Show Less
Open access

Conversational Writing

A Multidimensional Study of Synchronous and Supersynchronous Computer-Mediated Communication


Ewa Jonsson

The author analyses computer chat as a form of communication. While some forms of computer-mediated communication (CMC) deviate only marginally from traditional writing, computer chat is popularly considered to be written conversation and the most «oral» form of written CMC. This book systematically explores the varying degrees of conversationality («orality») in CMC, focusing in particular on a corpus of computer chat (synchronous and supersynchronous CMC) compiled by the author. The author employs Douglas Biber’s multidimensional methodology and situates the chats relative to a range of spoken and written genres on his dimensions of linguistic variation. The study fills a gap both in CMC linguistics as regards a systematic variationist approach to computer chat genres and in variationist linguistics as regards a description of conversational writing.
Show Summary Details
Open access

Chapter 5. Conversational writing positioned on Biber’s (1988) dimensions

← 204 | 205 →

Chapter 5.  Conversational writing positioned on Biber’s (1988) dimensions

5.1  Introductory remarks

In the previous chapter, the most salient features of conversational writing were narrowed down and their distributions in traditional writing, speech and a genre of asynchronous computer-mediated communication were contrasted. The chapter explored, for instance, the distribution of modal auxiliary verbs, inserts and emotives, and surveyed a number lexical and paralinguistic features of conversational writing. In the present chapter, we turn to investigating the overall lexico-grammatical patterns found in the corpus of conversational writing (UCOW). The purpose of the chapter is to position the UCOW genres “Internet relay chat” (IRC) and “split-window ICQ chat” on Biber’s (1988) dimensions of language variation (cf. Conrad & Biber 2001a), taking into account all of Biber’s 67 linguistic features (see table 2.1 for a complete list of these). For updated reference as regards face-to-face conversation, the subset sampled in the present study (the SBC subset) from the Santa Barbara Corpus of Spoken American English part 1 (SBC, recorded in the 1990’s) will also be studied and positioned on the dimensions, but reference will no longer be made to the corpus of ACMC (since, as mentioned in section 4.1, only feature count data is available for the ACMC corpus, not the comprehensive raw texts).

In section 3.5 of chapter 3 (Material and method), the procedure for calculating dimension scores was presented. To recapitulate, dimension scores were computed for each text in the UCOW (10+12 texts) and the SBC subset (14 texts). First, frequencies of all the linguistic features were recorded for each text and normalized to text lengths of 1,000 words. The normalized frequencies for IRC’s 10 texts, split-window ICQ chat’s 12 texts and the SBC subset’s 14 texts are documented in Appendix II (tables 5–7). Next, a table of descriptive statistics for each of the genres was compiled; see Appendix II (tables 1–3). The frequencies of the linguistic features in each text, and the mean frequencies in each genre, were then contrasted with the mean frequencies of Biber’s corpus as a whole (all 23 genres), summarized in Appendix II table 4. All frequencies were standardized to conform to a single scale, i.e. a mean of 0.0 and a standard deviation of 1.0, before the dimension scores were computed. Table 5.1 here summarizes the resulting dimension score statistics for the genres of principal concern in the present chapter: Internet relay chat, split-window ICQ and face-to face conversations ← 205 | 206 → from the SBC subset (also in this chapter called “face-to-face conversations SBC”). An analogous summary of dimension score statistics for each of Biber’s genres is found in Appendix VIII.

Table 5.1:  Descriptive dimension statistics for the UCOW genres and the SBC subset


Dimension 1:Informational versus Involved Production
Dimension 2:Narrative versus Non-Narrative Concerns
Dimension 3:Explicit/Elaborated versus Situation-Dependent Reference
Dimension 4:Overt Expression of Persuasion/Argumentation
Dimension 5:Abstract/Impersonal versus Non-Abstract/Non-Impersonal Information
Dimension 6:On-Line Informational Elaboration

The sections below will present the dimension score statistics for all genres on Biber’s six dimensions as graphic adaptations of the dimension figures found in Biber (1988: 128ff). Each dimension graph plots the genres of conversational ← 206 | 207 → writing and face-to-face conversations SBC in relation to the 23 genres Biber studied (17 genres of writing and 6 of speech; see Appendix I). The plotting of the new genres (Internet relay chat, split-window ICQ chat and face-to-face conversations SBC) on Biber’s dimensions follows the tradition developed in numerous post-Biber variation studies in that new genres are positioned in relation to Biber’s established genres’ dimension scores (Appendix VIII) without being part of the calculation of their mean92 or the conception of the dimensions in the first place,93 i.e. without the application of a new factor analysis (cf. Conrad & Biber 2001a: 41, 43–183, Biber 2008: 844). The reader is presumed to be familiar with this tradition when interpreting the dimension plots.

Table 5.2 summarizes the results of an analysis of variance (ANOVA) carried out among the new genres “Internet relay chat,” “split-window ICQ chat” and “face-to-face conversations SBC,” and the results of Biber’s tests among his genres (from Biber 1988: 127). As dimension scores for Biber’s individual texts are unavailable, tests of all genres in combination were not carried out. Pairwise t-tests among the new genres were performed with respect to the dimensions for which the ANOVA returned significant differences (Dimensions 1, 2 and 6), in order to establish which genres differ. Table 5.3 reports the p-values from the t-tests. For dimensions on which the ANOVA yielded no significant differences, no t-test was carried out and, consequently, no p-values are given in table 5.3; instead, the genres on those dimensions are indicated as not significantly different, “n.s.” ← 207 | 208 →

Table 5.2:  Results from ANOVA among the new genres and from Biber’s (1988: 127) tests among his genres. (The p-values for the new genres have been multiplicity adjusted)


Table 5.3:  Results from t-tests among the new genres. Values for probability (p)(no multiplicity adjustment needed)


The dimension score of a new genre essentially marks the genre’s position relative to the mean of all of Biber’s spoken and written genres (which constitutes the zero point of each dimension scale). The dimension scores may assign a genre to the positive or negative end of a dimension scale, but more important than the absolute dimension score is the genre’s position relative to neighboring genres and opposing genres, as will become apparent in the presentation of the dimension plots (figures 5.1a through 5.6b) and associated textual examples. Biber (1988: 129) notes that a proper interpretation of a dimension entails consideration of 1) similarities and differences among the genres, 2) the linguistic features constituting the dimension and 3) the underlying functional and situational parameters associated with the dimension. The graphic presentation of each dimension scale will thus be followed by a discussion of sample texts in the new genres and contrastive genres, with reference to features constituting the dimension, both positive and negative, and to the communicative functions they serve. ← 208 | 209 →

A summary of Biber’s factorial structure is found in table 5.4 (adapted from Biber 1988: 102–103, repeated here from section 2.3 for convenience). Please recall from the presentation of Biber’s (1988) dimensions, section 2.3, that underlying each dimension are the combined sets of features, i.e. the co-occurrence patterns that reflect underlying communicative functions. For Dimensions 1 and 3, the sum of the standard scores of features with negative loadings (features below the dashed line) has been subtracted from the sum of the standard scores of features with positive loadings to obtain a dimension score; for all other dimensions, the standard scores of relevant features have simply been added up. The respective loads of features in table 5.4 were not included in the calculations, other than as indicators of which features to add up and which to subtract, in order to produce the dimension score for each genre.

The genres of focal concern in the present chapter are the conversational writing genres Internet relay chat and split-window ICQ chat, contrasted with Biber’s (1988) numerous genres of writing and speech, as well as with face-to-face conversations SBC. ACMC will not be plotted on the dimension graphs nor discussed beyond this point in the present chapter, since, as mentioned, the unavailability of comprehensive raw ACMC texts renders further textual analysis unfeasible. Nevertheless, one important disclosure with regard to Collot’s ACMC corpus deserves to be made here, as it pertains to dimension scores.

