Conversationalization and democratization in a radio chat show: A grammar-led investigation

Nicholas Smith

University of Leicester, UK

This paper investigates the closely-related concepts of conversationalization and democratization in a specialized register, the biographical radio chat show, represented by BBC Radio 4’s (DID). To explore these concepts in the show we first undertake a data-driven stylistic analysis of ‘key’ parts-of-speech (POS) tags, i.e. statistically significant grammatical categories that distinguish a corpus of DID talk from a corpus of conversation. We then track these grammatical features over four sampling periods to see what changes occur, and the extent to which they evidence conversationalization and democratization. This process is done separately for guests and hosts, since they each have a different role in the participation framework of the show. Results suggest a clear overall movement across time towards conversational norms and levelling of the differences between guests and hosts. The features that evidence conversationalization are usually but not necessarily evidence of democratization, and vice versa. We relate the findings to the changing contextual environment of the show.

Keywords: conversationalization; democratization; Desert Island Discs; participation framework; key POS-tags

1. Introduction

The terms democratization and conversationalization have been used in previous research to refer to broad processes of change in public forms of discourse, particularly those used in contemporary western societies (Fairclough, 1995, 2014; Steen, 2003; Leech et al., 2009). It is clear from the way these terms have been applied that they are assumed to share a lot of common ground. However, it is less clear in what ways (if at all) they differ: Can a field of discourse become progressively more conversational without becoming more democratic? Or more democratic but not more conversational? Such issues have been partially touched on before, but not apparently explored in systematically sampled data across time.

1 This paper seeks to address this research gap. The data we use for the investigation are from broadcast talk, specifically interview transcriptions from the popular British radio show Desert Island Discs (DID), which has been running almost continuously on BBC since 1942. The basic format of DID has changed remarkably little over the years: essentially a host invites a guest, a prominent public figure, to imagine being marooned on a desert island, to choose eight pieces of music to accompany them, and to talk about their lives in relation to the music. In previous research DID has sometimes been classed generically as a chat show (Castell, 1999; Smith and Waters, 2018), although music also plays an important role (Brown et al., 2017). The programme is interesting from the perspective of style change because a) the turnover of hosts has been very low, and b) during the 1970s and 1980s its hosting station, BBC Radio 4, was the site of conflicting pressures over issues such as emotional disclosure on programmes, and popularization (‘dumbing down’, in some cases) vs. seriousness (Hendy, 2007, 2017). The corpus data used here come from a crucial period in these struggles, and this may have influenced the course of democratization/ conversationalization.1

We will explore conversationalization and democratization in DID by drawing on a corpus – or representative sample – of transcriptions from the show in four periods. We will compare this corpus with a corpus of conversation to determine how far the frequencies of grammatical word classes used by hosts and guests on DID across time differ from conversation, and what evidence the comparative data provide for changes of style.

2. Overlapping notions of change, and the context of Desert Island Discs

2.1. Democratization, conversationalization and sociability

Norman Fairclough, perhaps the person most strongly associated with democratization and conversationalization of discourse, defines the first of these terms as “the removal of inequalities and asymmetries in the discursive and linguistic rights, obligations and prestige of groups of people” (Fairclough, 1992: 201). Fairclough (p.201-207) identifies a number of ways in which linguistic democratization can be realized, including: • democratization of access, e.g. accepting people with regional accents for public speaking

roles in the media;2

1 As Doctor (2017) notes, the very introduction of programmes like DID during the 1940s was itself a step towards ‘democracy’ in British broadcasting. Up to that point music selections on the radio were controlled by the broadcasters themselves. 2 This aspect is similar to Coupland’ (2014) vernacularization: “shifts towards a more positive valorization of vernacularity” (p.85). Although diversification of accents is a salient area of the democratization of the BBC, it is beyond the (grammatical and stylistic) scope of the present paper.

2 • the removal of ‘markers of hierarchy and power asymmetry’ (1992: 203), e.g. terms of address such as Mr and Sir; • removal of sexist language, e.g. he as a generic third person; • conversationalization (see below).

Fairclough argues that democratization is often a convenient pretence, a tool for social control. For example, when a government switches to using non-hierarchical address terms in letters to its citizens, it can be more about making people think that society has become more egalitarian and democratic than actually making it so. For this reason it might be more appropriate to rename Fairclough’s concept pseudo-democratization, as Clayman and Heritage (2002: 339) do. Media researchers sometimes regard Fairclough’s interpretations of discourse change as unnecessarily ‘suspicious’. Scannell (1989), Tolson (2006) and Coupland (2014), for instance, treat the expansion of more personable, informal communicative styles in broadcasting not in terms of ideological agendas so much as responding to a genuine need for sociability, i.e. to attract audiences by coming across as not too distant from them. Scannell (1989) characterizes the evolution of public service broadcasting in the UK as a genuinely democratizing social practice, citing as an example the introduction of interviews that bring public figures to account. Audience participation programmes arguably represent a more recent development in media democratization (see Livingstone and Lunt, 1994).

Fairclough defines conversationalization as a “tendency towards a simulation of conversation in public interactions and texts” (Fairclough, 2003: 224), that is, for linguistic features associated with informal, private conversation to permeate public forms of talk, e.g., news broadcasts and print advertisements. Such features include colloquial vocabulary and generic use of you. Conceptually, an important aspect of conversationalization is the ways in which the discourse of ordinary people’s experience, or ‘lifeworld’ discourse as Habermas (1984) calls it, spread into institutional domains such as television, radio and the print media (Fairclough, 1992, 1995). A typical case might be a news interviewer aligning her/himself with the audience by simulating words from an imaginary ‘man in the street’, e.g. “an awful lot of people when you ask them say…” (Fairclough, 1995: 145).3

There are occasional suggestions in the literature that conversationalization and democratization are not necessarily aligned. Garrett and Bell, for example, alluding to Fairclough’s work, remark that “conversationalization may or may not serve the democratization of discourse” (Garrett and Bell,

3 Fairclough (1992) sometimes uses synthetic personalization, in ways that are very similar to conversationalization. Unsurprisingly there is also overlap between conversationalization and informalization, as used by other reasearchers, e.g. Pearce (2005) on discourse, Wouters (2007) on social mores. Similarly, in corpus studies of recent language change, colloquialization (Leech et al., 2009) refers to the adoption of conversational usage in public forms of discourse, cf. Hiltunen et al. (this issue).

3 1998: 12). While the expression ‘level playing field’ is frequently used in relation to the conditions of conversation, Fairclough provides a useful reminder that informal conversation, too, can play out unequally when participants have differential power (2003: 78). And in a discussion of developments in broadcast news programmes around the turn of the 21st century, Montgomery (2007) notes that the introduction of dual news presenters can have very non-conversationlike effects. These are particularly noticeable on television news: while each presenter in turn talks directly to the camera (i.e. the viewing audience), there are times when neither presenter appears to pay attention to the other, either visually or verbally (e.g. because of the need to follow an autocue). This development clearly contrasts with ordinary conversation, where it is normal practice to monitor one’s interlocutor while speaking or listening to them.

The following observations can be made about research in this field:

1. Democratization seems more open to subjective interpretation, and therefore more difficult to firmly establish, than conversationalization. For example, how can we tell if a putative case of democratization is a pretence or real? For which parties do we consider it to apply? With broadcast interview programmes, for instance (the subject of the present study), it is unclear whether democratization should refer to the discourse relations between interviewer and interviewee, between these participants and the audience, or among all three parties. 2. Despite the multiplicity of terms and differing views on underlying motives, a fairly consistent pattern is apparent: commentators generally consider that recent style shifts in public discourse, at least in western society, have been in the direction of the more conversation-like/democratic end of the pole (whatever the underlying motivation), rather than the opposite direction. 3. There appears to be good potential for corpus methods to explore democratization and conversationalization at a textual level. We can, for instance, compare the frequencies of linguistic features in a discourse area of interest against frequencies of the same features in a corpus of conversation, to see the extent of their agreement. We can also compare talk by one kind of participant against another (e.g. guests vs. hosts). If we sample the data at different points in time, we can assess the extent of evidence of movement towards (or even counter- directional to) conversational and/or more democratic norms. This is the approach of the present paper.

