series of leading yes/no questions to get the witness to agree to facts that either tend to incriminate them or call their credibility into question. For this reason, cross-examination is often derisively referred to as the testimony of counsel, punctuated with “yesses” and “nos” from the witness. Leading questions are used in pre-court proceedings like police interviews and depositions in order to create a record of the witness’s evidence from which they cannot later resile without contradicting themself. If the witness tries later to offer contradictory evidence, their previous statement can be used to impeach them and decimate their credibility. Similarly, and more controversially, investigators have been known to use leading questions in the interview of a suspect. Like in cross-examination, the goal of leading questions is to force the suspect to commit to certain facts about the crime in order to convince the suspect to confess. This practice is controversial because any resulting confession is arguably not ‘voluntary’; in these situations, it is the investigator, and not the suspect, who puts the relevant information into the interview discourse (Hill, 2003; Shuy, 1998).   The investigators in the police interview data regularly put evidence to witnesses in the form of leading questions, although possibly for different reasons than those described above. It is clear from the transcripts that, for the most part, the recorded interviews reflect only a portion of the conversation between the witnesses and the interviewers; the interviewer and the witness talk off tape and determine what, if any, relevant information the witness could offer. Once the interviewer determines what information was relevant to the investigation, they begin recording and focus the interview to elicit that information. Therefore, the interviewer already has a sense of what the witness is going to say before the                                                  69 Leading questions may also sometimes be used in direct examination in order to establish non-controversial facts for the record.   173 recording began. The interviewer then uses leading questions to direct the witness to the information in order to establish it on the record.   However, it is clear from the police interview data that, while the interviewers regularly put previously discussed information to the witness in the form of a leading question, they often do so inaccurately. The inaccuracies may have arisen because the witness’s information was unclear to begin with or because it was misunderstood by the interviewer. Regardless of how or why they arise, these types of inaccuracies require the witness to challenge the presupposition on which the leading question is premised in order to fulfill their institutional role. The data show that witnesses regularly use actually to do this.   Consider the data in (4.25), in which the witness AL is asked the leading question from (4.24a). Recall that (4.24a) contained the presupposition that both he and Casey Anthony smoked pot.   (4.25) AL: I think the only thing that really was putting, was being, going through,  like, or people were coming over to, to do was smoke.  EE: Smoke pot?  AL: Uh-hum (affirmative) EE: Okay. Did you guys, did she smoked put [sic] with you almost every day, or just before you went…  AL: I actually…  EE: … clubbing? AL: I actually kind of stopped smoking pot around, I don’t know around April.  So I uhm, I only remember a couple of times of me actually smoking week [sic].  I might have…  EE: With her? AL: With her.  I remember coming home a couple times with her being, she was    high.  Like when I get out of class at 1:00 am she was smoking with my roommates. ((453) AL3.67:17-68:6) AL cannot simply answer the question from (4.24a) with a yes or no answer because to do so would be to accept the presupposition that both he and Casey Anthony smoked pot. Rather,  174 his information is that, while Casey Anthony regularly smoked pot, he stopped doing so months previously. Therefore, instead of answering yes or no, he uses an actually-marked clause to challenge the presupposition underlying the leading question.   The witness JR was asked a leading question like that in (4.24b), which presupposed that Casey Anthony liked drugs. It is clear from JR’s answer, given in full in (4.26), that she did not know whether Casey Anthony did drugs and challenges that underlying presupposition with an actually-marked clause.  (4.26) YM: Okay.  Do you know anything about Casey uhm, what kind of drugs  Casey liked? If she like uhm, what she liked to use, or what she like to…  JR: (Sighs).  YM: … to pop, or what she liked to smoke? JR: No.  Actually, during the course of knowing Casey I never, ever recall her mentioning drugs at all.  I didn’t even know she did drugs.  I mean, I knew from the photos that she drank but…  YM: Uh-hum (affirmative). JR: … other than that I didn’t know anything about drugs.  Never was offered or anything.  ((573) JR7:21-8:6) Similarly, in (4.27), the interviewer’s question presupposes that it was Roy who found a snake in the swampy area near where they were working. The question was intended to determine at what point Roy spotted the snake. However, the presupposition that Roy found the snake was inaccurate; the witness, CG, responds that it was David that spotted the snake, not Roy. He uses an actually-marked clause to challenge the inaccurate presupposition.  (4.27) YM: And as far as you rec… can recollect, he was on, Roy was on to go take a,  to, to go relieve himself when he saw the snake? Is that, is that… CG: Uh, he was on his way up to use the restroom.  Uh, he walked by the snake.  It was actually David who spotted the snake.  YM: Uh-huh (affirmative) CG: And I actually don’t know if he, if it was before or after he went to the bathroom when he actually s… when David actually seen the snake and pointed it out.   175 ((305) CG2:23-3:5) In (4.28), the interviewer, YM, asks the witness, AL, how many times he drove Casey Anthony’s car, a leading question which presupposes he had driven her car in the past. AL’s evidence is that he never drove her car. He initially answers “I never drove…” and then restarts his turn with “I actually never drove her car.”   (4.28) YM: … about, about day-by-day, I kind of had a general question.  From the 16th,  up until the 27th, how many times did she drive, or did you drive her car, were you in her car? The 16th to the 27th?  AL: I never drove, I actually never drove her car.  ((435) AL3.19:1-4) I argue that, by using actually, AL’s challenge to the presupposition on which the leading question is premised is perceived as more polite than a direct challenge would be. Thus, in (4.25) through (4.28), the witnesses were able to challenge the inaccurate presupposition without acting in a manner inconsistent with their institutional authority.   Taglicht (2001: 3) observes that using actually to correct someone else serves to make the correction “less likely to offend.” Oh (2000: 24) similarly describes actually as a pragmatic ‘softener’ when “used with face-threatening acts” like challenging or contradicting. In his comprehensive survey of linguistic politeness, Watts (2003) notes that there are generally three types of politeness marking: ‘metapragmatic’ politeness, ‘classificatory’ politeness, and ‘expressive’ politeness. Unlike metapragmatic or classificatory politeness, the goal of expressive politeness is to produce explicitly polite language (Eelen, 1999). Watts (2003) observes that expressive politeness can be achieved where speakers use formulaic language to expressly adopt a respectful or polite stance to the addressee. Consider the following.    176 (4.29) a. Supposing you say to me      supposing…        <low burp> … \beg your pardon\  b. Yes, \can I come back to Mandy’s point\ because I think this is one  aspect of TVEI which has been totally underemphasized tonight…  (adapted from Watts, 2003: (2) and (3))  Watts points out that the formulaic phrase beg your pardon is obligatory in (4.29a) where the speaker does not want their behaviour to be seen as ill-mannered or boorish. Thus, he argues, the use of beg your pardon is socially obligatory in (4.29a). However, in (4.29b), no one would have considered the speaker to have been impolite by simply returning to the moderator Mandy’s point without expressly indicating his intention to do so. Watts observes, in this case, that the use of can I come back to Mandy’s point served a strategic social purpose; namely, an overt expression of politeness. In discourse contexts where this type of formulaic language is optional, its use signals that the speaker is strategically intending to present as polite and respectful. I argue this is what English speakers do when they use actually to challenge information or cancel implicatures that have arisen in the discourse; namely, I argue that actually in this context is a marker of expressive politeness.   