In chapter 4, Collot’s feature counts for the “ELC other” corpus of BBS conferencing were brought in, for contrastive purposes, to represent ACMC (see section 4.1 for an introduction to the corpus). In her 1991 study, Collot applies Biber’s (1988) methodology to compute dimension scores for the ACMC corpus, reportedly based on the standard scores computed, which Collot calls FDS, “feature deviation scores” (1991: 73, results also presented in Collot & Belmore 1996). Upon studying Collot’s (1991) dimension scores, however, the present author found a considerable mismatch between the dimension scores and the constituent standard scores. As a result of the mismatch, Collot’s (1991) dimension scores for the “ELC other” corpus fail to adequately represent the “ELC other” genre of BBS conferencing on several of Biber’s (1988) dimensions, most notably on Dimension 1. To remedy this situation, a new computation of dimension scores for BBS conferencing was carried out in the present study, based on Collot’s (1991: 69–70) “feature deviation scores” for the “ELC other” corpus. Table 5.5 presents the results of the new calculation. ← 209 | 210 →

Table 5.4:  Summary of co-occurring features on each dimension (Biber 1988: 102–103)


← 210 | 211 →

On Dimension 1, the correct dimension score (25.3) positions the “ELC other” genre of ACMC considerably closer to face-to-face conversations than what Collot (1991: 77) and Collot & Belmore (1996: 22) imply (which is c. 8.5). On other dimensions, the scores in table 5.5 represent less grave adjustments of Collot’s results, although significant enough to warrant their documentation here. On Dimensions 4 and 6, Collot’s (1991) dimension scores appear properly computed. Interested readers are referred to table 5.5, Collot (1991) and Collot & Belmore (1996) for further contrastive analysis. The present chapter now turns to the genres of synchronous and supersynchronous CMC and their positions on Biber’s (1988) dimensions of linguistic variation.

Table 5.5:  Corrected dimension scores for the “ELC other” corpus of BBS conferencing presented in Collot (1991) (“n.c.” means that a corrected value was not calculated)


5.2  Dimension plots

In the subsections below, the positions of the conversational writing genres “Internet relay chat” and “split-window ICQ chat” and of “face-to-face conversations SBC” (denoting the SBC subset annotated in the present study) are plotted on Biber’s (1988) six dimensions of linguistic variation (alongside Biber’s 17 genres of writing and 6 genres of speech) and discussed – one dimension per subsection (the first three genres based on table 5.1, and Biber’s 1988 genres on Appendix VIII). To follow up on the bar charts in chapter 4 (e.g. figure 4.4), the written genres are plotted in black in the dimension graphs of this chapter, the spoken genres in gray, and the conversational writing genres in white; see e.g. figure 5.1a. The black and gray dots may thus be regarded as the granular follow-up of the bars for writing and speech in chapter 4, respectively, whereas the white dots were each represented by their own bar in the chapter 4 figures. ← 211 | 212 →

The dimension plots in this chapter are all drawn from Biber (1988), but have been graphically adapted to make room for the new genres alongside Biber’s (1988) great number of established genres, in the format of the present publication. As in Biber (1988), the interpretation of a genre’s position always focuses on the scale of the y-axis, that is, the vertical position of the genre (Biber’s 1988 dimension plots accordingly only have vertical axes). To afford room for the multitude of genres here, however, the plots are slightly slanted; see e.g. figure 5.1a. This means that the genres are plotted in ordinal sequence one step apart on the x-axis, even though the x-axis, of course, is insignificant for the interpretation of a dimension and, consequently, has no scale.

5.2.1  Dimension 1: Informational versus Involved Production

The first of Biber’s dimensions is labeled “Informational versus Involved Production” (Biber 1988: 107). In the adapted dimension scale, figure 5.1a, the genres are plotted in ordinal sequence one step apart on the x-axis ranging from most informational at the bottom left end to most involved on the upper right (adapted from Biber 1988: 128).

Although most variationists after Biber heed the admonition that there is no overall absolute difference between writing and speech, they nevertheless agree that the first of Biber’s dimensions reflects a near-perfect literate vs. oral dichotomy (the dimension scores of written genres are plotted in black and spoken genres in gray in figure 5.1a). New in the Dimension 1 plot here, however, are the genres of conversational writing (Internet relay chat and split-window ICQ chat, plotted in white and with labels in capital letters) and the face-to-face conversations SBC genre (in gray).

The conversational writing genres score well into the involved end of the Dimension 1 scale, with split-window ICQ chat exhibiting a score beyond all other genres. The following text example, (1), taken from the split-window ICQ chat corpus, illustrates the patterns found in texts with very high dimension scores on Dimension 1. Several of the salient features in the text were explored in section 4.4 (in fact, as many as nine of the ten most salient features found in conversational writing in section 4.4 are features that loaded on Dimension 1 in Biber’s (1988) study). ← 212 | 213 →94

(1)<A>there is no reason to hurt me just cause i’m attractive
<B>nah, it’s more that you have that insecure charm
<A>hehe i know i do it so well
<B>girls think your pathetic, and thus flock to love you
<A>hey i’ll let you know that almost all of tamis friends think i’m hot
<B>you know i never saw that movie
<C>Fuck B stool my font
<B>heard it was really twisted
<B>no, you stole it from me off AIM
<A>C learn how to spell
<C>i wad this font on my odl comp
<B>i’ve had it since i started using AIM
<B>and then figured out hoe to do the font change thingy
<A>see thats why i have a black font

Split-window ICQ chat text 11 (UCOW) ← 213 | 214 →

Figure 5.1a:  Mean scores on Dimension 1 for all genres (capitalization denotes conversational writing). Dimension 1: “Informational versus Involved Production” (adapted from Biber 1988: 128).


← 214 | 215 →

Example (1) displays intense personal involvement among the three supersynchronous conversation participants. The split-window ICQ communication in UCOW is highly interactive and affective, and many of its high-frequency features belong to categories with strong positive weights on Dimension 1. Private verbs abound (know, think, saw, heard, learn, see), and several of them are followed by subordinator-THAT deletion (e.g. i know Ø i do it so well,95 you know Ø I never saw, girls think Ø your pathetic, friends think Ø I’m hot, heard Ø it was really twisted). Contractions are the norm (i’m, it’s, your, i’ll, i’ve, thats), with our without apostrophe or standard spelling. Most verbs are in the present tense (is, -’m, -’s, have, r, flock, let, -’ve, s, have) like the private verbs. Possibility modals (such as can) occur more frequently in split-window ICQ than in any other genre. Pro-verb DO frequently substitutes more elaborated constructions (do it so well, do the font change thingy) and general emphatics add force and mark certainty (just, more, really). The supersynchronous texts bristle with first and second person pronouns (me, i, my, you, your), but the most conclusive contribution to the high dimension score is brought by direct WH-questions (3.9 per thousand words, compared to 0.2 per thousand words in Biber’s corpus as a whole). All of these linguistic features together, contrasted with the sparsity of features with negative weights (infrequent use of nouns, prepositional phrases and attributive adjectives), yield a mean dimension score for split-window ICQ beyond all spoken genres (although the score is not significantly different from that of face-to-face conversations SBC, as indicated in table 5.3).