4 2.2. DID and its situational context 2.2.1. Genre classification and situational differences from conversation

DID defies a simple classification in terms of register or genre. In an entire book devoted to the musicological and sociological significance of DID (Brown et al., 2017), the only attempt at labelling the show generically is ‘biographical interview’, by a former producer (Drescher, 2017). In previous work involving the present author DID was tentatively classed as a ‘biographical chat show’ (Smith and Waters, 2018, 2019), since it exhibits some of the typical characteristics of chat shows. For the present study it seems instructive to attempt to itemize salient situational characteristics (cf. Biber and Conrad, 2009) of DID that have distinguished it from everyday conversation during most of its lifetime: see Table 1 (and Smith and Waters, 2018, for further detail).

Table 1. Situational characteristics or everyday conversation and DID Characteristic Typical conversation DID Planning Unplanned, Planned, but not scripted spontaneous (also: recorded and edited) Communicative purpose Maintaining personal Entertainment through exploring the relations guest’s life & music preferences Relations between Mutually-familiar Host and guest are usually strangers participants participants Discourse control of Equal discourse rights Host has overall control of the discourse: participants see further, Section 2.2.2

It should be pointed out that this brief description masks some important changes. The show was in fact scripted until 1955 (before the period of investigation in the present study). Further, as detailed by Hendy (2007, 2017), during the 1970s and 1980s the management of BBC Radio 4, which broadcasts DID, were being pressured by changing norms in broadcasting and society more generally to a) introduce more emotional frankness into programmes, and b) become more popular in style. On the first of these points, Hendy (2017: 162) argues that a changing “centre-ground of acceptable disclosure for guests… and acceptable probing among interviewers” did not seriously affect DID until after the death of the first host in 1985. Plomley, famous for a ‘genteel’ and formulaic style of questioning (see chapters in Brown et al., 2017), was replaced by former journalists, first (1985-88), then (1988-2006) who instituted several changes: more probing kinds of questioning, playing musical choices during the actual interview to enhance the discussion (rather than editing music in afterwards), and, in Lawley’s case, avoiding pre-interview lunch with the guest, so as to increase the spontaneity of the exchange. But on the pressure to popularize the style of Radio 4, Hendy claims that “despite the many awkward attempts at popularization… the underlying trend was undoubtedly for Radio Four to become steadily tougher over time” (Hendy 2007: 395, emphasis original), i.e. to maintain a high standard of discussion, and

5 resist being ‘dumbed down’. It will be interesting to explore stylistic change in DID against this backdrop.

2.2.2. Participation framework

Goffman’s (1981) notion of participation framework is designed to capture the structural relationships between participants in a given type of interaction. The concept has been quite widely applied to media talk, and been found particularly useful in interpreting the ‘behaviour’ of personal pronouns (e.g. Chang, 2002; Jautz, 2014; O’Keeffe, 2006).

A twofold participation framework for DID is proposed in Figure 1, representing pre-allocated roles in a) the short introduction segment, and b) the main part of the show, the interview. The use of an inner and outer circle here follows Burger (2005), cited in Jautz (2014): “The inner circle comprises the conversation proper and those who directly and interactively participate in this conversation” (Burger, 2005: 20). The diagrams also draw on schematizations of participation frameworks in media talk in Chang (2002: 124) and O’Keeffe (2006: 66).

a. Introduction b. Interview

Figure 1: Proposed participation frameworks in DID, in a) the introduction and b) interview part of the programme. Note: solid-line arrow: direct address; dotted-line arrow: direction of orientation, but not direct address.

The audience in DID is a broadcast audience rather than a studio-based one, thus their members play no active verbal role in the framework. However, at all times the audience are what Goffman (1981) calls ‘ratified’ participants. Indeed, even when they are not directly addressed, that is, during the interview proper (

Figure 1b), the talk is primarily designed for them (Hutchby, 2006; O’Keeffe, 2006).

6 Although during the interview the guest and host are in direct interaction, the relationship is asymmetrical. Ordinarily only the host can initiate topics, ask questions and control turn-taking (see Hutchby, 2006; Jautz, 2014). Guests are expected to answer, typically with turns longer than those of the host. Note also that we have greyed out the arrow from guest to host, as there is evidence that the nature of address here is more diffuse than from host to guest (see Section 4.2.5). This framework is more constrained than in the closely related genre of ‘talk show’, where (particularly on television) guests can sometimes ask questions, interrupt, or initiate a turn (cf. Tolson, 2001; Loeb, 2015).

2.2.3. Potential indicators of democratization and conversationalization on DID

What might democratization and conversationalization in DID consist of? In a paper focusing on style shifts in DID mainly in relation to turn-taking structure and length, Jucker & Landert (2015) identify several changes that appear relevant:

a) an increase in the average turn length of hosts b) a greater balance between host and guests in the proportion of words spoken in the interview c) a decrease in question intonation used by hosts d) an increase in hesitation markers by hosts

The increasing balance of contribution length between the participants suggests a democratizing tendency. It might also seem a conversationalizing one, but the figures Jucker and Landert (2015: 33) give for average turn length (overall, 28.1 words per turn for hosts, 58.5 words for guests) are far higher than everyday conversation (10.85 words, cf. Rühlemann 2019: 139). Items c) and d) in the list provide clearer evidence of conversationalization, and arguably democratization too: with growing numbers of ‘questions’ no longer being framed as questions but as gentle prompts to elicit talk from the guest, the latter should have more freedom in how to respond; similarly, increasing numbers of hesitation markers by hosts is an indication of a more spontaneous production style (Koch and Oesterreicher, 2011: 54-55) as in ordinary conversation, and freedom to talk in this way can be seen as democratizing.

The present study seeks to complement Jucker and Landert’s insights by taking a more grammatical focus on style change. While the methodology is data-driven (see Section 3; also cf. Hiltunen et al. this issue, Palander-Collin and Nevala, this issue), it will be useful to see if the results can be interpreted in relation to well-known features associated with democratization/conversationalization:

• title nouns (Mr, Mrs, Dr, etc.) that indicate hierarchical relations (cf. Leech et al., 2009: 260)

7 • more and less authoritarian expressions of obligation/necessity (e.g. must, need to/want to/should (cf. Leech et al., 2009: 109-111); • sexist pronouns (generic he) and gender-neutral pronouns (e.g. singular they), cf. Paterson (2014); • features of interactive involvement, e.g. second person pronouns, discourse markers, tag questions (cf. Biber 1988, 2006, Biber et al. 1999), vague language (Carter and McCarthy, 2006); • guests having the opportunity to ask questions (cf. Tolson, 1991: 180); • in-group markers (e.g. we) that risk overreaching speakers’ authority (cf. Fairclough, 1995, 2014).

3. Data and methodology

The following research questions guide the investigation:

1. On DID how does talk by guests and hosts, respectively, differ stylistically (grammatically) from conversation? 2. What grammatical indicators of conversationalization and democratization can be detected in host talk and guest talk across time? Do conversationalization and democratization appear to be travelling on the same path?

To address these questions the primary corpus data used are four samples of transcriptions from DID taken at four periods in time:

• two periods from the first host, Roy Plomley: 1960s and early 1980s (henceforth 1980s-A)4 • two periods from the second long-serving host, Sue Lawley: late 1980s (called 1980s-C) and

the early 2000s.5

The guests in the interview samples are all English, but demographically balanced for speaker age, gender, education and occupation, to ensure a reasonably consistent social mix. In each period 20 guest speakers were selected, with a target of 1,400-1,450 words per guest.6 Table 2 presents the word

4 The 1960s corpus consists of many programme fragments. While the sound quality is good, large chunks of the original are missing, and the time offset from the original recording is unknown. For these reasons, results from the 1960s still need to be treated with more caution than those from later periods. 5 For a short period between Plomley and Lawley the host was Michael Parkinson. 6 We used Wmatrix word counts because they take better account of multiword units than standard word counts based purely on whitespace (Rayson, 2008).