In English, politeness is often marked by the use of indirect speech, including hedges. When used to challenge information or presuppositions, a witness’s use of actually as a strategy to mark expressive politeness is motivated in two ways. First, since witnesses have limited institutional authority in police discourse (Heydon, 2004, 2005), they may be motivated to speak ‘politely’ simply as a result of their institutional role. Using actually mitigates a challenge, as described above, and is therefore itself a marker of expressive politeness. Speakers use actually to challenge the information proffered by their discourse partners by using actually to mark their information as reliable (i.e., I’m right), and leaving it  177 to the addressee to draw the appropriate conclusions (i.e., you’re wrong). As such, actually allows a speaker to challenge information indirectly and, therefore, more politely.  Second, the use of actually mitigates against the potential uncertainty effects of hedging. Hedges are “linguistic expressions that allow speakers to avoid being too direct in … [their] utterances” (Watts, 2003: 274). Hedges have the effect of suggesting the speaker is unsure about the proposition they introduce and effectively minimize the speaker’s commitment to the proposition they are making. Thus, where the speaker’s goal is to correct information for the record, the use of hedges is counterproductive because hedging may be interpreted as an indication of uncertainty. I argue that actually ameliorates the uncertainty effect of hedging by explicitly marking that the speaker has best evidence for their proposition. Thus, as a politeness strategy, the use of actually allows witnesses to confirm their commitment to the quality of their information while nonetheless expressing themselves with appropriately polite speech.   In (4.30), for example, the interviewer, EE, is asking the witness, AL, about Caylee Anthony’s location while Casey Anthony was staying at his apartment. Specifically, EE asks about an incident after July 5 when Casey Anthony indicated she and Caylee stayed in Cocoa Beach for a birthday party. AL’s evidence is that the Cocoa Beach incident occurred in June, not after July 5. AL challenges the presupposition that the Cocoa Beach incident occurred after July 5 by using actually and a series of hedges.  (4.30) EE: … You said that she, after the 5th [of July], or during the time she started  staying with you more, the child was never over there? You had said that uh, there was times when the nanny had the kid at Disney on a weekend pass?  AL: Yes.  EE: And then they did a weekend stay over at Cocoa Beach for a birthday  party?  AL: Yes.  178  EE: Can you tell me about those again?  AL: Uh, basically it was either uhm, I don’t think it actually was in July. I  think that happened in June.  EE: June?  AL: Yeah (affirmative).  But uh, I remember her bringing up uh, I asked her  one day.  She said she was at Disney.  And then I remember actually asking her the second day, right after that day, and she said that no, she’s going to now whatever parks that there are in Disney…  ((406) AL1.5:23-6:6)  Although appropriate for politeness purposes, the use of hedges like I don’t think/I think could have the unintended effect of indicating uncertainty; namely, that AL was not certain that the Cocoa Beach incident happened in June and not July. However, by introducing the hedged phrase with actually, AL is able to cancel the uncertainty effect of the hedging and foreground the reliability of his evidence to challenge the presupposition created in EE’s series of questions.   Similarly, in (4.31), JW’s question to the witness AH about how Casey Anthony behaves when she is drunk presupposes that Casey Anthony drank to excess when she was with AH. AH responds by pointing out that Casey rarely gets drunk, and that AH had only seen her really drunk “like maybe two or three times” despite the fact they regularly drank together.   (4.31) JW: Okay um and you seen her drink some alcohol?  AH: Um huh (affirmative) JW: Have you ever seen what does she act like when she’s you know under the influence?  AH: She actually doesn’t get drunk very easily.  JW: Um huh (affirmative) AH: Um I’ve only actually seen her real drunk like maybe two or three times and we drink together a lot. ((354) AH1.12:17-24)  AH uses the hedges like, and maybe, along with actually, in correcting JW’s information.  179  In (4.32), the witness KW’s initial evidence was that when he searched the wooded area where Caylee Anthony’s remains were ultimately found, he stayed by the fence line and looked into the woods. However, it became clear during the second interview that he ventured farther into the woods than he originally indicated. When JA repeated his initial information back to him for confirmation, KW corrected his earlier information using actually and a series of hedges, including you know, but, and I mean.  (4.32) JA: When you say the whole foot area, you said you walked, you were, you stayed  fairly close to the wood uh road and looked back as you walked. KW: Yeah I stayed, I stayed close to the road but I was looking at one foot area of the woods.  JA: Okay. KW: You know but I mean as far as where the fence line is I actually went, you know towards the woods.  I didn’t actually stay at the fence line. JA: So when you got back here on this side of the curb towards the school you walked off the road and back into the woods?  KW: Yeah I walked that far… JA: Okay. KW: … beyond the fence line, was looking the whole foot… JA: Okay. KW: … area to the woods.  You know I actually circled the whole, the entire street I circled. ((628) KW2.8:5-20)  The police interview data show that speakers use actually to indirectly challenge information proffered by their discourse partners that they believe to be inaccurate. Rather than directly challenging the controversial information (i.e., by telling their discourse partner their information is incorrect), which would violate politeness norms and, in the case of witnesses specifically, be inconsistent with their limited institutional authority, speakers use actually to emphasize or foreground the reliability of their own information. That is, speakers use actually to indicate that they are correct, rather than to expressly announce that the addressee is wrong. The speaker then allows the addressee to assess for themself the  180 implication of that contradiction. By using this discourse strategy, speakers are able to negotiate the common ground while maintaining politeness norms consistent with their institutional authority. To this end, actually performs the same discourse function as the direct experiential marker in Korean. Kim (2006) observed that speakers use the Korean direct evidential  -telako to carry out delicate discourse actions like challenging, disagreeing, or correcting. She argues that, by using -telako, the speaker is able to detach themself from the information contained in their utterance and expect the addressee to figure out its consequences (at pages 56-57). She gives the example in (4.33) below which involves a conversation between a professor (H) and a teaching assistant (S) about the workbooks necessary for a Korean language class. The professor is explaining that the students will not be able to do homework assignment number two because they have not yet been able to purchase their workbooks. The teaching assistant replies that she has directly observed, for other classes, that the workbooks are put on course reserve. That is, rather than directly challenge the professor’s conclusion that the students cannot do their homework, she puts an alternative (i.e, putting the workbook on course reserve) to the professor using -telako.   (4.33) Korean 01 H: number two <chayk-ul mos kwuhay-ss-unikka       book-ACC not get-PST-since     ha-l  swu-ka  eps-canh-a   ku-ci?   do-ATTR way-NOM not:have-you:know-IE  be:so-COMM  02  “Up to homework number two, we cannot do much about it since (the  students) did not get their (work) books, you know. Right?”   03  (.2)   181 04 H: number two <kkaci-nun nuc-eto  co:tha       up:to-TOP later-even:if okay:DC   “It’s okay even if (the students) submit their homework late up to  homework number two.”  05  (1.2)  06 S: talun class po-nikka,   other  see-when  07  book-ul reser:ve-lul ha-y-nwa-ss-te-la-kwu-yo?   book-ACC reserve-ACC do-CONN-put-PST-RT-INTROS-QT-POL “In the case of other classes, [I observed/noticed] the books were in reserve.”  08 H: [um]   “mm”  09 S: [college]-ey?   college-at   “At the college (library).”  10 S: [work-]   “Work…”  11 H: [kulen] pangpep-to iss-nuntey,   such way-also exist-but      >wuli-n  cikum  reserve ha-l   chayk-to eps-e-yo<     we-TOP  now    reserve do-ATTR  book-even not:have-IE-POL “There is such a way (to do), but we don’t even have any books to reserve now” (adapted from Kim, 2006: (10))  Kim observed that, in (4.33), the teaching assistant avoided directly challenging or disagreeing with the professor by couching her response as an observation based on direct (and presumably objective) evidence.  