The spoken genres that come closest to split-window ICQ chat on Dimension 1 are face-to-face and telephone conversations. Worthy of notice among them is the face-to-face conversations SBC genre, which scores higher than both the LLC face-to-face and telephone conversation genres (a position also noted by Helt 2001 for American telephone conversations). As can be seen in example (2), the SBC texts have a strong interpersonal focus, and some are fairly intimate in character. They display frequent private verbs (think, know, find, guess), some of which are followed by subordinator-THAT deletion (e.g. I don’t think Ø I am).

(2)Nathan:… Am I doing that right so far?
Kathy:… Mhm.
Nathan:… All the way down to that? ← 215 | 216 →
Kathy:… Mhm.

… I think.
Nathan:… I don’t think I am.

Do you?
Kathy:… And you’d have to have that plus or minus.

Nathan:I don’t know what I did to get that. Where did I get that square root of- um,

… ex squared.
Kathy:Because you brought this over here.

… You brought … three over here. … divided by three,

and then you have ex squared,

so if you want to find ex,

you have the square root of ex squared.
Nathan:… I guess all I can’t figure out is,

what the square root of negative two thir- .. thi- .. two thirds is.

Face-to-face conversations SBC text 9

Contractions (don’t, you’d, can’t) are more common in the SBC subset spoken American English conversations than in the LLC British counterpart, and more verbs are in the present tense (am, think, do, have, want, guess). Analytic negation is frequent and usually contracted (n’t). As in most spoken genres, first and second person pronouns together by far outweigh third person pronouns, although in IRC and split-window ICQ chat, first and second person pronouns individually show superior distribution over third person, as seen in section 4.2. Features with negative weights on Dimension 1 boost the mean dimension score with their low frequencies. Prepositional phrases, for instance, are markedly infrequent in both the SBC and split-window ICQ texts; in the SBC texts they are on average 61.1 per thousand words, and in ICQ only 42.0 per thousand words. Nouns are equally rare in both genres of conversational writing, and slightly more frequent in SBC. The type/token ratio is lower in face-to-face conversations SBC, although more notably the mean word length is higher. All in all, the distribution of features renders a very high mean Dimension 1 score for face-to-face conversations SBC, although not as high as for the split-window ICQ chats.

Dimension 1 is one of the dimensions most clearly associated with a literate/oral dichotomy, and we find that both conversational writing corpora reside in the oral end. Interestingly, however, Internet relay chat scores lower than face-to-face and telephone conversations. Let us find out why by way of an example. ← 216 | 217 →

(3)<chatty’1> hi cheeky1 hows things:)
<{{{odew[[[[[> hey allllll
<Cheeky1> ok thanx and u chatty1?
<{{{odew[[[[[> how u guy doing
<chatty’1> hi mad max how u doing:))
<Sexy-Xhick> hey
<Sexy-Xhick> grrrrrr
<Sexy-Xhick> grrggr
<Sexy-Xhick> ?
<Sexy-Xhick> ;
<Cheeky1> grrrr u
<chatty’1> gret thanks cheeky
<Cheeky1> well eat u
<Cheeky1> lol
<|mad_max|> fine, chatty ….. and u+
<|mad_max|> ?
<chatty’1> great thanx max
<|mad_max|> glad to hear
<Cheeky1> me 2
<chatty’1> hey cheeky we finally got your weather:(
<REVOLI> hi everyone how are you?
<Cheeky1> lol
<Cheeky1> lucky u
<Cheeky1> its horrible here at the moment
<Cheeky1> so wet and muggy
<chatty’1> same here think the house is about to float away
<Cheeky1> lol
<Cheeky1> get those paddles out
<Guest75862> hello
<chatty’1> hehehe got the floaties just incase

Internet relay chat text 3b (UCOW)

As the flow of the IRC conversation is rapid, messages are kept very short and occasionally consist of only one keystroke. Example (3) illustrates how the competition for attention calls for minimal turns, and how these turns manifest themselves with abundant abbreviations and misspellings. Paradoxically, the irregularity of spelling renders a seemingly varied vocabulary, i.e. a great number of types, at least according to the discretion of available type/token ratio calculators (the mean TTR of IRC is 54.9), as seen in section 4.3. The type/token ratio of Internet relay chat is thus comparable to that of written texts such as press reportage, editorials and reviews (whose means range from 54.4 to 56.5) – a factor that inevitably slightly reduces the dimension score of both conversational writing genres. ← 217 | 218 →

Interlocutors in public IRC channels are concerned with finding conversation partners; greetings abound, and the conversations rarely evolve beyond superficiality. The IRC chatters appear less inclined than the split-window ICQ chatters to share personal information, and private verbs common in the ICQ chat (think, know, feel, etc.), where interlocutors are previous acquaintances, are more rare in the IRC corpus. Along with the private verbs go THAT deletions, which are markedly few in the IRC chats compared to the split-window ICQ chats (although found in the IRC example (3) as think Ø the house is about to…).

A surprising, counter-intuitive, finding in the IRC texts is the relatively low frequency of contractions (30.8 per thousand words), as noted in section 4.4. The manual tagging ensured that no apostrophe was needed for them to be found, yet they turn out to be fewer than in the face-to-face conversations, telephone conversations or split-window ICQ chats (which range from means of 46.2 to 55.0 contractions per thousand words). The relative rarity of analytic negation (including its contracted form n’t) in IRC partly helps to explain the low frequency of contractions (as noted in section 4.4). Another likely explanation for the rarity of contractions, however, lies not with the contractions themselves, but rather with the irregular orthography, and lexico-grammatical homogeneity, of the IRC tokens, as well as with the prevalence of inserts and emotives, and nicknames used as address terms. A great number of tokens (besides pronouns I and u) are one mere keystroke long (such as ?, ;, 2), while other tokens are the result of fingers resting on a keyboard key for the entire turn (grrrrrr). This irregularity yields an abundance of tokens, some of them nonsensical; that is, lexically empty. As standard scores are based on relative differences in frequencies per thousand words, several of the linguistic features with positive weights on Dimension 1 score low in the IRC chats (this also partly accounts for the relatively low scores of e.g. DO as pro-verb and demonstrative pronouns). Their common denominator is simply a vast number of irregular tokens, frequently consisting of one mere keystroke (but also repeated greetings, compliments, phatic expressions and attention-attracting tropes such as hi green, hey, gret thanks cheeky, grrrr u). The IRC tokens, moreover, represent a collection of rather few of Biber’s (1988) linguistic features (nine Biber features, for instance, have no textual representation in IRC). In addition, roughly every fifth word in IRC is an insert, an emotive or a nickname address term, i.e. belongs to a category not included in the dimension score calculations. Markedly frequent categories included, however, are direct WH-questions (how are you?) and indefinite pronouns (anyone), general forms of address, which act together with prevalent first and second person pronouns to raise the Dimension 1 score of Internet relay chat. ← 218 | 219 →

At the literate end of Dimension 1 are the texts of traditional writing. Contrasting them with the chatted texts immediately elucidates two of the major distinctions between traditional writing and conversational writing – the salience versus sparsity of nouns, and the sparsity versus salience of first and second person pronouns. In example (4), from the official documents genre, more than every fourth word is a noun (roughly every fifth in the whole official documents genre, compared to roughly every ninth in the split-window ICQ texts).