8 counts by guests and hosts, while Table A1 provides figures on speakers by demographic group. For details, see Smith and Waters (2018).7

Table 2. Speakers and word counts in the DID corpus 1960s 1980s-A 1980s-C 2000s Speakers Words Speakers Words Speakers Words Speakers Words Guests 20 23,279 20 27,114 20 27,407 20 27,501 76.1% 83.3% 80.4% 72.2% Hosts 1 7,311 1 5,427 1 6,675 1 10,589 23.9% 16.7% 19.6% 27.8%

By keeping the sample size of guest talk reasonably consistent, we can infer that host talk proportionally increased over time (although not monotonically). A similar finding is reached on their DID sample by Jucker and Landert (2015), who additionally report (p.33) that average turn length increases markedly for hosts and moderately for guests.

The corpus against which DID is compared is the demographically-sampled Conversation section of the British National Corpus Sampler, from 1994. Henceforth we call this corpus “BNC1994-Conv”. Ideally we would have liked to use conversational corpora from the same periods as we used for DID, but no such data could be found. Both corpora are transcribed orthographically using simple XML- style tags for notable features, e.g. for laughter, for overlaps.

The DID talk will be analysed in a two-step process. The first step uses the method of ‘key POS [part- of-speech] tags’ within the Wmatrix tool (Rayson, 2008; Smith and Waters, 2019) to determine statistically significant differences in frequency between grammatical word classes in DID talk and BNC1994-Conv. Guests’ and hosts’ contributions are compared to conversation separately. This use of key POS-tags works in a similar way to multi-dimension analysis (MDA) (cf. Biber 1988, Nini 2019, for instance), but focuses on particular grammatical features rather than functionally-related clusters of features. Grammatical tagging in Wmatrix is performed by the CLAWS4 POS-tagger (Garside and Smith, 1997), with an accuracy rate of around 97%.

The “C7” list of tags (http://ucrel.lancs.ac.uk/claws7tags.html) used by CLAWS4 captures some useful distinctions for exploring democratization/conversationalization, e.g. titular nouns (NNB), multiple personal pronoun forms. On the other hand, unfortunately it does not always distinguish

7 For the list of speakers and their demographic characteristics, see: https://doi.org/10.25392/leicester.data.11635914.v1.

9 discursive from associated grammatical features: for example, multiword discourse markers such as you know and I mean have to be identified manually by inspecting personal pronoun or verb tags. Similarly, complementizer and relativizer uses of that are conflated under one tag (CST), but can be manually disambiguated.

The next analytical stage is to explore change over time among the grammatical categories that stand out in the previous step. This time close manual analysis will be used where possible, to correct tagging errors and to facilitate interpretation of the results in relation to democratization/conversationalization.

4. Results

The main focus of this section is a description of stylistic changes in DID that have at least a prima facie connection to conversationalization and democratization. However, as a point of departure for this analysis we provide a sketch of the distributions of grammatical features in guest and host talk on DID by comparing each in turn with conversation.

4.1. DID participant talk in comparison to conversation

We first explore significant grammatical differences between guest talk and conversation, and host talk and conversation, across all four sampled time periods, using Wmatrix key POS-tags, and the BNC1994-Conv to represent conversation. The raw results from Wmatrix (a list of overused and underused tags, sorted by significance) have been reorganized into broad word classes, cf. Table 3 and Table 4.

10 Table 3:

POS-tags used with significantly different frequency in DID guest talk and conversation, in all four periods, grouped by general word class. Log-likelihood cut-off point 6.63 (~ p<.01).

Overused by DID guests Overused in Conversation

Personal pronouns Personal pronouns I, me: PPIS1, PPIO1 you: PPY we: PPIS2 it: PPH1 possessive pronoun (e.g. my, our): APPGE he/she, him/her: PPHS1, PPHO1 they, them: PPHS2, PPHO2 Nouns common noun, sg., pl.: NN1, NN2 WH-forms wh-general adverb (e.g. when, how): RRQ Articles wh-determiner (e.g. what, which): DDQ article, e.g. the: AT singular article a/an: AT1 Present tense verbs is, are: VBZ, VBR Adverb/adjective-related does, do: VDZ, VD0 general adjective: JJ has, have: VHZ, VH0 degree adverb: RG, RGT lexical verb. e.g. give(s): VV0, VVZ ordinal adv./adj., e.g. next, first: MD Modals Prepositions core modal: VM general preposition: II BE going to: VVGK preposition of: IO Other Past tense verbs interjection, e.g. oh, yes, mhm: UH was: VBDZ negative particle not/n’t: XX lexical verb e.g. gave: VVD locative adverb: RL had: VHD determiner-pronoun: this, that: DD1 Coordinators/complementizers coordinator and: CC that as complementizer/rel. pronoun: CST

Other hesitations & truncated words: FU

Shading: tags that are not equivalently overused/underused in host talk compared to conversation. (sg.=singular, pl.=plural)

11 Table 4: POS-tags used with significantly different frequency in DID host talk and conversation, in all four periods, grouped by general word class. Log-likelihood cut-off point 6.63 (~ p<.01). Overused by DID hosts Overused in Conversation

Personal pronouns Personal pronouns you: PPY I: PPIS1 possessive pron (e.g. your): APPGE we : PPIS2 he/she, him/her: PPHS1, PPHO1 WH-forms they, them: PPHS2, PPHO2 wh-general adverb (e.g. when, how): RRQ wh-determiner (e.g. what, which): DDQ Present tense verbs wh-degree adverb (how + adj./adv.): RGQ is/are: VBZ, VBR am: VBM Nouns has: VHZ common noun sg.: NN1 lexical verb. e.g. give(s): VV0, VVZ Articles Modals singular article a/an: AT1 core modal: VM Adverb/adjective-related BE going to: VVGK ordinal adv./adj., e.g. next, first: MD Other Prepositions hesitations & truncated words: FU general preposition: II negative particle not/n’t: XX preposition of: IO

Past tense verbs was, were VBDZ, VBDR

Modals let’s: VM21

Shading: tags that are not equivalently overused/underused in guest talk compared to conversation. (sg.=singular)

It can be seen that guest and host talk share a common set of differences from conversation (shown by unshaded tags), but also there are areas where they consistently diverge from each other (shaded tags). For example, relative to conversation, guests overuse first person pronouns (including singular I/me and plural we) and underuse second person pronoun you, whereas hosts do the reverse. Both guests and hosts underuse third person pronouns. These patterns can be understood as straightforward outcomes of the show’s focus on the guest’s life: the guest can attend to this using the first person, while the host needs to elicit such information by addressing the guest directly (see further, Sections 4.2.5 and 5). The same motivation appears to explain the overuse of wh-question forms in host talk,

12 relative both to guests and to conversation, cf. (1) and (2). Questions are a routine resource for interviewers to extract factual and emotional responses from interviewees.

(1) What did you do when you left school? (Roy Plomley to Olivia Manning, 1960s) (2) How was it at home? (Sue Lawley to Gordon Ramsay, 2000s)

Regarding verbs, both guest and host talk use more past tense, and less present tense, than conversation. An important factor here appears to be that the biographical focus of DID is largely retrospective, with hosts eliciting and guests producing descriptions and narratives using past tense; e.g. (3) answers the host question in (2):

(3) Erm, I mean I really, cocooned myself and cut myself off (Gordon Ramsay, 2000s)

In addition, guests and hosts both overuse features associated with nominal style, including common nouns (e.g. record, audition, school), indefinite articles (a/an), and prepositional phrases (e.g. bought through a factory; a year at the Palladium). This suggests that DID talk is more information-focused than ordinary conversation. However, it is noteworthy that guests – but not hosts – use significantly more adjectives (e.g. evaluative adjectives wonderful, great; and career-related adjectives professional, young) than conversation does, as well as more degree adverbs (e.g. very, absolutely). These findings seem to reflect the show’s communicative purposes of eliciting details and evaluative insights by the guest on their life.