Both the Korean direct evidential marker -telako and the English best evidence marker actually function similarly in discourse contexts. Both are used to mitigate the effects of challenging information proffered for inclusion in the common ground, whether that information is expressed explicitly, by implication or through a presupposition. By  182 foregrounding the reliability of their own evidence, the speaker is able to challenge the accuracy of the information provided by their interlocutors in a way that maintains politeness expectations.   4.1.3.3 Persuasion The police interview data show that, in addition to using actually to launch a challenge, speakers can use actually to defend their evidence in the face of a challenge; this is true whether the challenge to their evidence is direct or indirect. To this end, actually in the police interview transcripts functions to emphasize the speaker’s information, as previously observed in Quirk et al. (1985).70 Although there are few examples, the police interview transcript data shows that speakers can use actually to defend their information where it has been directly challenged by their discourse partner. However, speakers are also observed to use actually as a tool for persuasion where the challenge to their evidence arises indirectly.  The data in (4.34) is the clearest example in the police interview transcripts of a direct challenge to the witness’s information. In this case, the interviewer, AP, is not a member of law enforcement but is herself a witness in the investigation. Both AP and LB were volunteers who participated in the search for Caylee Anthony through Texas EquuSearch. After Caylee’s remains were found, an issue arose with respect to statements made and certain Texas EquuSearch documents prepared by LB. Specifically, LB advised the media that the area in which Caylee’s remains were ultimately found had been searched in late September 2008 and the remains had not been there at the time. The implication of this                                                  70 I define ‘persuasion’ as a response to a challenge, which distinguishes it from the emphasis function of actually discussed in section 4.1.2.  183 statement was that Caylee Anthony’s remains had been moved to that location after the search had taken place. This was important for the purposes of the investigation because Casey Anthony remained incarcerated between October 14, 2008 and December 11, 2008, the date on which the remains were found. Had the remains been moved between those dates, it would mean that someone other than or in addition to Casey Anthony was involved in the crime.  To confound things further, investigators were suspicious of the search result form prepared by LB and submitted to Texas EquuSearch as a result of the September 2008 search of the material area. Although it purported to report the results of the September search, it was not submitted until well after the fact. Most importantly, however, it listed one of the searchers, AP, by her maiden name. AP was adamant that she did not use her maiden name in her dealings with Texas EquuSearch, that LB was the only person from the search team who knew her maiden name, and that LB did not know AP’s maiden name until October at the earliest (more than a month after the impugned search had been conducted), when AP emailed LB from an out-of-date email address. Based on AP’s information, investigators believed that LB manufactured the September search result form only after Caylee’s remains were found and that her claim the remains had not been present in September 2008 was not credible.  LB refused to cooperate with investigators and did not make herself available for questioning. As as result, the investigators arranged for a controlled telephone call between AP and LB in which AP would confront LB about the suspicious report form and LB’s claims to the media. In (4.34), AP confronts LB about LB’s use of her maiden name.    (4.34) AP: I mean definitely, like, and I thought about it and I looked back and all my   184 emails, Laura, and, you know what I saw, I saw that you’re the only person; I didn’t email you; you wouldn’t have known my name until October, Laura, and that search was in September, so how?  LB: No, you actually told me, you actually told me…  AP: And you (inaudible).  LB: … your maiden name in Florida. AP: No, I didn’t, no, I didn’t because I was married by then. I wouldn’t have said, this is my maiden name. I told you my name is Anne.  ((546) AP/LB2.6:17-25) When directly confronted about AP’s maiden name, LB uses actually to defend her evidence that AP had told LB her maiden name prior to October. That is, LB uses actually to justify her information in the face of a challenge.  The data in (4.35) through (4.40) occur in the context of a series of exchanges between Casey Anthony’s mother Cindy Anthony and an FBI interviewer, SB. The data show that Cindy Anthony uses actually to defend her evidence even though there appears to be no direct challenge in the discourse. At the time the interview occurred, Casey Anthony had already been charged with murdering her child, though Caylee’s remains had not yet been found. In these extracts, Cindy Anthony is attempting to persuade the investigators that the decision to assume Caylee had been murdered was in error and that they should continue their investigation as if Caylee were a missing person. Even though the interviewers had not directly challenged Cindy Anthony’s information at that point in the discourse, it is clear from the context in which discourse was situated that the investigators did not believe Caylee Anthony was still alive.  To put the specific discourse cited below into context, Cindy Anthony’s information was that, when Casey Anthony initially advised her that Caylee Anthony was missing on July 15, 2008, Casey Anthony explained that Caylee had been abducted a month prior by her “nanny”. None of Casey Anthony’s friends or family had ever seen or met the nanny and  185 knew her only by the nickname “Zanny”. For a period after Caylee Anthony was initially reported missing, Casey Anthony refused to speak directly about the abduction, instead advising Cindy Anthony that she had been providing “clues” as to Caylee’s abductors and her whereabouts throughout the month between the date Caylee had been abducted and the date she was reported missing. Casey Anthony encouraged Cindy Anthony to reflect on those conversations to determine the clues that she had revealed.  In the interview excerpts below, Cindy Anthony is explaining to SB the results of that reflection. Specifically, Cindy Anthony has convinced herself that the name “Zanny” does not refer to a specific person but rather is a code-name for “babysitter”; that is, Cindy Anthony was convinced Casey Anthony used the name “Zanny” to refer to whoever was minding Caylee at a particular time. Moreover, by referring to a specific apartment depicted in a series of photographs as “Zanny’s apartment”, Cindy Anthony believed that Casey Anthony meant to indicate that it was the apartment of whoever was minding Caylee. Cindy Anthony then learned, during the course of the investigation, that the apartment depicted in those photos belonged to Casey Anthony’s ex-boyfriend, Ricardo Morales (RM), and her friend, Amy Huizenga (AH). Based on this knowledge, Cindy Anthony reasoned that RM and AH were the “babysitters” who abducted Caylee; i.e., that “Zanny” was a code-name for RM and AH. To put the conversation between Cindy Anthony and SB further into context, I include an earlier portion of the interview, in (4.35).    (4.35) CA: Um, so I… that stood out in my mind that several months ago before all of  this happened.  You know she said that that was Zanny’s apartment.  So, the picture right here, Zanny’s apartment.  And again Zanny may not be Zaneida.  Zanny may be referring to the nanny or the baby sitter someone watching over them.  Um, so Zanny’s apartment right now to me in that photograph is a big  186 clue okay.  That picture right there.  This was again stated before it all happened.  That, that was Zanny’s apartment.   SB: And how old is this picture? CA: Probably… this is, this dress is only since the beginning of the year so probably between March and May.  Taken somewhere in that time.   SB: Okay.  CA: Okay.  And um, and like I said that’s…  SB: Casey, Casey told you that was Zanny’s apartment? CA: Yes, I asked her specifically you know where that was taken ‘cause I was interested in the drums.  You know because I… I didn’t recognize that apartment.   SB: Uh-huh (affirmative). CA: Well, I just found out um, a week ago that that’s Ricardo’s apartment.  Ricardo was someone that Casey had been seeing and had just broken up with at the…  SB: So, this is Ricardo…  CA: … the week… Ric.. .the week before Caylee went missing. SB: Okay how, how do you know… how did you find out this is Ricardo’s apartment? CA: Because um, Amy told me.  Amy was the person that took me to Casey on the 15th of July.  And Ricardo confirmed that when he was at my house a week ago that, that… that indeed is his apartment.  SB: Okay.  CA: Okay, so that’s Ricardo’s apartment.  SB: Right.  CA: That’s a fact.  SB: So, it wasn’t…  CA: Ricardo Morales. SB: … Zan… it wasn’t Zanny’s apartment like Casey said it was somebody else’s? CA: Yeah someone else’s.  Ricardo Morales’ apartment.  