(4)As before, the record made during the enumeration lists all buildings, residential premises and temporary places of abode, and all households occupying them, as the basis of the enumeration is the household schedule. The number of structurally separate dwellings (that is, houses or flats or other quarters built or adapted for separate occupation and forming a private and structurally separate unit) was obtained as previously, together with the number of households with sole occupation or sharing such dwellings, and the number of living rooms occupied by each household.
Official documents LOB H: text 1

First and second person pronouns are extremely rare in official documents; only every hundredth and every thousandth word belongs to these categories respectively (Biber 1988: 254), compared to nearly every tenth and every twentieth word respectively in split-window ICQ chat. The common characteristic of texts on the literate end is their informational density, resulting from production circumstances that permit careful planning, redrafting and selective word choice. The texts have no affective content and consequently very few private verbs, THAT deletions and emphatics (e.g. just, really, for sure). Contractions and direct WH-questions are completely absent from the official documents genre of the LOB corpus,96 whereas they are pervasive in the chatted UCOW texts. On the other hand, attributive adjectives are twice as common (residential, temporary, separate, private, sole), and prepositional phrases are more than three times as common in official documents as in chatted texts. Words in official documents are on average one character longer than in chats, but their TTR fails to compete with chats, for reasons addressed in section 4.3. In sum, the near-absence of features with positive weights on Dimension 1 combined with the impact of features with negative weights (nouns, attributive adjectives, prepositional phrases and word length) results in very low dimension scores for traditional writing, as illustrated in figure 5.1a. In short, judging from the mean dimension scores, traditional writing is informational whereas ← 219 | 220 → conversational writing is involved. Now, let us briefly consider the genre-internal spread of scores along Dimension 1 and find out if this proposition holds.

Besides the mean dimension scores, table 5.1 (in section 5.1) indicates the genre-internal variation, i.e. minimum and maximum scores, of texts in the genres of IRC, split-window ICQ and face-to-face conversations SBC, in analogy with the descriptive dimension statistics for Biber’s genres, found in Appendix VIII (taken from Biber 1988: 122–125). As mentioned in section 3.5, dimension scores were, in fact, computed not just for each new genre, but for each text in the genres. To illustrate the spread of these scores, figure 5.1b plots the mean and range of scores along Dimension 1 in each genre studied (along with those of Biber’s genres, Biber 1988: 122–125). The texts of split-window ICQ chat, for example, range on Dimension 1 from a minimum dimension score of 19.3 to a maximum of 66.4, a range that is illustrated in figure 5.1b as whiskers around the mean score of 47.2. Our focal concern now is to contrast the whiskers of writing (black dots) and those of conversational writing (white dots).

Figure 5.1b:  Spread of scores along Dimension 1 for all genres (capitalization denotes conversational writing). Dimension 1: “Informational versus Involved Production” (adaptation of Biber 1988: 172–177 and 122–125, supplemented with the new genres).


← 220 | 221 →

A quick look at figure 5.1b strengthens the proposition that traditional writing is informational whereas conversational writing is involved – even the least involved texts of IRC and split-window ICQ chat have dimension scores that exceed those of most of the written genres. The least involved IRC conversations, however, show considerable overlap with personal letters and some overlap with professional letters (but as these letters are not available more specific analysis is unfeasible). Three more written genres are close on the heels of IRC: romantic fiction, general fiction and religion. A closer look at the standard deviation of these three, however, suggests that their texts are fairly tightly distributed around the mean (s.d.<10; see Appendix VIII), which is also the case for IRC (s.d. 7.1; see table 5.1), meaning that only few of their texts overlap on the on the “involved” end of the scale. Compared to the spoken genres, IRC is truly intermediate.

Split-window ICQ chat surpasses all other genres not just in mean dimension score, but also regarding the extent of its scores into the involved end. One third of the ICQ conversations are more “oral” and involved than the LLC face-to-face conversations, although only one conversation in isolation surpasses the SBC subset texts. More striking at the other end is that even the least involved split-window ICQ chat conversation (a statistical outlier with a dimension score of 19.3) has very little in common with traditional writing, except for personal letters. The split-window ICQ chats are highly interactive, personal and affective, which apparently is a characteristic displayed in some of the personal letters and a few of the professional letters, but among the written genres from LOB, only a few general fiction texts display any resemblance – naturally deriving from dialogue such as example (5).

(5)‘But the Old Man doesn’t care for using double-barrelled names, as he calls them. And I think I agree with him. That’s why I use just the plain “Lee” on my cards. But if you think’ – and his expression changed quickly to deliberation – ‘that I should use the Stratford-Lee, just out here I mean, then of course-’‘Oh Lord, no,’ I said, perhaps just a little too abruptly. ‘There are far too many double-barrelled names out here as it is.’ He sat back again, obviously satisfied. ‘I’m inclined to agree with you, sir,’ he said.
General fiction LOB K: text 2

The split-window ICQ chat texts above the bottom outlier text range from dimension scores of 32.7 to 66.4, a range which clearly separates supersynchronous chat conversation from all written genres and renders only analogies with face-to-face and telephone conversation fruitful.

In conclusion then, on Dimension 1, conversational writing most closely resembles face-to-face and telephone conversation, and although the Internet relay chat features are occasionally difficult to interpret and show some parity with e.g. ← 221 | 222 → personal letters, the results indicate firmly that the conversational writing genres are distinct from traditional writing and that supersynchronous conversational writing occasionally even appears to exceed the involvement of speech. However, no single dimension in itself will account for the full range of variation in language, a fact that will become clear as we move on to consider the second of Biber’s dimensions.

5.2.2  Dimension 2: Narrative versus Non-Narrative Concerns

Several scholars who pioneered the analysis of patterns in writing and speech discussed the varying patterns in terms of speech styles (Ervin-Tripp 1972, Hymes 1974, Brown and Fraser 1979). Other scholars studied linguistic features across social groups and situations and came up with labels for basic discourse dichotomies such as high and low varieties (Ferguson 1959), nominal versus verbal styles (Wells 1960), elaborated versus restricted codes (Bernstein 1970), formal versus informal registers (Ervin-Tripp 1972, Irvine 1979) and planned versus unplanned discourse (Ochs 1979). Chafe (1982) was one of the first to empirically identify sets of co-occurring features that characterize written and spoken texts into underlying dichotomous dimensions, which he labeled “integration vs. fragmentation” and “detachment vs. involvement” (as seen in chapters 2 and 4), and Tannen (1982a, 1985) discussed linguistic variation in terms of oral versus literate discourse – the type of discourse distribution partly elucidated by Biber’s Dimension 1 (Biber 1988). However, not many of the scholars were able to account for the range of variation among written and spoken texts as regards narrative concerns.97 Biber’s discovery of the second dimension of variation therefore threw light on a continuum that distinctly separates fiction genres from other written genres and distinguishes among genres of writing and speech, but only in intricate ways without association to divisions of literacy and orality.

Dimension 2 is labeled “Narrative versus Non-Narrative Concerns” (Biber 1988). Genres with high positive scores on Dimension 2 are all associated with past-time narration, whereas genres with high negative scores are similar to each other only in that they lack narrative concerns. As seen in figure 5.2a, the fiction genres cluster by themselves in the bottom left corner (most narrative) and an array of genres share the upper right end of the scale (non-narrative), with Internet relay chat standing out at the top.98 Intermediate in the continuum is a variety of written and spoken genres. ← 222 | 223 →

Figure 5.2a:  Mean scores on Dimension 2 for all genres (capitalization denotes conversational writing). Dimension 2: “Narrative versus Non-Narrative Concerns” (adapted from Biber 1988: 136).


← 223 | 224 →

The linguistic features with a bearing on Dimension 2 were shown in table 5.4. The five fiction genres are all characterized by a high concentration of past tense and perfect aspect verbs, public verbs (e.g. insist, mention, say), third person pronouns, synthetic negation (no, neither, nor) and present participial clauses – typical markers of narrative action. In romantic fiction these features co-occur with very high frequencies, as illustrated in example (6).

(6)He reached over into the back and lifted out his bag.