The overuse by guests of coordinators/complementizers and and that, relative to conversation, is not paralleled by hosts. The vast majority of cases of and are clause rather than phrase coordinators, and appear to reflect the fact that interviewees can, and are expected to, be more expansive in their turns than participants in conversation ordinarily are.

(4) the thing I find about this music is it totally unwinds me, and I make decisions when I’ve heard this music, it just relaxes me and it has, pleasant memories too of, someone (Carla Lane, 1980s-A) (5) Host: Host: Did you fear for your life there in the Soviet Union? Guest: No not at all, no I just thought I’d probably be, chucked out, and the awful fear that a writer has, that your notes will be confiscated (Colin Thubron, 1980s-C, interviewed by Sue Lawley)

Surprisingly, conversation surpasses guest talk, but not host talk, in frequency of interjections (e.g. mhm, yes), a highly interactive feature. Meanwhile, the tag FU represents ‘unclassified’ words, which

13 are mostly hesitation markers (er, erm) and truncations (words cut short, e.g. because of a self- correction or an interruption). Relative to BNC1994-Conv, guests overuse these features, while hosts underuse them. This latter finding supports views in the literature (e.g. Bell and van Leeuwen, 1994) that host talk in broadcast interviews is more fluent than guest talk.

Another possible surprise is that the most overt indicator of democratization in the CLAWS tagset, the tag for title nouns (e.g. Mr, Mrs, Miss, Doctor) did not feature as a key POS-tag in any of the datasets.

4.2. Stylistic changes across time

We now turn to a diachronic view of salient linguistic features of DID talk identified in the preceding section, this time using closer, mostly manual analysis. The focus will be on closed-class grammatical items, as the noun- and adjective-related categories yielded very many examples and no overall pattern of change in the automatic POS-tagging analysis. When we consider democratization, we do so mainly from the perspective of the guest and host, although in the case of continuers and generic pronouns (Sections 4.2.3 and 4.2.5) there is a stronger case to include the audience’s perspective too.

4.2.1. Wh-questions

Although wh-questions are a feature that consistently sets host talk apart from ordinary conversation (cf. Table 4) and guest talk (Table 3), their frequency in fact halved across the span of the DID corpus, with a particularly pronounced decline after Sue Lawley became host: see Figure 2 and Table A2. The results reported here are frequencies per million words (henceforth: pmw) and based on manually-corrected concordance searches for all POS-tags representing potential wh-question forms (what…?, how…?, why…?, etc.).

14

Figure 2: Wh-questions by hosts: frequencies pmw (manually corrected)

Many of Roy Plomley’s wh-questions are formulaic, in that they tend to be fact-finding closed questions, rather than aimed at exploring feelings, and to apply to the guest’s life in chronological order (e.g. “Where were you born?”, “What was your first job?”). By the early 1980s, however, his formulaicity has reduced, and we see variations such as open-ended questions (6), and invitations to talk that are expressed as statements, cf. (7):

(6) How much does a game depend on the cue? (to Steve Davis) (7) now Mollie I’m so fascinated by your childhood, home medicines, you were sewed up for the winter, now that’s a lovely story (to Mollie Harris)

Questions by Sue Lawley, by contrast, are more likely to elicit emotional response, e.g. (8). She sometimes prefaces her questions by revealing her own attitudes, cf. (9):

(8) How was it at home? (to Gordon Ramsay, 2000s) (9) Er it just worries me all of this psychology because if in the end you’ve got everybody, dressing in erm crumpled, clothes, unironed creased linen and silk er that’s twice too big for them, wh what does that say about us? (to Katharine Hamnett, 1980s-C)

Lawley frequently uses non-interrogative prompts, including significantly higher use than Plomley of the coordinator but to introduce a challenge to what the guest has said; (10) is clarification-seeking

15 but empathetic, (11) is more face-threatening with traces of democratic accountability, inviting the guest (a newspaper editor) to explain how he made his way to the top of his profession.8

(10) But I defy that not to have had some effect on you you must have been quite squeezed (Lawley to Anne Fine, 2000s) (11) But certainly you didn’t get to the, the dizzy heights you’ve achieved w without seeing off the competition (Lawley to , 2000s)

From the guest side, by contrast, it is striking that across all the sampled data there are only two clear instances of wh-questions uttered by guests. In (12) the guest has just explained that his wife deserted him as a result of his single-track devotion to writing books: the question is intended to move on from an uncomfortable topic. The guest in (13) expresses surprise that the host had uncovered some obscure information about him.

(12) Don’t know how she stuck it for 30-odd years, right, what’s your next question? (, 1980s-C) (13) where did you get that story from? (Matthew Bourne, 2000s)

Thus it seems that there has been little movement in guests’ scope to ask questions.9 This observation strengthens the case that the relationship between the host and their guests is, and has consistently been, asymmetrical: the exchanges remain unlike ordinary conversation (where either party is free to ask questions), and in this sense have democratized very little.

4.2.2. Tense, and verb-derived stance expressions/discourse markers

Section 4.1 reported that, relative to conversation, DID participants consistently overuse past tense and underuse present tense. Figure 3 (cf. Table A3) gives a diachronic view of tense development in the show. Note that the frequencies are based on automatic tagging.10 They do not include modals (e.g. can, should), and disregard aspectual distinctions (so that, for example, present tense subsumes present perfect, present progressive and present non-progressive).

8 Cases of but by hosts: 13 in the 1960s, 19 in 1980s-A, 67 in 1980s-C and 155 in the 2000s. 9 Yes-no questions are relatively difficult to find in the corpus, but it seems guests only produce a handful of cases: they are mostly echo or checking questions, rather than the canonical type. 10 Tags for present tense: VV0/VVZ/VBM/VBR/VBZ/VH0/VHZ/VD0/VDZ; for past tense: VVD/VBDZ/VBDR/VHD/VDD.

16

Figure 3. Tense: frequencies pmw (automatic tagging)

We see that guests and hosts mirror each other’s tense selections fairly closely across periods. After a brief surge in past tense in 1980s-A, there have been successive shifts from past to present tense. By the 2000s, guests’ use of present and past tense is close to parity (48.5% and 51.5% respectively), while for the host Sue Lawley, present tense (55.1%) even surpasses past tense (44.9%) slightly. The results suggest a more conversational profile, but not necessarily a more democratic one.

Among the tense results, one area that seems important for conversationalization/democratization is verb-related expression of thoughts and emotions. In, for example, their explanations for choosing a piece of music, guests in more recent years often provide lengthy accounts of the meaning of the piece to them, and present tense can be used to convey these thoughts, e.g. (14), seemingly with more immediacy than past tense would provide (Biber et al., 1999: 457). It is particularly with cognitive verbs in the present tense and a first person subject that guests and hosts express their stance towards different situations, e.g. epistemic stance (judgments about the probability of a situation being the case) and affective stance (attitudes and emotions); cf. the use of I suppose, I think, I want in (14) and (15).

(14) Well the first one I suppose is very evocative of the great outdoors and when I hear it, it’s one of those pieces of music of high emotion, er you can listen to it in a car, you can listen to it, in a room, anywhere you like at any time and it will transport you to, a high field, and a clear blue sky… (Alan Titchmarsh, 2000s)

17 (15) Well I want to hear a lot about your wartime experiences in a minute, but let’s have your fourth record I think we’ve got to, what is it? (Sue Lawley to Dame Vera Lynn, 1980s- C)

The results for present tense also include discourse markers such as I mean, I see, you know, you see, which “function to mark the state of knowledge between participants” (Carter and McCarthy, 2006: 901), and are thus highly interactive, e.g.:

(16) I can’t talk for these other countries, but I mean er if you're, banned, in England, you’re banned er everywhere (Henry Cooper, 1960s) (17) Host: when you, originally set up obviously you know you, became the feed and he became the comic Guest: Yeah I became the straight man and he became er the comedian (Sue Lawley and Ernie Wise, 1980s-C) (18) but actually the dancing was pretty good, you know, we weren’t going around smashing things up (Tracey Emin, 2000s)

Figure 4 and Table A4 demonstrate the changing frequency patterns of verb-derived stance and discourse markers (the latter consisting of I mean, I see, you know, you see and (do) you know what I mean).11

Figure 4. Stance expressions and discourse markers: frequencies pmw (manually corrected)

11 Unlike Biber et al. (1999), we also include pronoun+verb stance expressions with a non-clausal noun phrase as object, e.g. I regret that; I love the lyrics; I remember the day vividly.