Okay which is very key because the address that she took the detectives to on Glenwood.  Ricardo lives in that complex.  Ricardo and Amy are roommates now.  I just found out that Ricardo and Amy are roommates. (CA3.26:15-27:23)  With the full context of the conversation in mind, it is clear Cindy Anthony uses actually p in the excerpt below to defend her information with respect to Caylee Anthony’s abduction in the face of SB’s anticipated skepticism.   (4.36)  CA: … I mean I heard Ricardo is one of Casey’s friends.  You know realized   187 they had been seeing each other a little bit.  And did not know that Ricardo’s apartment was actually what Casey was referring to as the baby sitter, Zanny’s apartment.   SB: I’m still… I’m still a little bit confused.  CA: I know you are.  SB: Clarify for me.  CA: I know you are. SB: Why… why you’re… why do you think at this point that Ricardo’s apartment is Zanny’s apartment? CA: Because Casey stated that the picture of Caylee was in the apartment of Zanny, which I’m thinking Zanny could be a generic name for the baby sitter.  Someone that was watching Caylee for the day.   ((18) CA3.38:19-39:6)  It is clear from the totality of the exchange that Cindy Anthony knows SB will be skeptical of her information, in particular given that Casey Anthony had already been charged with murder. In the exchange in (4.37), she specifically compares the clues she is offering to pieces of a jigsaw puzzle.   (4.37) CA: ...  But I think the clue here is a seasonal. And I know you guys don’t look  on clues or whatever.  You’re not supposed to be putting a jigsaw puzzle piece together.  But sometimes that’s what we have to do hen [sic] we’re trying to figure stuff out… ((13) CA3.29:9-30:8)  She further attempts to persuade SB that AH was involved in Caylee’s abduction. First, she explains that, before Cindy learned about Caylee’s disappearance, Casey Anthony advised her that Zanny had been in a car accident and totaled her vehicle. After learning of Caylee’s disappearance, Cindy Anthony learned that the witness AH had been in such a car accident. She concludes that Casey Anthony had been referring to AH when she described Zanny’s accident.  (4.38) CA: … This is key because Casey had told me that when Zanny had totaled her  car in the accident that her father had bought her a new car.  And she told us it was a Ford Focus a 2008, silver Ford Focus.  I don’t know what Amy drives I  188 have no idea.  I never say [sic] Amy’s car.  But I think it’s very pertinent if Casey is indeed giving us clues as to what's going on.  And she’s trying to lead us to these people so that we can find and realize who it is so we can watch ‘em and maybe catch ‘em or we’re gonna catch ‘em off guard and get Casey home… Caylee home safely to us.  It’s all kind of makes sense.  You see where I’m going with this?  Casey said that Zanny’s car had totaled and that her father had bought her a car.  Amy in her own words told me that her car had been totaled.  The dates, I don’t know if they match up and it could’ve been something that Casey was throwing out there...  (CA3.47:20-48:6)  With this background in mind, Cindy Anthony attempts to persuade SB, in (4.39), that the information about the car and the car accident were intended to be clues about Caylee’s abductors.   (4.39) CA: New car.  Again, I don’t know the make of Jen71… of hers.  And Casey may  have picked a random car.  I’m not sure if that’s the clue or not.  That the 2008 silver, Ford Focus is actually what she’s driving or someone else is driving.  I don’t know if the actual car or the fact it was the car accident.  And her [JG: Amy’s] father is the one, is actually the clue Casey’s trying to give us… ((24) CA3.50:21-25) I argue that actually’s function as a tool for persuasion falls out from its semantics; just as with the corrective and challenging functions observed above, it allows the speaker to emphasize the reliability of their information even in the face of a direct or perceived challenge.  In Cindy Anthony’s case, her attempt to impose her convictions about Caylee’s disappearance on SB were unsuccessful; indeed, SB uses actually to both challenge her information and to persuade her that her information is unbelievable, as shown in (4.40).                                                  71 This conversation occurs in the context of CA listing all of Casey’s friends in order to determine which, if any, drive a silver 2008 Ford Focus, since that is the type of car Casey Anthony referred to as “Zanny’s”. Jen is the other woman who supposedly travelled to Tampa with Casey Anthony for work.   189 This challenge occurs later in the interview, after Cindy Anthony’s persuasion attempts given above.   (4.40) SB: I’ll tell you something right up front, Cindy.  She’s… she’s still not…  she… if she had time to go to work, if she had time to see friends…  CA: But I don’t think she was working during this time. SB: I understand that, but we’re talking about a week and a half two weeks period of time where her time was completely occupied.  Is that what we’re supposed to believe at…  CA: She…  SB: … this point she hasn’t… CA: She’s told me during that time.  I mean from my perception not knowing Caylee’s missing.  SB: Uh-huh (affirmative) CA: She was always doing something with work or visiting with her friends.  And the core friends she’s visiting… SB: But now she’s telling you from… from where she’s… the conversation she had now, think back mom what I was telling you, I was trying to give you clues.  CA: Right. SB: Or is… does that mean that you’re supposed to believe now that she was actually in a position where she had no contact with anybody except by phone?  They would let her talk on the phone as much as she wanted?  CA: I don’t know.  SB: But she couldn’t…  CA: I don’t know, I cannot…  SB: … get out and talk to anybody…  CA: I couldn’t elaborate.  SB: …by herself? ((64) CA3.177:3-178:2)  Speakers’ ability to use actually to attempt to persuade their discourse partners that their information is correct is consistent with the discourse function of Wanka Quechua -mi observed in Floyd (1999). Recall the data from (1.5), repeated as (4.41), where the speaker is responding to allegations that he fathered a child.   (4.41) Wanka Quechua  mana-m chay ya’a-pa-chu   not-DIR.EV that 1-GEN-NEG  190  “That (the child) is not mine.”  (Floyd, 1999: 73, cited in Aikhenvald, 2004: 318) The direct evidential marker is unexpected in (4.41) because there is no suggestion the speaker witnessed the event itself. Rather, Aikhenvald described the speaker’s goal in (4.41) as an attempt to “impose his conviction on others” (2004: 318; see also, Mushin, 2013). That is, like actually, Wanka Quechua -mi can be used as a tool of persuasion in addition to marking evidential information.   That Wanka Quechua and English speakers can use -mi and actually, respectively, for the purposes of persuasion is consistent with their analyses as markers of direct evidentality (for -mi) 72 or best evidence (for actually). The speaker can use -mi or actually to highlight that their information for p is direct or best, respectively, and should therefore be preferred to any information that tends to contradict it. This function is observed in response to a challenge about the reliability of the speaker’s evidence or where the speaker believes their report will be met with incredulity. 4.1.4 Negotiating the common ground, summary   I have shown above that speakers use actually to correct and challenge information proffered for inclusion in the common ground, and to defend their information against such a challenge. I argue that these functions fall out from the semantic analysis presented in the previous chapters; specifically, by marking that the speaker has direct evidence for the proposition they are making, actually allows the speaker to foreground the reliability of their evidence in the face of contrary evidence. Thus, the speaker is able to mitigate the effect of                                                  72 For the purposes of this discussion, I assume that Wanka Quechua -mi marks direct evidence rather than best evidence because that is how Floyd (1999) and Aikhenvald (2004) analyse it. I leave to future research the question of whether it is more properly characterized as a marker of evidence strength, like Faller (2002) argues for -mi in Cuzco Quechua.   191 challenging the addressee’s information by indicating that the speaker has good reason to believe what they are saying without directly expressing skepticism. As such, actually can be used to achieve delicate interactional goals while maintaining expected politeness norms. This is particularly important where the speaker has limited institutional authority in the discourse.73 In the next section, I show that English speakers also use actually to control the discourse topic.  