“But not yours, Mrs. Landry. I attend only to the lower members of your household.”

He said it quite without rancour, and I was positive none was intended.

“But you could be mine,” I insisted.

He inclined his head. “I could, yes. But I would advise you to see your own man, one who knows and understands you.” He shut the door and leaned down through the window to ask, “Are you coming in, Mrs. Landry?”

Romantic fiction LOB P: text 15

The majority of verbs in example (6) are in the past tense (reached, lifted, etc.). Public verbs are prevalent in connection with dialogue (said, insisted), and the reference to characters in third person naturally carries the story forward. Synthetic negation is more common in fiction than in other writing, although not a decisive factor in (6). Present participial clauses add description and narrative action to stories, e.g. Seizing a piece of carpeting Mr. Herman attempted to… (LOB P: text 1), as well as conclusive import to the score – these clauses are nearly absent in non-fictional writing, even more infrequent in speech and completely absent from the chats.

Narration is by definition concerned with the rendering of (human) events, which the narrator communicates directly to the reader/listener. In fictional texts, authors (or their characters) are the narrators, but, to the extent that narration occurs, any communication can take on a narrative flavor. In face-to-face conversation, for instance, speakers typically switch back and forth between the rendering of past events and the discussion of current matters. Consequently, as the positive end of Dimension 2 indicates high density of narrative markers, and the negative end the absence of the same, we can see why the means of face-to-face-conversations (from both LLC and SBC) assume an intermediate position in the continuum, coinciding as they do on the score of -0.6.

On Dimension 2, no linguistic features carry a negative load in the calculation of dimension scores; yet, the paucity of relevant positive features adds up to negative numbers for many of the genres (as explained in section 3.5). Split-window ICQ chat is about as non-narrative as telephone conversations and professional ← 224 | 225 → letters, whereas Internet relay chat is the least narrative of all. Interpreted differently, split-window ICQ chat is more narrative than Internet relay chat, even though neither of them is particularly concerned with narration. The slightly more narrative concern of ICQ chat is partly explained by looking at the relationship between the interlocutors. The ICQ chatters in UCOW are previous acquaintances, most of them high school classmates with friends in common and occasional stories to tell, as in part of example (7).

(7)<10>did u go to the dance last weekend
<J>uhh.. nope.. not reallly.. not much of a fan of dances
<J>but i heard it sucked anyways
<J>the only thing i hated about the dance when i didnt even go there is that i learned that michael took out my sis.. :-(
<J>nah.. i dont think so.. me and spencer is pretty cool.. so i really didnt care much that he went out wit my sister
<10>does he like her
<J>though they both say it wasnt a “date” because they went as “friends”
<10>oh ok
<10>oh well he might get with her
<10>what would u do if he did
<J>i would do absolutely nothing
<J>my sister has her personal life.. i stay away from her personal life because she needs to live her life without me interferring.. ya know?

Split-window ICQ chat text 9 (UCOW)

IRC interlocutors, on the other hand, are not previously acquainted with each other, at least very rarely in real life, and therefore have few common referents. In public channels, their main concern is with finding conversation partners through superficialities (repeated greetings, compliments, phatic expressions and attention-attracting tropes), as seen in example (3) in this chapter. IRC communication in public channels therefore rarely evolves into the narrative state, where interlocutors share stories or relate to events in the past.

Judging from the spread of scores along Dimension 2, illustrated in figure 5.2b, IRC has the highest concentration of non-narrative texts, but two genres have a few texts that surpass IRC’s range into the non-narrative end: professional letters and academic prose. As the professional letters in Biber’s 1988 study are unavailable for scrutiny, we will briefly look at academic prose. Exemplifying the absence of features is tricky, but the non-narrativeness of academic prose might be found in texts such as example (8). ← 225 | 226 →

Figure 5.2b:  Spread of scores along Dimension 2 for all genres (capitalization denotes conversational writing). Dimension 2: “Narrative versus Non-Narrative Concerns” (adaptation of Biber 1988: 172–177 and 122–125, supplemented with new genres).


(8)Changes in voltage accompanying fluctuations of coolant temperature according to equation 6 vary only slightly with concentration and are proportional to the temperature change. Values at various oxygen concentrations of [FORMULA] together with apparent changes in oxygen level for temperature fluctuations of 14 100 C at 500 C are presented in Table 1.
Academic prose LOB J: text 1

Example (8) from academic prose and example (9) from IRC, below, despite their apparent lexical and functional differences, remarkably share the non-narrative space diametrically opposed to fiction. Neither (8) nor (9) has any past tense, perfect aspect or public verbs, third person pronouns,99 synthetic negation or present participial clauses;100 that is to say, they completely lack the typical ← 226 | 227 → markers of narration. Example (9) shows how IRC communication typically is concerned with the immediate present, in this case with greetings as participants are joining and leaving the channel.

(9)<}}melons{{>\/\/elcome Back ^^katy^
<scorpio_>hello }}melons{{
<^^katy^>ty mels…man this is slow
<Guest_162>hello ladies
<}}melons{{>yes ot is
<Rich23>later alls
<^^katy^> bye rich
<Rich23>[bye ^^katy^]
<ROCK>hello ladies
<Farkles>hey, rock
<cristie>hi rick
<Chaser>hi christie
<scorpio_>hm not much talking in here tonight
<cristie>hi chaser
<Syl>Hi Y’all
<scorpio_>just lots of coming and going

Internet relay chat text 1a (UCOW)

The spoken genre on the non-narrative end, broadcasts, also largely lacks markers of narration but is somewhat noteworthy for its unexpected location. The genre stems from the London-Lund corpus radio broadcasts recorded in the 1960–70s, which reported on events actually in progress (sports and events commentary, wildlife commentary and a physics demonstration). The description of sports, events and demonstrations occurred almost exclusively in the present tense and the present progressive, reflecting the ongoing action. Since then, however, radio and television broadcasts have developed into an array of modes: more affective commentary, interviews with players on past events, interaction among commentators with reflections on past events etc., which more likely would place modern broadcasts among the genres with intermediate status on Dimension 2.

To sum up Dimension 2, we can conclude that no functional analysis of variation across writing, speech and conversational writing is adequate unless it also takes into account the narrative dimension, the dimension that places fiction on one end and non-fiction, whether in the form of prose or conversational writing, on the other. We found that both genres of conversational writing largely lack markers of narration and, even though brief passages of narration are ← 227 | 228 → found in the split-window ICQ chats, the genres generally display interaction with the immediate time as the main concern. In the sections below, Biber’s further dimensions will each shed further light upon the genres of conversational writing, as we are ultimately aiming for an adequate, multifaceted description of conversational writing in relation to the textual variation of the English language.

5.2.3  Dimension 3: Explicit/Elaborated versus Situation-Dependent Reference

When communication participants share time and space, as in face-to-face conversation, reference to external objects, events and people is frequently made by temporal deixis (yesterday, tomorrow), spatial deixis (inside, outside) and pronominal reference to objects or humans sharing the same space. In most such circumstances, no confusion arises as to which time, space, object or person is referred to, as the referent is either immediately discernible or can be inferred from shared knowledge or the shared physical context, i.e. by situation-dependent, exophoric, reference (Halliday & Hasan 1976, Biber 1988), as explained in section 4.5. When, by contrast, communication participants (sender and receiver) are temporally and/or spatially separated, reference to such objects usually must be made explicitly, in an elaborated, endophoric, way (Halliday & Hasan 1976). Biber’s (1988) Dimension 3 distinguishes among texts with precisely these separate functions.