18

Guests appear to have led the way in expanding use of both kinds of expression. Epistemic stance remains the dominant type, just as it is in conversation (Biber et al., 1999: 666), and it is this type that has risen most dramatically. The dip in stance markers in 1980s-A is probably because past tense was more prevalent in that period (Figure 3). The rise of discourse markers, meanwhile, has been continuous, with most of the expansion occurring with the highly interactive you know. Among guests it rises from 62 cases in 1980s-A (2,287 pmw) to 141 in the early 2000s (5,127 pmw). Among hosts you know does not appear until Sue Lawley’s stewardship, increasing from just two instances in 1980s-C to 35 in the 2000s (3,305 pmw). Meanwhile you see is vanishingly rare: none of the hosts uses it at all, and guest usage shrinks from a highpoint of 38 cases in 1980s-A (2,287 pmw) to just one case in the 2000s. Thus participants appear increasingly free to express their viewpoint in a spontaneous, conversation-like manner.

4.2.3. Interjections

Interjections are another feature associated with interactivity. Regarding the surprising differential frequencies of interjections in the automated analysis (Section 4.1), firm results from manual analysis of guests’ use are still pending. Hosts’ use of interjections, however, falls into more regular patterns, including a clear preponderance of continuers (or backchannels). Continuers indicate attentiveness and support to an interlocutor, but without seizing the floor (Biber et al., 1999: 1094). An example is Roy Plomley’s overlapping use of Mhm in (19):

(19) Guest: we had to go back because they they were afraid that the war would cut us off, you know Host: Mhm Guest: and that he wouldn’t get back to, Bucharest (Olivia Manning and Roy Plomley, 1960s)

Figure 5 and Table A5 reveal interesting patterns of change among interjections, particularly continuers.

19

Figure 5. Interjections: frequencies pmw for hosts (manually corrected)

From Plomley to Lawley the host’s use of interjections plummets, with just a modest upswing in the 2000s. Plomley’s use of continuers is particularly high, whereas Lawley largely avoids them, suggesting an anti-conversational shift. One possibility is that Lawley, a former journalist, adopts a convention from news interviewing of withholding vocal acknowledgements so as to make clear that the talk is intended primarily for the audience rather than themselves (Clayman and Heritage, 2002: 121). Thus arguably – from the audience perspective as opposed to the guest’s – Lawley’s restraint with continuers may be perceived as more democratic than Plomley’s liberal use.

4.2.4. Hesitation markers and truncated words

Hesitations and truncations (both represented by the tag FU) were found in Section 4.1 to be more prevalent among DID guests, and less frequent among hosts, than in conversation. In each corpus, hesitations are transcribed as er or erm, (20), while truncations are surrounded by , cf. Error! Reference source not found.:

(20) when I was erm about fifteen or sixteen (Bob Champion, 1980s-C) (21) there was a lot of, p pressure from them, older men (Tracey Emin, 2000s)

Figure 6 and Table A6 show the observed diachronic development in DID, although it should be kept in mind that i) editorial processes may have led to dysfluent material being removed from the broadcast episodes; ii) it is difficult to maintain complete consistency in transcribing these

20 phenomena; and iii) examples caught by the tag FU do not capture the whole range of dysfluency phenomena.

Figure 6. Hesitation markers and truncated words: frequencies pmw (manually corrected)

The figures for guests’ hesitations are unexpected in that the highest rate appears to be in the 1960s, followed by a general decline. Truncations are much less frequent and only show signs of increase in 1980s-C and the 2000s. Meanwhile among hosts, hesitations and truncations increase between the Plomley and Lawley eras. By the early 2000s, there is no significant difference between Lawley’s frequency rate and BNC1994-Conv, and she approximates reasonably closely to the rate for guests she was interviewing at that time. While there is some evidence of rapprochement between host and guest talk from the early 1980s onwards, the results need to be treated with caution. As well as the above-mentioned caveats, it is possible that in the earlier years editing out dysfluencies from recordings was technically more challenging.12

4.2.5. First and second person pronouns

Table 3 and Table 4 revealed sharp discrepancies in pronoun usage between guests and hosts in relation to each other and to conversation, specifically an overuse of first person singular and plural by guests and second person by hosts. Over time, however, there appears to have been some erosion of these differences: see Figure 7 and Table A7. Note that the results here include all relevant pronominal forms (I/me/my, we/us/our and you/your) but exclude repaired instances (e.g. the first I in

12 Derek Drescher, who produced DID from 1976 to 1986, reports removing ‘longueurs’ from tape recordings with a razor blade (Drescher, 2017: 85).

21 I I know you joined the army again) and pronoun-derived discourse markers (you know, I mean, etc.), which were covered in Section 4.2.2.

Figure 7. First and second person pronouns: frequencies pmw (manually corrected) (Note: 1sg.=first singular; 1pl.=first plural)

The most striking findings appear to be: among hosts, an increase of first person singular by Sue Lawley, which intensifies over time, and a decline of second person by the end of her stewardship; and among guests, increasing use of second person pronouns. Less conspicuously, however, among the instances of first person plural there appear to be changing patterns in who we refers to, and these could have implications for democratization. We investigate each of these topics in turn. i) First person singular by hosts

We have already seen evidence of the first of these developments in the form of growing use of first- person singular stance markers (I think, etc.) in Section 4.2.2. To these we can add, for example, stance expressions containing modals, cf. (22), which conveys awareness of a guest’s emotional reaction to a piece of music; a playful response to guest’s comments about their music selection, cf. (23); and, in (24), the host constructing herself as an ordinary person with simple assumptions about seasonal dishes. Such first-person instances contribute to constructing the DID host as an engaged dialogic partner with the guest, increasingly drawing on a feature of everyday conversation.

(22) It moves you still, I can tell (Sue Lawley to Matthew Bourne, 2000s)

22 (23) Guest: Now, this is a tearjerker, Host: My handkerchief is ready (Sue Lawley to Alfred Wainwright, 1980s-C) (24) Host: Let’s talk food, Gordon just get the, sort of, juices running, before we, do anything else, it's spring, so I think, cos I’m boring and traditional I think lamb and minted peas and jersey royal potatoes I suppose what do you think? Guest: (Sue Lawley to Gordon Ramsay, 2000s) ii) Second person pronouns by guests and hosts

The increase of second person among DID guests has been partially explored in an earlier paper (Smith and Waters, 2018). It was found that i) when they are quoting speech or thought by someone

(including themselves) guests use you almost entirely with specific, deictic reference;13 ii) they rarely use you to address the host specifically, rather their main use of you is generic and conversely this occurs almost entirely outside of quotation; iii) in comparison to its main competitor, generic one, generic you significantly increased between 1980s-A and the 2000s.

We now extend this line of investigation to guests in the 1960s and to hosts in all periods. The results given in Table 5 are for generic you and one only, since (deictic) singular addressee you is relatively stable.14 Proportional frequencies are provided to illustrate the competition among generic forms. Note also that the figures for generic use include indeterminate cases, i.e. where a clear distinction between generic and specific (deictic) use seemed impossible.

We see that among guests the increase of generic you relative to one has been continuous since the 1960s. Examples include:

(25) curiously enough if you do a little tiny English play about, erm schoolmasters in Norfolk […] you’re immediately selected as just the right chap to write an enormous widescreen epic (John Mortimer, 1960s) (26) one of the things that one, remembers, from one’s early, days, the most impressionable days, in London (Peter Pears, 1960s) (27) it isn’t until of course you look back, that you realise quite er how much you’ve, assimilated over the years (Ray Cooney, 1980s-A)

13 For example: (i) My grandchild said That’s like you, Nana (Sheila Hancock, 2000s). 14 Clear cases of singular addressee you by guests are strikingly infrequent: 11 instances in the 1960s, 3 in 1980s-A, 9 in 1980s-C and 6 in the 2000s. See further, section 5.