4.2 Actually and topic control In addition to the discourse functions described above, speakers use actually to control the discourse topic; to introduce or change the topic, to resist topic change, and to begin a discourse-relevant digression. More generally, actually can be used to signal that the speaker is “initiating a distinct move, which deflects the discourse from the path it was following” (Taglicht, 2001: 10) or that it suggests “topic movement” (Clift, 1999, 2001; see also, Oh, 2000). To this end, actually functions like the Korean direct evidential -telako as observed in Kim (2006). However, actually’s topic control function is sensitive to its syntactic position; only parenthetical actually can mark that the speaker has direct evidence that what they are about to say is relevant to the discourse and that their contribution is felicitous.  Just as the topic control function is limited syntactically to parenthetical actually, it is also limited in terms of which discourse participants can make robust use of it in police interviews. I previously noted that witnesses have limited institutional authority, and this extends to their ability to control the discourse topic. Heydon (2005) observes that, in police                                                  73 Recall from chapter 2 that the majority of actually tokens (69%) were produced by witnesses.   192 interviews, the interviewer has exclusive authority to dictate the discourse topic. Indeed, the police interview data show that, while both interviewers and witnesses use actually for topic management, witnesses’ use of actually to this end is limited by their institutional role. Only interviewers are observed to use actually to introduce and change topics, while both interviewers and witnesses use actually to resist topic change and to mark that they are about to embark on a relevant digression.   4.2.1 Introduce a new topic  Speakers use actually to introduce a new topic into the field of discourse. Indeed, the police interview data show that actually is regularly used at the beginning of the information-gathering phase of the interview; i.e., when the interviewer first begins to question the witness about their evidence. Heydon (2004, 2005) showed that police interviews are organized in a tripartite framework comprised of the Opening, Information Gathering, and Closing (2005: 73). Interviewers regularly use actually to mark the transition from one phase of the interview to another, as exemplified in (4.42), where the interviewer, YM, is explaining to the witness, SP, the purpose of their visit. The interview with SB is one of the few that occurred in the witness’s home as opposed to in a formal interview room.   (4.42) SP: Find a seat and…   YM: Well thank you.  SP: … this is the best I can do for you.  SB: That’s not a problem.  JA: Alright, we’re good. YM: That’s quite alright.  Actually, we’re, we’re here, we’re hoping just to take a bit of your time and kind of, I, I’ve been meaning to come up here and talk.  Here, have a seat.  Have a seat. ((559) SP2.4:3-10)  193 YM uses actually to transition from general conversation to the Opening Phase of the interview. Within the interview, interviewers use actually to signal an explicit shift in discourse topic. In (4.43), the interviewer, YM, asks a series of questions to determine which building in a particular apartment complex another witness, Sean, used to live in. He then indicates he is going to change his line of questioning by noting “actually, I’ll backtrack…;” that is, he uses actually to indicate that he is changing the discourse topic.   (4.43) YM: Where does Sean live in relation to where Tony used to live at Sutton  Place? It was the same building, right?  CC: Uhm, if you go out of our, like you know how you go into the walkway to  get to our building?  YM: Uh-hum (affirmative).  CC: Uhm, you’re, actually do you know where the volleyball court is?  YM: Yes.  CC: It’s that building right there that like if you look out his uh, balcony you  see the back lake, and then to the left there’s the pool and the sand, uh, volleyball court.  YM: Okay. So it’s not the same building?  CC: It’s not. Uhm, I don’t know the building number offhand, but it’s uh…  YM: If I come out, if I come out of Sutton Place towards that parking lot where  you used to live, go down the stairs and there’s a parking lot…  CC: Yeah (affirmative)  YM: … which building was it?  CC: You just go straight.  YM: Uh-hum (affirmative).  CC: And then you dead end and it’s that one right there on the right. YM: The one there on the right? Okay.  And you, you mentioned something  about uhm, actually I’ll backtrack.  While Casey was staying at the apartment, and while you… ((215) CCa6:18-14) YM’s use of actually in (4.43) could also be analysed as self-correction as discussed in section 4.1.3.1. That is, the actually could just as easily mark that YM is correcting his earlier question (you mentioned something about…) by indicating that he intends to ask about something that happened while Casey was staying at the apartment. I argue that YM’s use of  194 actually in (4.43) and the examples in section 4.1.3.1 differ with respect to what is being corrected. In section 4.1.3.1, the speaker corrects information they previously gave, while in (4.43) and (4.44) they correct the choice of topic they intend to pursue. In (4.44), the interviewer, CE, notes for the record that the witness, TM, provided a number of photocopied documents and that they would go through those documents throughout the course of the interview. He then changes his mind and indicates, using actually, that he intends to go through the forms immediately.   (4.44) CE: Okay.  Alright, first I’d like to go through uhm, we had met with you and  your attorney Mr. NeJame there in Orlando a couple of times regarding this paperwork.  Ah, looking into who may have been searching that area of Suburban Drive.  And we were able to find some photocopies that you were able to provide us so we could start following up on who was in that area and we appreciate that.  Ah, today with your cooperation, we were able to actually find some of the missing documentation.  Um, which is good and we’ll get to that here in a short period of time.  Um, I’ve done a property form and we’re gonna take approximately eleven articles from here today, is that your understanding?  TM: Ah, yeah that’s what I understand.  CE: Okay.  Super and I’ll provide you with those forms?  TM: Right.  CE: Excellent. We can go through those during the course of the interview.  Actually, we’ll go through ‘em right now… ((494) TM2:1-15) Although likely related to the self-correction function discussed in section 4.1.3.1, the data in (4.42) through (4.44) show that interviewers use clause-initial actually to introduce a discourse topic for discussion.  4.2.2  Resist topic shift   Because the witness has limited institutional control over the discourse topic, they generally are unable to introduce a new topic. However, where the interviewer moves off a  195 discourse topic before the witness has given all of the information they believe is relevant, they can use actually to maintain the current topic until their information is given in full. In (4.45), for example, the interviewer, YM, is canvassing with the witness, CS, certain comments from Casey Anthony that she had intended to move out of her parents’ house.   (4.45) YM: Okay, so June 17th she started saying about her and her mom moving to a  different house (inaudible)?  CS: No, just her and Caylee.  YM: Oh, okay. CS: And but the mom would come and look at it with them.  Which I found  out was not true at all.  They had never even looked at it.  YM: You ever meet Amy Huizenga, her friend?  CS: Uhm, yes.  YM: She ever say anything about moving in with Amy, or? CS: No. Uhm, I actually, one more thing, I just went over to Casey’s house with her parents and her friends… ((610) CS17:24-18:9) CS’s evidence is that Casey Anthony had intended to move out with Caylee, but that her mother, Cindy, had gone out to look at potential houses with her. YM asks CS whether CS had heard Casey Anthony mention that she intended to move in with her friend, Amy, but CS had not finished giving his evidence about Cindy Anthony’s involvement in the housing search. He uses actually to resist the topic shift until he can provide all of his information before moving on to the next discourse topic.  Similarly, in (4.46), CA uses actually to resist a topic shift away from the issue of whether or not she took Caylee to the beach at a particular time. CA uses actually to maintain the topic to explain that she learned through conversations with Casey Anthony’s friends that they believed Caylee was at the beach with CA, even though that was not true.   (4.46) CA: I never took Caylee to the beach.  EE: Okay yea well I don’t want to appear like I’m trying to (Over talking…)  196 CA: No, no, no that was something because um… actually I had spoken with Ryan on the 3rd and he thought that I had Caylee at the beach that week and the week that I was off and I said no I haven’t been to the beach all week and I haven’t seen Caylee.  ((84) CA1.15:15-20)  Even though they are limited in terms of their authority to introduce or change topics in police interviews, witnesses are able to maintain a current topic (or, in other words, resist a change in topic) where they have not finished proffering their information; they can use actually to accomplish this.   4.2.3  Discourse relevant digressions  Similarly, where a speaker is about to embark on a digression that might not be obviously relevant to the discourse topic, the police interview data show they can introduce their digression with actually. This is particularly true for witnesses, who have little independent institutional authority to dictate the discourse topic. Taglicht observed that actually can be used to show that a speaker is “‘going off on a tangent’, that a thought has just come into his head which changes the direction of the discourse, at least momentarily” (2001: 9). I argue that, while actually marks the beginning of a tangent, only speakers with institutional power can use it to introduce a new discourse topic entirely (as shown above); speakers with less institutional authority use actually to mark not only the fact of the digression, but that its content is relevant to the topic under discussion.  For example, in (4.47), the interviewer, YM, asks the witness, AL, if he recalls being in Casey Anthony’s car during a particular time frame. AL begins to answer, then digresses to talk about a conversation he had with his promoter, introducing his digression with actually.    197 (4.47) YM: Now you’ve never been in, and do you recall being in her car in that time  frame?  EE: Let me go back to my calendar.  AL: I think so. YM: Hmm?  AL: I think so.  There was a day, I there was a day that actually, uhm, I just actually had a conversation yesterday with uhm, with one of the prom… one of my promoters.  YM: Uh-hum (affirmative) AL: We were, uh, well uh well, it was around the same time that I was also looking for a house.   YM: Uh-hum (affirmative) AL: Or for an apartment or whatever.  The two dudes that I work with were also looking for a place to stay.  So I’m pretty sure that there was one day when I was meeting with them and she tagged along.  Uhm, and she met me I think at the house, at their apartment.  And then we left with her car I think and my friend’s car.  And there was a time when I was in the car and there was a time when I was in his car. ((436) AL3.19:24-20:16) AL’s information about having a conversation with his promoter is not obviously relevant to the question posed to him; namely, whether he recalls being in Casey Anthony’s car. However, as he explains, the digression is relevant because it explains that, during the time in question, he and two colleagues were looking for a place to live and he rode in Casey Anthony’s car when they went to look at a potential apartment. His conversation with his promoter reminded him of that information. Thus, the information he gave, while appearing to be a digression, was nonetheless relevant to the question posed.   Similarly, in (4.48), the witness, CA, is challenging the interviewer, SB, about the investigators’ efforts in following up on crime-line tips. At this point in the investigation, the investigators have charged Casey Anthony with Caylee’s murder, but CA is insistent that the investigators should continue to search for Caylee as a missing child. CA is specifically confronting SB about the investigators’ failure to take a particular tipster seriously. In the  198 course of the confrontation, she indicates that the tipster has involved her 14 year old son, introducing that information with actually.  (4.48) CA: I just…  SB: … understand that. CA: … I hope they are doing everything they can because she um told me that right after that person did her… the sketch.  Um, actually she’s involved her 14 year old son in this too, you’re well aware of that.  Her 14 year old son..  SB: Right. CA: … was also brought to the McDonald’s, statement taken.  I mean most mothers wouldn’t bring in their children into something like this unless there was… they would have pretty you know definite feelings about it. I know I certainly wouldn’t involve my kids if… on a whim on something of this magnitude. ((9) CA3.11:14-23) The tipster’s decision to involve her son in the investigation is not obviously relevant to the topic under discussion; namely, whether the investigators are valuing her information appropriately. However, it becomes clear that CA made the invocation in order to bolster her position that the investigators were wrong not to take the tipster seriously.   In (4.49), the interviewer, EE, is asking the witness, WW, about certain comments attributed to him by a person named Ryan Huffman. WW explains that he does not know anyone named Ryan Huffman, but then backtracks to indicate he knows a friend of a friend named Ryan, but he does not know Ryan’s last name. EE then asks whether WW may have made the alleged comments to the mutual friend, to which WW responds that “actually, I heard Ryan got arrested.”   (4.49) EE: Crimeline tip.  Winter Park arrested a kid on 7/26 I guess that this call  came in, which would be last, it was last weekend.  It’s last Saturday.  This Ryan Huffman.  “Suspect states Will Waters, Casey’s supposed boyfriend, mentioned in casual conversation that Casey had been in an accident and brought her car in for repair.  Afterwards unknown mechanic, unknown garage, stated the car smelled like a dead body.  No further details.”  Ring any bells?  199  WW: Who said…  EE: Any conversations like that where?  WW: No, never.  I’ve never heard anything about a dead body.  EE: Okay.  WW: I don’t know a Ryan Huffman.  EE: Alright.  WW: Wait.  Is Ryan, have you met Ryan? Is he a tall guy?  EE: (Inaudible). WW: Because my buddy Dusty uhm, my buddy Dusty has a Ry… a buddy named Ryan but I don’t know his last name.  EE: Any, any social circle conversations where you may have been talking about the case and maybe it was misunderstood?  WW: (No verbal response)  EE: You buddy Dusty has a friend named Ryan? WW: Yeah (affirmative).  Actually I heard, I heard (making tapping sounds) Ryan got arrested.  Do you want me to, can I call Dusty and ask him what uh, Ryan’s last name or? ((612) WW16:1-23) Ryan’s arrest is not obviously relevant to the question under discussion, except insofar as it explains why WW does not believe he could have had an opportunity to speak with Ryan about the Casey Anthony case.   The police interview data show that speakers, and in particular witnesses, use actually to introduce a digression that is not obviously relevant in order to exert some control over the discourse. In these circumstances, actually marks that their speech act is relevant to the greater discourse.    4.2.4 Evidentials and topic control cross-linguistically   I have shown about that English speakers use actually to control the discourse topic, by introducing or changing topics, resisting topic change, and marking relevant digressions. At a discourse level, I argue that actually signals the speaker has reliable evidence that their contribution is relevant and felicitous at that point in the discourse. In the police interview transcripts, the interviewer is the institutionally powerful discourse participant and can use  200 their institutional authority to direct the topic to whatever they determine is relevant. Although their institutional authority is limited, witnesses are similarly able to use actually to control the topic (in particular, to resist a topic shift or to embark on a discourse-relevant digression), where it is felicitous to do so at that point in the discourse.  By using actually to control the discourse topic, English speakers use actually to perform the same type of discourse management functions as speakers of Korean and Mandarin. Kim (2006) observes that Korean speakers use ] -tela not only to indicate the source of their information, but to mark discourse transitions, such a digression or side-track in the on-going discourse and interaction. She gives the data in (4.50) as an example. In (4.50), two friends are discussing the fact that N’s child is on the waitlist for a private school. In line 13, N uses -tela to mark that she is digressing from the topic directly up for discussion.   (4.50) Korean  01 S: icey  twulccay-twu  hakkyo ka-se com iss-ta    Now second-also school  go-and a:little exist-TRANS     o-keyss-ney?    come-DCT:RE-FR   “(Your) second (child) must be also coming home soon from school?”  02 N: kulenikka kyay  kyay  han il nyen tongan   so  that:child that:child about one year during  03  ilpan  hakkyo-ey cip-e  neh-ki-lo hay-ss-e   Japanese school-to put-CONN put-NML-to do-PST-IE   “So, (we) decided to put him, him in a Japanese school for about a year.”  04 S: uh kulay-ss-e?    be:so-PST-IE   “Uh, was that so?”  05 N: awu ya? yay-nun private school-i waiting list-ya  201   wow hey this:child-TOP            -NOM        -be:IE     cikum-kkaci-twu    now-until-also   “Wow, hey, he is still in the waiting list for the private school.”  06 S: kulay?   be:so:IE   “Is that so?”  07 N: waynyamyen yay-nun nai-ka  eli-ki  ttaymwuney   because this:child-TOP age-NOM young-NML because  08  public-ul mos cip-e  neh-canh-ni:¿             -ACC not put-CONN put-you:know-INTERROG   “(We) cannot put him in public (schools) because he is too young, you  know?”  09 S: kule-chi   be:so-COMM   “(That’s) right.”  