Explicit reference is by nature independent of physical context; the wording of a legal act (an official document), for instance, must be equally valid whether read in court, where proceedings are taking place, or in the street of a different city. Furthermore, it must be independent of temporal context, as its wording will stay valid for a long period of time, across generations of readers. To attain this general validity, highly specific reference is needed. Elaborate constructions involving WH relative clauses, pied-piping relative clauses, phrasal coordination and nominalizations are stacked up so that no doubt arises among readers as to intended referents (Biber 1988: 110). The resulting texts are highly integrated and informational and, except as markers of text-internal deixis, time and place adverbials are rare. The linguistic features mentioned are the very features that in co-occurring distribution carry positive weight on Biber’s Dimension 3; see table 5.4.

On Dimension 3 (figure 5.3a), the features with positive weight are contrasted with the absence of features with negative weight and vice versa. ← 228 | 229 →

Figure 5.3a:  Mean scores on Dimension 3 for all genres (capitalization denotes conversational writing). Dimension 3: “Explicit/Elaborated versus Situation-Dependent Reference” (adapted from Biber 1988: 143).


← 229 | 230 →

The genres with positive dimension scores are texts with explicit and elaborated reference, found in the bottom left-hand corner of figure 5.3a. Genres with negative dimension scores are texts with situation-dependent reference (frequent time and place adverbials and adverbs), located towards the upper right. As before, a chain of genres share intermediate positions on the continuum, and written and spoken genres intertwine. A faint pattern emerges, however, according to which traditional writing mainly resides towards the bottom left and speech and conversational writing towards the upper right. A few contrastive examples will serve to shed light on the distribution of texts and the status of conversational writing on Dimension 3.

Example (10), part of a legal amendment from official documents, illustrates vigorous use of constructions used for explicit reference: a WH-relative clause (which provides for), a pied-piping construction (for which), and several nominalizations101 (subsection, section, detention, indictment, imprisonment) – all in less than 100 words. Together with the near-absence of features with negative weight (no time and space adverbials, only one adverb), the text carries a dry, formal tone with large amounts of information packed into each phrase and clause.

(10)In subsection 2 of section fifty-three of the Children and Young Persons Act, 1933 (which provides for the passing of a sentence of detention for a specified period in the case of children or young persons convicted on indictment of certain grave crimes therein mentioned) for the words from “an attempt to murder” to “grievous bodily harm” there shall be substituted the words “any offence punishable in the case of an adult with imprisonment for fourteen years or more, not being an offence the sentence for which is fixed by law”
Official documents LOB H: text 14

Conversely, in a sports radio broadcast, like example (11), reference is typically situation-dependent. That is to say, referents are not identified explicitly or elaborately, but must be inferred from the context of the message, and listeners are forced to invent a mental image of the setting and situations. The speaker assumes the listener’s keen familiarity with the physical context, and the communication is thus perfectly functional despite the spatial divide between broadcaster and listener.

(11)Gowling is beaten #
Dearden comes in on it # ← 230 | 231 →
stabs a foot at it #
and this was quite a fair effort [pause]
no room to swing the foot #
straight into the hands of Stepney #
in the goal to our left thats
Manchester goal # [pause]
from Stepney’s hands it comes out to Kydd #
Kydd now #
with the ball #
on the half-way line # [pause]

Broadcasts LLC 10: text 2

Similarly, just as in face-to-face conversation among people who share knowledge, telephone conversation partners share time and the auditory surrounding, despite being spatially separated. As broadcasts and telephone conversations prove highly situation-dependent on Dimension 3, it seems as though the shared time (the co-temporality) overrides spatial proximity (co-spatiality) as a determiner for situation-dependence. This certainly holds true for Internet relay chat, in which participants are globally dispersed. IRC example (12) displays several time adverbials (later, recently, later) and many adverbs, but no place adverbials.

(12)<Guest22>tks see you later ulsterman
<big-dog>ßÿè o/~ bk
<Guest_404>my son is 9 and a half 22
<yazzie^>Be Right Back
<UlsterMan-Away>Hurry Back yazzie^
<yazzie^>Awwwwwwwwwwwwwwww thanks
<Guest_404>bye UlsterMan
<Guest22>they still do need us, don’t they
<yazzie^>Laughin Out Loud
<Guest_404>indeed 22
<Guest_404>my son was recently diagnosed with Tourettes 22
<Guest22>the children in australia get to be off from school for easter
<Guest22>what is that excactly

Internet relay chat text 4b (UCOW)

On Dimension 3, IRC is positioned in the vicinity of telephone conversations whereas the mean of split-window ICQ chat is slightly closer to face-to-face conversations. Conversational writing, such as IRC and split-window ICQ, is ← 231 | 232 → temporally simultaneous, ranging from synchronous to supersynchronous communication (the latter with completely overlapping turns). The co-temporality factor seems to be at play here, but not conclusively as certain fiction genres and personal letters make almost as frequent use of situation-dependent reference. To sort this out, we will take a look at example (13), which sheds light on the experienced shared space of the split-window ICQ interlocutors.

(13)<1>I dont like you anymore
<A>what u could never not like me
<1>Oh yea?
<1>would you stop with the font… you’re freaking me out!
<A>juice JUICE
<1>blah blah blah
<1>you piece of crap!!
<1>uhh… okay whatever… freak
<1>wow… shot down… that hurt
<1>i hate you :(

Split-window ICQ chat text 1 (UCOW)

ICQ chatters may be spatially distant in real life but their conversation takes place in a visually proximate interface on their computer screens. As the interlocutors inhabit this virtual space, the text of their interaction carries the social situation as well as their relationship to the situation and the objects under discussion (cf. Yates 1996). In chapter 2, this shared space was defined as the semiotic field (Halliday 1978), which, besides carrying the text, allows interlocutors to express themselves graphically beyond actual words (Yates 1993). Example (13) displays ample devices whereby the semiotic field is disarranged (in the original color script); font choice, color and size, upper case, repeated exclamation marks and an emotive. Among the grammatical features with a bearing on Dimension 3, we find a complete absence of positive features in example (13) but several adverbs (anymore, never, out, even, down).

The adverbial distribution in example (13) largely resembles that of face-to-face conversations, exemplified in (14) by a face-to-face conversation from SBC with a similar low score on Dimension 3. In example (14), situation-dependence ← 232 | 233 → is evident in the speakers’ exophoric reference to shared time (stay up late, do this anymore, what can I do now) and their abundant use of adverbs (up, usually, like). In combination with the absence of features for explicit reference (no WH-relative clauses, no phrasal coordination and no nominalizations), such features result in low dimension scores on Dimension 3 for conversational writing and face-to face-conversations alike.

(14)Mary:… God,

I said I wasn’t gonna do this anymore.

… Stay up late.

… Kinda defeats the purpose of getting up in the morning.
Alice:… I know.

.. And it’s a hard habit to break.

Usually I don’t
Mary:It is.

s- Usually I don’t stay up late.

… But it’s like,

if I’m up after midnight

.. It’s just like,
Alice:… Hm.

… Yeah yeah.
Mary:What can I do now.

Face-to-face conversations SBC text 7

We can conclude then, that situation-dependent reference arises in contexts where both co-temporality and co-spatiality are at play, whether separately or in synergy. A few fiction genres manage to evoke a sense of shared time and space between characters and readers, through authors’ vivid descriptions involving frequent use of adverbs. More explicably, however, situation-dependent reference occurs in the spoken genres and in conversational writing, where time, and a real or virtual space, is actually shared.