23 (28) You just need clothes that work for you in any situation I think (Betty Jackson, 2000s)

Table 5: Generic you and one (excluding quoted use and discourse markers) Participant Generic Freq. measure 1960s 1980s-A 1980s-C 2000s pronoun Guests you n. 67 140 259 211 pmw 2,980 5,486 10,197 7,672 proportion 74.4% 88.6% 94.9% 98.6%

one n. 23 18 14 3 pmw 1,023 705 551 114 proportion 24.0% 11.4% 4.9% 1.4%

Hosts you n. 2 14 24 52 pmw 274 2,580 3,596 4,911 proportion 100.0% 93.3% 100.0% 96.3%

one n. 0 1 0 2 pmw 0 184 0 189 proportion 0.0% 6.7% 0.0% 3.7% Note: ‘proportion’ refers to the proportion of generic you and one combined.

Drawing on Fairclough (1998, 2014), we can argue that you allows the guest to connect their experiences and attitudes to ‘lifeworld’ discourse, that is, discourse associated with the everyday world of ordinary people, in a way that the audience can empathize with (even if the situations themselves are somewhat extraordinary, as in (25)). This is characteristically conversational and arguably also democratizing, particularly in comparison to one (as in (26)) – an option that Fairclough

(1998) describes as more formal and ‘middle class’, even though it is generic by default.15

For a similar reason, perhaps, hosts too seem to warm to generic you, from the early 1980s onwards, cf. (29) and (30):

(29) and you can’t these days assemble all your cast in the library? (Roy Plomley to PD James, 1980s-A)

15 The relatively empathetic qualities of you are also discussed in Bolinger (1979) and Gast et al. (2015).

24 (30) You cannot it seems to me be a sex symbol if you’ve got a sense of humour (Sue Lawley to Cilla Black, 2000s) iii) First person plural by hosts

From a democratization perspective we will be particularly interested in instances where the host makes assumptions, or implies exclusions, that might seem overreaching or ‘undemocratic’ towards one of the parties. After examining examples of we/us/our in the data, the following typology of referents was developed:

• Generic, as in (31); the scope of reference here is wider than the DID participation framework, including people not listening to the programme; • Host + guest + audience, cf. (32); • Host + guest (audience seemingly not included), cf. (33): • Host + audience (guest seemingly not included), cf. (34):

(31) And then we get people like I mean the famous names Stella McCartney Alexander McQueen going abroad you know (Sue Lawley to Betty Jackson, 2000s) (32) Shall we hear the second record you’ve chosen? (Roy Plomley to Mary Archer, 1980s-A) (33) Bob, we were talking about your accent (Sue Lawley to Bob Hoskins, 2000s) (34) Host: We used to marvel how you found a brand-new plot for us every week Guest: Well, somehow I managed to do it (Roy Plomley to Gale Pedrick, 1960s)

Sometimes these types are difficult to differentiate, e.g. (35) could refer to host+guest, to host+guest+ audience, or be generic.

(35) Now Carla Lane, we know, i is a nom de plume (Roy Plomley to Carla Lane, 1980s-A)

In such cases all reasonably probable readings were counted (so the sum of the types sometimes exceeds the actual number of cases of first person plural). The results are presented in Table 6.

25 Table 6: Probable referents of first person plural pronoun (we/us/our) used by hosts 1960s 1980s-A 1980s-C 2000s n. proportion n. proportion n. proportion n. proportion Generic 0 0% 3 18% 6 16% 20 54% Host+guest+audience 13 50% 10 59% 22 58% 12 32% Host+guest 4 15% 7 41% 12 32% 13 35% (excluding audience) Host+audience 9 35% 2 12% 7 18% 3 8% (excluding guest) Total (actual) 26 17 38 37 Note: Manually-corrected results

Despite the categorization difficulties, there seems to be a clear increase in generic we, and few cases where either the guest or the audience was likely to be excluded from the referential scope. But, more importantly for assessing democratization in DID, do there appear to be any cases of ‘abuse’ of we, i.e. where one type of participant might feel marginalized, or unfairly assumed to be included in the scope of the pronoun? (cf. Pennycook 1994).

Example (34) above is not generic but assumes that the audience shares the host’s professed astonishment at the guest’s scriptwriting exploits. It could be an illustration of why some commentators (e.g. Hendy, 2017) have described Plomley’s style as ‘chummy’ and ‘obsequious’ towards his guests. Arguably it marginalizes those audience members who do not share such admiration.

In the Lawley era there are around a dozen cases each in 1980s-C and the 2000s where use of we assumes audience familiarity – but not wonderment – with the host’s life and exploits. Examples (36) and (37) both occur with television celebrities and presuppose universal awareness about the guest with we all.

(36) Host: you’ve had the nose job since Guest: Yes Host: We all knew about the nose job I mean it was all over the papers… we would have noticed anyway though cos the nose did change, radically (Sue Lawley to Cilla Black, 1980s-C) (37) But don’t you long for one of those, beautiful gardens that we would all expect you to have, a kind of, broad expansive dreamy herbaceous bordered, beautiful thing (Sue Lawley to Alan Titchmarsh, 2000s)

26

In (38), spoken to a fashion designer, Lawley’s generic use of we assumes that a lifestyle that seasonally fits in with fashion cycles is the universal norm.

(38) How you do you set about, designing something, new, I mean, we’re all wearing things in one season and you’ve got to think one if not two seasons ahead … (Sue Lawley to Katharine Hamnett, 1980s-C)

However, it can also be argued that a natural consequence of the unnatural situation of a celebrity interview (Bell and van Leeuwen, 1994: 27) is that hosts will need to flatter their guests sometimes, to maintain their willingness to appear on the show. Occasionally pretending that all, rather than simply some, audience members are absorbed in the guest’s life, or share common behaviours with the guest or host, seems to be an understandable, only mildly non-democratic cost of this process.

5. Discussion

The use of a carefully designed DID corpus and the key POS-tags method in this study have been effective in highlighting consistent grammatical/stylistic differences between guest talk, host talk and conversation. Many of the observed differences (e.g. in first and second person pronouns, wh-forms, nouns, adjectives, conjunctions) can be related to the situational characteristics and the participation framework (cf. Goffman, 1981) of the show. The key POS-tags also provided useful pointers for investigating stylistic changes in DID over time.

We now review the observed changes in relation to conversationalization and democratization, considering first where the two processes appear to be congruent, then where they appear less congruent. In doing so we relate the findings to other research, including on the evolving background of BBC Radio 4 and wider social change.

Most of the features undergoing change appear to represent convergence of conversationalization and democratization. These include wh-questions (by hosts); tense-derived stance markers and discourse markers; truncations; hesitation markers (by hosts); first person singular (by hosts) and generic you. The increase by both hosts and guests of stance expressions and discourse markers (especially you know) is an indication that over time DID has given greater weight to expressivity and conversation- like interactive involvement. The reduced fixation by hosts on wh-questions brings them more in line with conversational frequency rates and can be seen as democratizing, in that less mechanical prompts offer guests greater freedom in how to respond. This finding complements Jucker and Landert’s (2015) observation of a decline in frequency of question intonation by DID hosts.

27 The increase we observed in host’s dysfluency features similarly concurs with Jucker and Landert (2015). It suggests rising, conversation-like levels of spontaneity and, supporting democratization, a reduced burden to produce unnaturally flawless talk. However, while guests correspondingly make more truncations, puzzlingly their use of hesitation markers decreases.