10 N: kunikka kuntey waiting-   so  but    “So, but the waiting-“   11  kunikka kyoyukyel-i  kangha-n ile-n    so  education:fervor-NOM strong-ATTR like:this-ATTR     ciyek-eyse-nun,    area-in-TOP   “So, in this kind of area with strong education fervor,”   12 S: uh   “Uh”  13 N: kulen  ccok-ulo mwe-l  ha-nun  key   such direction-to what-ACC do-ATTR  thing     nas-keyss-te-la   [hhh    better-DCT:R-RT-INTROS   “[I noticed] that it would be better to do something in such fields (related  to children)”  14 S:      [e: kule-chi  202        yeah be:so-COMM        “Yeah, (that’s) right”  15 N: e mwe soakwa-twu,   yeah what pediatrics-also,   “Yeah, also for pediatrics”  16 S: [uh   “Uh”  17 N: [ahwu:: soakwa-ka  way ilekhey   wow  pediatrics-NOM why like:this     pakulpakalha-ni    crowded-interr   18  cangnankam-to thwusengi-kwu   toy-also  full:of-CONN   “Wow, why is the pediatrics office so crowded like this? (It’s full of toys,  and,”  19 S: uh kulay?   uh be:so:IE   “Uh, is that so?”    (N’s children are making noise in the background)  20 S: ay-tul-i-ya cikum yeph-eyse mwa-la-kwu kule-nun ke?   kid-PL-be-IE now side-at  what-DC-QT say-ATTR thing   “Are they (your) kids who are talking next to (you) right now?”  21 N: e.   “Yes.”  In addition to marking that N has direct information, -tela marks that N is digressing from the discourse topic. Thus, its use is dependent not only on the speaker’s source of information, but also the speaker’s interactional goal in the circumstances.   Wang et al. (2011) show that Mandarin speakers also use both qishi (‘actually’) and shishishang (‘in fact’) to control discourse. Specifically, they show that qishi and shishishang function to mark a boundary in discourse to indicate a partial shift in topic (at p. 243). While  203 they do not analyse qishi and shishishang as evidentials, it is nonetheless interesting that they appear to function similarly to actually at a discourse level. The data in (4.51) occurred in the context of a television debate. The main topic in the talk is the Democratic Progressive Party’s abolition of the name of a memorial hall. A, the program host, asks B to comment first. B does so, and then shifts the topic initiated by keshi (‘but’) to talk about the bribery of the DPP. C, another invited guest, then uses qishi to shift the topic further to give his opinions on justice in democracy (Wang, et al., 2011: 249).  (4.51) B: …qu hanwei   huozhe  sibao   zhe   jiu   quanwei  shidai       go defend   or     stick to ASP old  authority age    de  paibian, dangran bu yinggai.\   ASSOC tablet   of course NEG should   “Defending the old order of the authoritarian past is of course not a good  thing.”   A: ..en.__     DM   “Um.”   B: keshi,_ weile zijide guanwei,/   But for self official.position    qu dang tanfu jituan de  zougou,__   go act as corrupt group ASSOC stooge     …na  <H geng bu yinggai H>,__      that  more NEG should    … na shi xialiu beibi.\        that cop nasty mean   “But acting as a stooge for currently-operating corrupt agencies is worse yet –  it’s low-down and dirty.”   C: …qishi oh,__       actually prt    … sheme Jiaozuo minzhu?__  204        what called   democracy    … minzhu limian you ji ge  hen   zhongyao de dongxi,__        democracy inside have several CL  very important ASSOC things   “Actually, what exactly is democracy? There are a number of important  elements…” (adapted from Wang et al., 2011: (14)) Thus, qishi functions like actually in English and like -tela in Korean insofar as it can be used to shift discourse topics and, as a result, exercise control over the discourse.   4.2.5 Actually and topic control; the problem for a unified analysis  I have shown that the English speakers whose speech is recorded in the police interview transcripts use actually to perform two main discourse functions. As demonstrated, speakers use actually to negotiate what should and should not be included in the common ground. For this discourse purpose, the unified analysis of actually in (3.36), repeated as (4.52) below, applies equally to parenthetical actually and clause-internal actually.  (4.52) i.   [[Actually φ]]c,w is felicitous iff the speaker in c has best evidence for [[φ]]c,w, in c.  ii.  If [[actually φ]] c,w is felicitous in c, then [[Actually φ]] c,w = 1 iff [[φ]] c,w = 1.  Where the unified analysis breaks down is with respect to actually’s topic control function. From a topic control perspective, speakers use actually to control the discourse topic. Speakers with greater institutional authority are able to make more robust use of actually as a marker of topic control, though witnesses are able to use it to resist topic shift and to mark discourse-relevant digressions. While both parenthetical actually and clause-internal actually can be used equally to negotiate the common ground, the topic control function is sensitive to syntactic position and is only available with parenthetical actually. Rather than tweak the semantic analysis to reflect the fact that the topic control function is only available with parenthetical actually, I appeal to English syntax to explain this  205 sensitivity. In this section, I set out some suggestions about how its syntactic position may permit parenthetical actually (and not sentence-internal actually) to mark topic control. I suggest that the fact that only parenthetical actually is felicitous with speech acts is likely due to its syntactic position. This is just a first pass; I leave a fully fleshed out syntactic analysis to future research.   I assume that sentence-internal actually is adjoined to the VP as in (4.53). For parenthetical actually, I follow Dehé & Wichmann (2010, 2011) and assume that it is adjoined to the clause as in (4.54a) and (b).   (4.53)  CP[you actually VP[told me your maiden name]]  (4.54) a. CP[actually CP[ you told me your maiden name]]  b. CP[CP[you told me your maiden name] actually] The proposition in (4.53) is felicitous where the speaker has best evidence for the prejacent (you told me your maiden name). In (4.54), both (a) and (b) are felicitous either where the speaker has best evidence for the prejacent or where the speaker has best evidence that the speech act of asserting the proposition is felicitous at that point in the discourse.    Because it is adjoined to the clause, one possible analysis is that parenthetical actually is a candidate to occupy syntactic space in the ‘grounding layer’ of Wiltschko’s (2014) universal spine. The universal spine promotes a hierarchically organized series of core functional projections, the including the grounding layer.   (4.55) Universal Spine Hypothesis    RespP                        GroundP    206                                       S    (from Wiltschko & Heim (to appear): (34)) For Wiltschko (2014) and Wiltschko & Heim (to appear), the S node in (4.55) represents the traditional clausal architecture. Above that, however, is the grounding layer relating to the speaker’s attitude towards the proposition. The responding layer is dedicated to letting the addressee know what the speaker wants them to do with the utterance (Wiltschko & Heim (to appear): 16) Wiltschko & Heim (to appear) argue that confirmationals like English eh target the grounding layer of the hierarchical structure. Although it remains to be investigated, parenthetical actually may similarly occupy the grounding layer and mark that the speaker has best evidence that the speech act of asserting the proposition is felicitous at the time it is made.   4.3 Summary of actually’s pragmatic function  In this chapter, I have shown that speakers use actually for two main discourse purposes: to negotiate the common ground and to control the discourse topic. While bare assertions will generally suffice, speakers use actually to emphasize their evidence where discourse contexts require them to do so. I have shown that, in the police interview transcripts, four such discourse contexts are attested: where the hearer fails to respond appropriately to the speaker’s information when it was initially presented in a bare assertion, for self-correction, to challenge the information of others, and for the rhetorical purpose of persuasion. Both clause-internal and parenthetical actually are shown to perform this  207 function, which includes the well-documented contrastive reading regularly ascribed to actually in the literature. I argue that the emphasis function (including the contrastive function) falls out from the semantic analysis of actually set out in section 3.2. By foregrounding the strength of their own evidence in the face of contrary information, the speaker is able to emphasize that they are ‘right’ without explicitly insisting their discourse partner is ‘wrong.’  The police interview transcripts also show that speakers use parenthetical actually to control the discourse topic. That is, speakers use parenthetical actually to introduce or change topics, to resist topic change, and to embark on discourse-relevant digressions. Because of the institutional power imbalance between discourse participants in police interviews, interviewers were shown to make more robust use of this discourse control function, particularly with respect to introducing and changing topics, although witnesses were observed to use actually to resist such topic change where they had not finished giving their evidence on a particular issue. Finally, I showed that the topic control function is only available with parenthetical actually and propose some first steps towards a syntactic analysis which might account for this distinction.      