Noteworthy with regard to the spread of scores along Dimension 3, in figure 5.3b, are the whiskers of academic prose (Biber 1988: 174). Most texts in the genre are informationally dense, highly explicit and elaborated, with relative clauses galore, but to the extent that text-internal deixis is needed, time and space adverbials do occur, which for these texts yield a remarkable spread of scores. As the reader has perhaps noted in the present volume, academic prose indeed ranges from being extremely explicit to being fairly situation-dependent in character. On the next dimension to be considered, academic prose displays another record spread of scores, but more importantly, the two genres of conversational writing diverge slightly and assume positions on opposite sides of the zero point. ← 233 | 234 →

Figure 5.3b:  Spread of scores along Dimension 3 for all genres (capitalization denotes conversational writing). Dimension 3: “Explicit/Elaborated versus Situation-Dependent Reference” (adaptation of Biber 1988: 172–177 and 122–125, supplemented with new genres).


5.2.4  Dimension 4: Overt Expression of Persuasion/Argumentation

On Dimension 4, figure 5.4a, which Biber’s early work calls “Overt Expression of Persuasion” (Biber 1988: 111), it seems only fair that academic prose scores below the zero point – scholarly publications, per traditional definition, should stay neutral and non-opinionated, as “author-evacuated” prose (Geertz 1988: 9) by tradition is “the standard of credibility in academia” (Surman Paley 2001: 31, also discussed by Elbow 1991 and Johns 1997). Dimension 4 has only features with positive weights, and when these add up to high frequencies in a text, the text is considered marked with persuasive and argumentative force; that is to say, it contains a speaker’s or writer’s expression of “likelihood or advisability” (Biber 1988: 148). Conversely, when the features are markedly infrequent, the text has no overt expression of persuasion or argumentation. All the same, the first sentence in this paragraph elucidates how academic prose occasionally may contain a should (as in should stay neutral and non-opinionated), a necessity modal which ← 234 | 235 → adds argumentative but not necessarily persuasive force to an utterance. In later work, Biber seems to address remarks of this kind, and similar confusion arisen with regard to the dimension, by renaming the dimension “Overt Expression of Argumentation” (e.g. Biber 1995: 159). More recently, however, the two stances are combined with a solidus: “Overt Expression of Persuasion/Argumentation” (Conrad & Biber 2001b: 35), which is the label adopted here.

In figure 5.4a, the genres with overt expression of persuasion/argumentation are found in the top right end, those that lack this character drop down into the left bottom corner, and a chain of moderate genres range in between. The features carrying weight on the dimension were listed in table 5.4: to-infinitives, prediction and necessity modals (e.g. will, would, shall, ought, should, must), suasive verbs (e.g. agree, ask, beg, recommend), conditional adverbial subordinators (if, unless) and split auxiliaries (they are objectively shown to…).102

The main concern here will be to shed light on the slightly divergent positions held by the genres of conversational writing in the continuum. Judging from figure 5.4a, neither IRC nor split-window ICQ chat is overtly persuasive/argumentative, but the mean of split-window ICQ chat is higher than that of IRC (although not significantly so; see table 5.3). As it turns out, several split-window ICQ chat texts contain animated discussion, which, as we shall see, is instrumental in bringing the mean for split-window ICQ chat above the zero point (recall that the zero point, in all dimension plots, constitutes the mean of all of Biber’s written and spoken genres, listed in Appendix I).

The low R2 values of Biber’s genres on Dimension 4 (16.9%) and of the added genres (11.0%; see table 5.2) indicate that the importance of the dimension is relatively small in distinguishing among the genres, which tells us we should not read too much into the scores.103 Neither of the conversational writing genres is statistically different from face-to-face conversations SBC (as indicated by the p-values in table 5.3), but some IRC and split-window ICQ chats differ from each other more than others. ← 235 | 236 →

Figure 5.4a:  Mean scores on Dimension 4 for all genres (capitalization denotes conversational writing). Dimension 4: “Overt Expression of Persuasion/ Argumentation” (adapted from Biber 1988: 149).


← 236 | 237 →

Looking into the situational parameters of the communication representing the slightly diverging conversational writing genres here, one underlying functional distinction can be made: the IRC chatters are just beginning their acquaintance with each other, whereas the ICQ chatters are cultivating their real-life acquaintances. Public IRC channels (from which the IRC texts were drawn) often function as scenes for participants to pick up partners for private chat. The communication consists of repeated greetings and self-presentation schemes whereby the user’s age, sex and location are presented, but no participants know each other well enough to bring forward arguments or animated discussion; see example (15).

(15)<rockhard>so i changede it too rock hard
<raindancers>i remember your name i think
<rockhard> u girls are from the uk right
<^^katy^>chanel he’s here…lmao
<angeldelightt> hello
<}}melons{{> \/\/elcome Back angeldelight
<chanel> what?!
<SNOWMAN>how many of u is dancing in the rain?
<raindancers>hugssssssssssssssssssssssssss angel
<^^katy^>matt is on…lmao
<^^katy^>just wait…hehe
<raindancers>just me snowman
<^^katy^>yeah…hahah told ya!
<angeldelightt>hey rain
<chanel>ah ha ha
<^^katy^>hey baby!
<chanel>wel li’m happy for ya
<[MATT]>hey angel :)
<chanel>hey baby goin my way?!
<^^katy^>thanks chanel…
<chanel>Laugh Out Loud

Internet relay chat text 1b (UCOW)

Example (15) is typical of the non-argumentative, non-opinionated style adopted in public IRC. The example contains no linguistic features with a bearing on Dimension 4, although naturally, the features are not completely absent from all the IRC texts.

The split-window ICQ chatters in UCOW, on the other hand, as explained in the analysis of Dimension 2, are friends in real life who share stories and common referents. Their interaction deals with matters shared outside of the ← 237 | 238 → immediate present; they are aware of each other’s attitudes and occasionally dare to challenge them. Their discourse is often opinionated, even to the point of being adversarial, albeit usually exchanged with an ironical glint, as observed in split-window ICQ examples (1) and (13) earlier in this chapter. The adversarial discussions observed in examples (1) and (13), however, come from texts with fairly low dimension scores on Dimension 4 (-0.3 and -0.9, respectively). This means that very few of Biber’s features of overt expression of persuasion/argumentation are found in them; rather, in those texts, argumentation manifests itself in other, more refined, ways. To find out what kinds of expressions bring about positive scores on Dimension 4, we will instead turn to example (16). The example comes from a text that ranks among split-window ICQ’s highest on the persuasive/argumentative pole (2.7, cf. figure 5.4b). It is a non-adversarial, rather supportive text – a motivated discussion on plans for the weekend.104

(16)<5>So i dont know if we shoulf still save $$$ – Which i dont have – or if i need to rush out and get her something with $$ (that idont have)
<5>i know
<5>but i told her we should just celebrate v day some other time ..?
<5>kinda – we didnt really know if tim and val could go til today/last nite cause they had to get out of p[lay practice.
<E>yea I see what you are saying. your in a bad bind man did you how did that go over or was she cool with that cause if she was then I would be like hell yea o i c.
<5>yeah i think we’ll do that – but that may be like celebrating – pause thought- sweet to what? oh – yeah and put the me/her stuff on hold?

← 238 | 239 →
<E>I would do that hell man you and your gf and a couple of friends in the [mountains] that would be sweet. LOL. going up to the [mountains] for the weekend over vday.