Changes among personal pronouns suggest further conversationalization and levelling of the differences between hosts and guests. Both types of participant draw on generic you as a way of relating situations described to the experiences of ordinary people. (Fairclough, 1995, notes a similar tendency of ‘lifeworld discourse’ in political interviews on the same radio station.) Arguably, Sue Lawley’s utterances using first person singular similarly construct her as a down-to-earth conversational partner involved in eliciting the guest’s ‘story’. In the case of we, despite an increase in overreaching generic uses (where the host assumes knowledge or behaviour that not every audience member would share), these cases were few in number and could be interpreted as innocuous rather than undemocratic.

As to areas where conversationalization/democratization might not be aligned, one candidate is the use of continuers by hosts, which dramatically declines from Plomley to Lawley. To the guest, a host who steadfastly withholds continuers might seem less conversationally cooperative and democratic than one who freely uses them. But it can also be argued that through such restraint a host recognizes the guest’s talk as intended primarily for audience members (Clayman and Heritage, 2002: 121), and is therefore being more democratic towards them. This is a practice more associated with news interviews than chat/talk shows (Loeb, 2015), and might have been influenced by Lawley’s journalistic background. A second area where the processes might not converge is singular addressee (deictic) you from guest to host. Unlike hosts, and unlike conversational interaction (where the deictic use dominates, cf. Carter and McCarthy, 2006: 377), it is consistently rare for guests to use deictic you (unless they are quoting someone); instead they make their yous generic, opaque (‘anybody, including you, who might be in that situation’). For this reason we greyed out the arrow from guest to host in the interview participation framework (Figure 1b). The guests’ motivation appears to be similar to the host’s suppression of continuers: to allow the audience to feel included, rather than on the outside of a private conversation.16

As well as suggesting democratization/conversationalization overall, the results provide supportive textual evidence to historical accounts (especially Hendy 2007, 2017) of cultural changes in BBC

16 Another possible area of non-convergence, which we are still investigating, is the expanding share of talk produced by hosts relative to guests. Jucker and Landert’s (2015) word counts show a clearer pattern of change than ours, and moreover they find that length of speaker turns increases over time. This might be democratizing (participants have more time to express themselves), but it is highly unlike conversation, where turn length is generally much shorter.

28 Radio 4 since the 1970s. Although we have focused on grammar, our examples of wh-questions lend support to Hendy’s narrative of increasing emotional frankness on Radio 4 that eventually affected DID, particularly with the changeover of hosts. In contrast to Plomley’s rather formulaic and fact- based questions, Sue Lawley’s questions and prompts invite guests to disclose their emotions more. At the same time she also contributes more of her own sentiments and spontaneity, which may have been facilitated by the slight changes to programme format introduced in the mid-1980s, such as not meeting the guest before the interview (Section 2.2.1). In a book of her reminiscences of DID, Lawley herself hints that she sought a conversation-like feel, stating “like all good conversations, it works best as a partnership” (Lawley, 1990: 3, emphasis added). Even so, Hendy and others have emphasised that compared to other public service broadcasters and against a general background of informalization in British society, Radio 4 has been slow to evolve. In our corpus we find evidence of interactional conservativeness in the persistent near-absence of wh-questions by guests. This is quite striking given that already in the 1980s, on BBC television chat shows such as Wogan it was not uncommon for guests to infringe interactional conventions by asking questions and interrupting (Tolson, 1991).

There are, naturally, limitations to our study. We lack detailed information on the social group balance in the BNC1994-Conv comparison corpus, and it is possible that some differences we have found are artefacts of it including a wider social mix of participants than on DID. Moreover, we have not analysed BNC1994-Conv in equivalent depth to our DID corpus. Analytically, the use of key POS-tags and corpus search methods, though effective, cannot replace close interactional analysis of turn-taking and other stuctural features. Finally, we have not explored the evolution of DID under its most recent hosts, and . We aim to address such issues in further work on DID using an interdisciplinary approach and extended corpus sampling.

6. Conclusion

Despite some issues in disentangling democratization and conversationalization, the study has found corpus evidence that what has become more conversational on DID is generally but not necessarily more democratic, and vice versa. Moreover, the key POS-tags method looks promising as a launchpad or support for investigating stylistic and interactional variation and change in other types of media discourse.

29 References Bell, Philip, van Leeuwen, Theo, 1994. The Media Interview: Confession, Contest, Conversation. University of New South Wales Press, Sydney. Biber, Douglas, 1988. Variation across Speech and Writing. Cambridge University Press, Cambridge. Biber, Douglas, 2006. University Language: A Corpus-based Study of Spoken and Written Registers. John Benjamins, Amsterdam/Philadelphia. Biber, Douglas, Johansson, Stig, Leech, Geoffrey, Conrad, Susan, Finegan, Edward, 1999. The Longman Grammar of Spoken and Written English. Longman, London. Biber, Douglas, & Conrad, Susan, 2009. Register, Genre, and Style. Cambridge University Press, Cambridge. Bolinger, Dwight, 1979. To catch a metaphor: You as norm. American Speech, 54 (3): 194-209. Brown, Julie, Cook, Nicholas, Cottrell, Stephen (Eds.), 2017. Definining the Discographic Self: Desert Island Discs in Context. The British Academy, London. Burger, H., 2005. (3rd edition), Mediensprache. Eine Einführung in die Sprache und Kommunikationsformen der Massenmedien. De Gruyter, Berlin/New York. Carter, Ronald, McCarthy, Michael, 2006. Cambridge Grammar of English: A Comprehensive Guide. Cambridge University Press, Cambridge. Castell, Sarah, 1999. Phone-ins and chat shows. In: Childs, P., Storry, M. (Eds.). Encyclopaedia of Contemporary British Culture. Routledge, London, pp. 392–292. Chang, Phoebe, 2002. Who’s behind the personal pronouns in talk radio? Cartalk: A case study. In A. Sánchez-Macarro (Ed.). Windows on the World: Media Discourse in English. University of Valencia Press, Valencia, pp. 115–152. Clayman, Steven, Heritage, John, 2002. The News Interview: Journalists and Public Figures on the Air. Cambridge University Press, Cambridge. Coupland, Nikolas, 2014. Sociolinguistic change, vernacularization and broadcast British media. In: Androutsopoulos, J. (Ed.), Mediatization and Sociolinguistic Change. Berlin, De Gruyter, pp.67- 96. Doctor, Jenny, 2017. From Forces’ Choice to Desert Island Discs: The BBC's promotion of personal choice in wartime. In: Brown, J., Cook, N., Cottrell, S. (Eds.), Definining the Discographic Self: Desert Island Discs in Context. The British Academy, London, pp. 51-66. Drescher, Derek, 2017. Personal Spin B. In: Brown, J., Cook, N., Cottrell, S. (Eds.), Definining the Discographic Self: Desert Island Discs in Context. The British Academy, London, pp. 83-86. Fairclough, Norman, 1992. Discourse and Social Change. Cambridge: Polity. Fairclough, Norman, 1995. Media Discourse. Hodder, London. Fairclough, Norman, 1998. In Bell, A., Garrett, P. (eds.). Approaches to Media Discourse. Blackwell, Oxford, pp. 142-162.