208 “I can’t explain something that I don’t know” LB (AP/LB2.10:23)      5. THE PRAGMATICS OF APPARENTLY AND SUPPOSEDLY   I previously observed that the goal of witness interviews is to obtain accurate, reliable information on the basis of which the interviewer can advance the investigation. Thus, the discourse goal of the cooperative witness is to ensure that only accurate, reliable information is accepted in the common ground, since it is that information the investigator will subsequently rely on in the course of the investigation.74 In chapter 4, I argued that witnesses can use actually to mark when they have best evidence for their proposition and, by doing so, advocate for its inclusion in the common ground. Witnesses may also have information which may be helpful to the investigation but for which they are either uncertain or skeptical about reliability. In order to achieve the discourse goal of providing accurate information, the witness must proffer such information in a way that makes explicit the fact that they are not committed to the veracity or reliability of the prejacent and are not proposing it for inclusion in the common ground. In other words, just as actually permits witnesses to align themselves with the principal/author role with respect to certain information, witnesses must also be able to express that they are merely the animator of the information. In this chapter, I argue that English speakers can use the indirect evidentials apparently and supposedly to do so.  In section 3.3, I argued that apparently marks that the speaker’s evidence for the prejacent is indirect on the evidence type dimension of meaning, while in section 3.4, I                                                  74 For various reasons, witnesses may attempt to deceive the interviewer and deliberately provide inaccurate or unreliable information. For the reasons discussed in chapter 1, I leave issues of deceptive language to other research.   209 showed supposedly is a dedicated reportative that marks the speaker’s information as ‘not best’. In this chapter, I examine the pragmatic effect witnesses achieve by using apparently and supposedly. I argue that speakers use apparently and supposedly to introduce information that they believe is important to the investigation but to which they cannot or will not commit because the information requires further investigation or because the information is known to be untrustworthy. It may be important to establish what the witness was told even if the content of that communication was untrue and witnesses use apparently and supposedly to do this. I argue that these discourse uses fall out from the semantic analyses of apparently and supposedly proposed in sections 3.3 and 3.4.  5.1 The pragmatic use of apparently   In section 3.3, I argued that apparently makes a contribution on the evidence type dimension of meaning and, specifically, that the speaker has indirect evidence for their proposition. I further argued that apparently differs from must in two crucial respects: first, and most importantly, apparently does not entail the truth of the prejacent and second, apparently can be used as a reportative.75 I argue that, given its semantics, speakers are able to use apparently to achieve the interactional goal of distancing the speaker from the information they are nonetheless compelled to proffer during the investigation. Apparently, therefore, allows the speaker to share information without indicating a concomitant belief in its accuracy or reliability. That is, apparently allows the witness to share information without proposing the propositional content for inclusion in the common ground.                                                   75 As shown in chapter 3, must p is only felicitous as a reportative where the information reported does not directly settle p and the speaker still requires some reasoning in order to settle p.   210  In (5.1), YM is interviewing Casey Anthony’s mother’s supervisor, DP. DP describes Cindy Anthony’s behaviour up until the day they found the car. The car in question was Casey Anthony’s car, which had been impounded from a local parking lot while she was supposed to have been out of town. YM asks a follow-up question about that day at which point DP makes it clear that she was not present when Cindy Anthony learned the car had been found. DP nonetheless goes on to provide the information she has, that Cindy Anthony learned about the car during a telephone call with George Anthony, introduced by apparently.   (5.1) DP: Um, but she… I always thought there was something underlying it ‘cause  she kept saying, well she’s trying to bond with the baby and I think she feels I’m too close to the baby and she’s trying to get her away from me.  So I thought well, maybe so you know.  So, that went on until the day they found the car.  YM: Tell me about that. DP: Hm, I was in my office which is in the back, she sat here.  Apparently she got a call from George that there was a letter from the impound lot that the car… that Casey’s car was in the lot.  I guess she started screaming on the phone to George and swearing and everything.  So…  YM: Screaming at George? DP: Well like I don’t know this is just what they told me.  She was upset at George and saying, well go get the God damn car, about that kind of stuff.  YM: Okay. ((A54) DePo6:4-16) By using apparently to proffer her information about Cindy Anthony’s telephone call, DP marks not only that she knows the information indirectly through hearsay but also that she makes no commitment to its reliability. When YM asks her a follow-up question (Screaming at George?) she reiterates that she is only repeating what she was told.   In fact, the proposition DP introduced with apparently was true; both Cindy Anthony and George Anthony independently described the incident and confirmed that DP’s information was accurate, as shown in (5.2).   211  (5.2) a. CA: … the 15th George goes and picks up the letter from the post office.  The  registered letter. I get a call from him somewhere between 12 and 12:30 somewhere around there. Saying, guess what that registered letter’s for and I said what? He said, it’s a tow company. And I said what do you mean, he said, Casey’s car’s been at this Johnson’s Tow Company since… And I’m going, oh my God!... (CA3.196:23-197:3)          b. GA: …Well as I open up the letter it said that the vehicle was towed on June  30th, 2008 from the corner of Chickasaw… or not from… from the corner of Goldenrod and …  SB: Colonial.  GA: … Colonial Drive. Nine fifty-eight I believe was the time it was towed. I  don’t know if that was AM or PM ‘cause it wasn’t on there it wasn’t… we still have …  SB: It just said 9:58? GA: Nine fifty-eight. I immediately called my wife I said, guess what? Casey can’t be in Jacksonville.  She says, why? And I said, ‘cause you know that letter that certified? I got in front of me I said, the car was towed by Johnson’s Towing and it’s down on Narcoossee Road.  My wife said, you’re kidding I said, I am not kidding. I said listen I’m gonna drive down there and find out what information I can find out. But you know I’ll call you back as quick as I can. ‘Cause now she’s a little like, what the hells going on?  (GA4.65:16-66:4)  As I argued in section 3.3, unlike must, apparently does not entail the truth of the prejacent. If p is false, must p is false as well (because must p entails p). Apparently p, on the other hand, can be true regardless of whether p is true or false. In (5.1), the prejacent happened to be true. In (5.3), however, it is not. In (5.3), Cindy Anthony’s colleague, DB, proffers that Casey Anthony had worked at Sports Authority, which was untrue.     (5.3) DB: Um, the daughter was apparently working at Sports Authority for a short  time there.  Apparently the one over at Waterford Lakes and she apparently was pulled to one I think on OBT or Orange Avenue or something.  YM: And this is according to?  DB: Cindy.  YM: Okay.  212 DB: Yeah. Yeah. Well the daughter came in one time and said she was going to work to Sports Authority or so she stated. ((A11) DB5:17-24)  It becomes clear through the course of the investigation that Casey Anthony never worked at Sports Authority. However, it is irrelevant whether DB knew or did not know that the prejacent was false at the time she made the utterance in (5.3); (5.3) is felicitous regardless of whether DB knew that Casey Anthony never worked at Sports Authority because, unlike must, apparently does not entail the prejacent. Therefore, by using apparently, the speaker proffers the information without making any commitment whatsoever to the veracity of the prejacent.76   The ability to distance oneself from the reliability of the information is particularly important when the point of the witness’s information is that somethi