30 Fairclough, Norman, 2003. Analysing Discourse: Textual Analysis for Social Research. Routledge, London. Fairclough, Norman, 2014. (3rd edition). Language and Power. London, Longman. Garrett, Peter, Bell, Alan, 1998. Media and discourse: A critical overview. In Allan Bell, A., Garrett, P. (Eds.) Approaches to Media Discourse. Blackwell, Oxford, pp. 1-20. Garside, Roger, Smith, Nicholas, 1997. A hybrid grammatical tagger: CLAWS4. In Garside, R., Leech, G., McEnery, A. (Eds.), Corpus Annotation: Linguistic Information from Computer Text Corpora, Longman, London, pp. 102-121. Gast, Volker, Deringer, Lisa, Haas, Florian, Rudolf, Olga, 2015. Impersonal uses of the second person singular: A pragmatic analysis of generalization and empathy effects. J. Pragmat. 88: 148-162. Goffman, Erving, 1981. Forms of Talk. Blackwell, Oxford. Habermas, Jürgen, 1984. Theory of Communicative Action, vol. 1, trans. McCarthy, T., Heinemann, London. Hendy, David, 2007. Life on Air: A History of Radio Four. Oxford University Press, Oxford. Hendy, David, 2017. Desert Island Discs and British emotional life. In: Brown, J., Cook, N., Cottrell, S. (Eds.), Definining the Discographic Self: Desert Island Discs in Context. The British Academy, London, pp. 155-172. Hiltunen, Turo, Räikkönen, Jenni, Tyrkkö, Jukka, DATE. Investigating colloquialization in the British parliamentary record in late 19th and early 20th century. Lang. Sci.VOLUME, PAGES. Hutchby, Ian, 2006. Media Talk: Conversation Analysis and the Study of Broadcasting. Glasgow: Open University Press. Jautz, Sabine, 2014. Who speaks and who is addressed in radio phone-ins? J. Pragmat. 72, 18-30. Jucker, Andreas, Landert, Daniela, 2015. Historical pragmatics and early speech recordings: Diachronic developments in turn-taking and narrative structure in radio talk shows. J. Pragmat., 79, 22-39. Koch, Peter, Oesterreicher, Wulf, 2011. Gesprochene Sprache in der Romania. Französisch, Italienisch, Spanisch. 2. aktual. und erweiterte Auflage. De Gruyter, Berlin. Lawley, Sue, 1990. Sue Lawley’s Desert Island Discussions. Hodder, London. Leech, Geoffrey, Hundt, Marianne, Mair, Christian, Smith, Nicholas, 2009. Change in Contemporary English: A Grammatical Study. Cambridge University Press, Cambridge. Livingstone, Sonia, Lunt, Peter, 1994. Talk on Television: Audience Participation and Public Debate. Routledge, London. Loeb, Laura, 2015. The celebrity talk show: Norms and practices. Discourse, Context and Media, 10, 27-35. Montgomery, Martin, 2007. The Discourse of Broadcast News: A Linguistic Approach. Routledge, London.

31 Nini, Andrea, 2019. The Multi-Dimensional Analysis Tagger. In: Berber Sardinha, T., Veirano Pinto, M. (Eds.), Multi-Dimensional Analysis: Research Methods and Current Issues, Bloomsbury Academic, London/New York, pp. 67-94. O’Keeffe, Anne, 2006. Investigating Media Discourse. Routledge, London. Palander-Collin, Minna, Nevala, Minna, DATE. Person reference and democratization in British English. Lang. Sci. VOLUME, PAGES. Pearce, Michael, 2005. Informalization in UK Party Election Broadcasts 1966-1997. Language and Literature, 14 (1), 65-90. Pennycook, Alistair, 1994. The politics of pronouns. ELT Journal 48 (2), 173-178. Paterson, Laura, 2014. British Pronoun Use, Prescription, and Processing: Linguistic and Social Influences Affecting They and He. Palgrave Macmillan, New York. Rayson, Paul, 2008. From key words to key semantic domains. International Journal of Corpus Linguistics, 13 (4), 519-549. Rühlemann. Christoph, 2019. Corpus Linguistics for Pragmatics: A Guide for Research. Routledge, London. Scannell, Paddy, 1989. Public service broadcasting and modern public life. Media, Culture and Society, 11 (2), 134-166. Scott, Mike, 1997. PC Analysis of key words - and key key words. System, 25, 233-245. Smith, Nicholas, Waters, Cathleen, 2018. From broadcast archive to language corpus: Designing and investigating a sociohistorical corpus from Desert Island Discs. ICAME Journal 42 (1), 167-89. Smith, Nicholas, Waters, Cathleen, 2019. Variation and change in a specialized register: A comparison of random and sociolinguistic sampling outcomes in Desert Island Discs. International Journal of Corpus Linguistics 24 (2), 169-201. Steen, Gerard, 2003. Conversationalization in discourse: Stylistic changes in editorials of The Times between 1950 and 2000. In: Lagerwerf, L., Spooren, W., Degand, L. (Eds.), Determination of Information and Tenor in Texts: Multidisciplinary Approaches to Discourse 2003, Münster: Nodus Publikationen, pp. 115-24. Tolson, Andrew, 1991. Televised chat and the synthetic personality. In: Scannell, P. (Ed.), Broadcast Talk. Sage, London, pp. 178-200. Tolson, Andrew (Ed.), 2001, Television Talk Shows: Discourse, Performance, Spectacle. Lawrence Erlbaum, Mahwah, New Jersey. Tolson, Andrew, 2006. Media Talk: Spoken Discourse on TV and Radio. Edinburgh University Press, Edinburgh. Wouters, Cas, 2007. Informalization: Manners and emotions since 1890. Sage, London.

32 Appendix

Table A1 Guest speakers in the DID corpus. 1960s 1980s-A 1980s-C 2000s Overall (1960-69) (1981-85) (1988-90) (2000-06) Guests total 20 20 20 20 80 Under-50 10 10 10 10 40 Over-50 10 10 10 10 40 Female 8 8 8 8 32 Male 12 12 12 12 48

Lower educ.,a lower occ.b 5 5 5 4 19 Lower educ., higher occ. 5 5 5 5 20 Higher occ., lower educ. 2 2 2 3 9 Higher educ., higher occ. 8 8 8 8 32 a educ. = educational group, based on length and type of schooling (elite/non-elite) b occ. = occupational group, based on estimation of speaker’s need to use standard English in their job; see Smith and Waters (2018).

Table A2 Wh-questions. 1960s 1980s-A 1980s-C 2000s n. pmw n. pmw n. pmw n. pmw Guests 0 0 1 1 Hosts 255 34,879 160 29,482 123 18,427 160 15,110

Table A3 Tense frequencies (from automatic tagging). 1960s 1980s-A 1980s-C 2000s n. pmw n. pmw n. pmw n. pmw Past Guests 1,602 68,817 2,312 85,270 1,997 72,865 1,861 67,670 Hosts 1,122 62,645 1,135 94,527 1,651 75,805 1,755 62,140

Present Guests 371 48,198 173 41,860 377 60,240 807 63,816 Hosts 458 50,745 513 31,878 506 56,479 658 76,211

Table A4 Stance expressions and discourse markers (DMs). 1960s 1980s-A 1980s-C 2000s n. pmw n. pmw n. pmw n. pmw Stance Guests 254 10,911 205 7,561 316 11,530 330 12,000

33 Hosts 26 3,556 19 3,501 30 4,494 63 5,950

DMs Guests 102 4,043 166 6,122 190 6,933 224 8,145 Hosts 4 547 7 1,290 50 7,491 97 9,160

Table A5 Interjections (including continuers) by hosts. 1960s 1980s-A 1980s-C 2000s n. pmw n. pmw n. pmw n. pmw Interjections 422 57,721 294 54,174 22 3,296 122 11,521 Continuers 251 34,332 247 45,513 14 2,097 67 6,327

Table A6 Hesitation markers and truncated words.

1960s 1980s-A 1980s-C 2000s n. pmw n. pmw n. pmw n. pmw Hesitations Guests 659 28,309 490 18,072 573 20,907 440 15,999 Hosts 28 3,830 13 2,395 63 9,438 113 10,671

Truncations Guests 116 4,983 83 3,061 118 4,305 228 8,291 Hosts 18 773 10 369 28 1,022 80 2,909

Table A7 First and second person pronouns. 1960s 1980s-A 1980s-C 2000s n. pmw n. pmw n. pmw n. pmw 1sg. Guests 1,757 75,476 1,856 68,452 2,023 73,813 1,964 71,416 Hosts 36 4,924 34 6,265 63 9,438 171 16,149

1pl. Guests 196 8,420 253 9,331 211 7,699 207 7,527 Hosts 26 3,556 17 3,132 38 5,693 39 3,683

2nd Guests 229 9,837 338 12,466 534 19,484 452 16,436 Hosts 683 93,421 508 93,606 646 96,779 818 77,250 Note: 1sg.=first singular; 1pl.=first plural

34