Split-window ICQ chat text 4 (UCOW)

Common for the ICQ chat texts with high scores on Dimension 4 is that they, like example (16), contain an element of advice-giving or a discussion of alternative options. The discourse resembles overt expression of persuasion/argumentation in that propositions are modulated by necessity and prediction modals, even though the propositions here are more of an encouraging nature, than of an overtly persuasive or argumentative kind. In example (16), expression of such moderate persuasion/argumentation is carried out by four prediction modals (would, ‘ll), two necessity modals105 (shoulf, should), four conditional adverbial subordinators (if) and a few split auxiliaries. A distribution of this density is not typical of the ICQ texts, but it illustrates the ways in which several ICQ texts deviate from most IRC texts: the ICQ chatters discuss matters from their shared real-life context, which occasionally brings about supportive or challenging argumentation, whereas the IRC chatters rarely exchange views in animated ways.

On this fourth dimension of variation, the spread of scores is extensive in most genres; see figure 5.4b. The texts of the conversational writing genres are only moderately spread. Nevertheless, the diverging ranges of the two conversational writing genres offer a measure of support to the conclusions drawn in the discussion of examples (15) and (16); the IRC texts indeed trail down into the non-argumentative domain, whereas the split-window ICQ chats, to some degree, extend into the persuasive/argumentative domain. ← 239 | 240 →

Figure 5.4b:  Spread of scores along Dimension 4 for all genres (capitalization denotes conversational writing). Dimension 4: “Overt Expression of Persuasion/Argumentation” (adaptation of Biber 1988: 172–177 and 122–125, supplemented with new genres).


Two genres outscore Internet relay chat in being non-argumentative; press reviews and broadcasts. In press reviews, opinions are expressed as if they were the correct view, and therefore the genre largely lacks markers of overt argumentation (Biber 1995: 162, also unexpectedly noted for direct mail letters by Connor & Upton 2003). Live broadcasts, on the other hand, contain inherently non-opinionated, non-persuasive discourse, near-void of the linguistic features on Dimension 4 (Biber 1988). Example (17) from broadcasts serves to illustrate the non-argumentative low extreme: a broadcast from the launching of a submarine, simply reporting a current and immediate progress of events.

(17)Her Majesty’s speaking to him now #
the end of [dhi:] line #
of presentations #
on the Admiralty side #
( – – – music)
now Her Majesty # ← 240 | 241 →
prepares #
to come onto [dhi:] the dais #
Broadcasts LLC 10: text 7a

Along the other end of the scale, several genres surpass split-window ICQ chat in overt expression of persuasion/argumentation, press editorials and professional letters being the most marked types of argumentative discourse (Biber 1988). As the professional letters are unavailable for sampling, below is a brief passage from press editorials to illustrate the dense argumentative force that is largely lacking in the conversational writing corpora. The example is part of a political appeal containing one infinitive106 (to drive), three necessity modals (should) and two split auxiliary constructions (should not despair, should not encourage), all in less than 50 words.

(18)He should not despair of keeping a large part of his copper revenue. O’Brien has praised the valour of Katanga soldiers. Tshombe should not encourage them to drive the point home. Instead of putting up a desperate resistance he should spend an hour reading the Nigerian Constitution.
Press editorials LOB B: text 2

Observing the spread of scores along Dimension 4 in figure 5.4b, we might again, in passing, remark on the extension of academic prose, in this case into the persuasive/argumentative top (cf. Biber 1988: 175). Despite the genre’s moderate mean and non-persuasive intention, a host of academic texts exceed most other genres in persuasive/argumentative force. On the next dimension, however, academic prose recovers its prototypically formal and untainted status by assuming, together with official documents, a position distinct from all other genres, a position diametrically opposed to conversational writing.

5.2.5  Dimension 5: Abstract/Impersonal versus Non-Abstract/Non-Impersonal Information

The labels of Biber’s dimensions of variation have altered slightly over the years as contrastive genres and perspectives have brought new light to them. Dimension 5, for instance, was labeled “Abstract versus Non-Abstract Information” in Biber (1988), “Abstract versus Non-abstract Style” in Biber (1995) and Conrad & Biber (2001b), and “Impersonal versus non-impersonal style” in Biber et al. (1998), although in essence the variant labels point to one ← 241 | 242 → and the same dimension of variation: that between academic prose on the abstract end and telephone conversations on the non-abstract end. New in the Dimension 5 plot here, figure 5.5a, are the genres of conversational writing and face-to-face-conversations from SBC, which essentially prove all these variant dimension labels adequate, but shed further light on the non-abstract, non-impersonal, i.e. the rather concrete and personal, end of the scale and revaluate informality.

Dimension 5 has only features with positive weights, although as before, the absence of the features yields negative scores. Figure 5.5a plots the mean dimension scores along Dimension 5. The genres with high positive dimension scores, academic prose and official documents, are found in the bottom left corner, and a number of other written genres cluster in intermediate position in the continuum. Conversational writing and face-to-face conversations SBC reside in the upper right corner (with no significant internal difference; see tables 5.2 and 5.3), closely interspersed with the other conversational genres and followed by personal letters, spontaneous speeches and the fiction genres.

The linguistic features with a bearing on Dimension 5 were given in table 5.4 above (from Biber 1988: 102–103). A high concentration of the features brings about the formal and complex technical style typically found in academic prose: conjuncts (e.g. alternatively, conversely, moreover), passives, past participial clauses (e.g. based on the current rate, the value…), WHIZ deletions (e.g. the value Ø based on the current rate) and certain adverbial subordinators (e.g. since, while, whereby). These co-occurring forms are devices whereby authors present information with reduced emphasis on the agent, forms that either demote the agent to a by-phrase or elide the agent altogether, and instead typically give prominence to a non-animate referent or an abstract concept. In academic prose and official documents, conjuncts and adverbial subordinators mark logical relations among clauses and serve to make complex reasoning explicit. Example (19) illustrates the frequent use of passives and complex clause constructions for dense integration of information in academic prose. No human agent is detectable in the discourse. ← 242 | 243 →

Figure 5.5a:  Mean scores on Dimension 5 for all genres (capitalization denotes conversational writing). Dimension 5: “Abstract/Impersonal versus Non-Abstract/Non-Impersonal Information” (adapted from Biber 1988: 152).


← 243 | 244 →

(19)The relations estimated were between the rates of change of the ‘wage drift’, the level of ‘excess profit’, the level of ‘excess demand’ and the rate of change in productivity. It may be pointed out that in our model productivity makes a significant contribution to the explanation of the spread between earnings and wage rates, when all variables are expressed as levels, but ceases to be a significant factor in our least squares computation in which variables are subjected to a first-difference transformation.
Academic prose LOB J: text 44

Dimension 5 is the third dimension, after Dimensions 1 and 3, to reflect a literate/oral dichotomy, i.e. a pattern emerges here, in which genres of traditional writing mainly plot in the lower, abstract end of the figure and speech in the upper, non-abstract end. The pattern closely resembles Dimensions 1 and 3 in that, apart from speech, only fiction genres and letters score in the vicinity of conversational writing. Judging from Dimensions 1, 3 and 5, therefore, we can determine that conversational writing adheres to the oral domain of the polarity, where the mean of one of its genres even takes the over-all lead (on Dimension 1, split-window ICQ, and on Dimensions 3 and 5, IRC). On Dimension 5, the absence of linguistic features with highly integrative textual functions entails the presence of features with concrete referents and active agents (although the latter features are not tagged specifically in Biber’s 1988 study, and therefore not counted). Face-to-face conversations and conversational writing typically deal with matters of immediate, current relevance to participants. Agents are animate or tangible objects, and the topics discussed are concrete, as in example (20) from SBC of the interaction among the same interlocutors as in example (8), chapter 4, who are cooking a meal together.

(20)Pete:They aren’t particularly stringy.
Marilyn: Oh.
Then just snap em.
Roy: That probably looks like a three-person salad bowl,
Pete: I’ll just and put em,
and put them…
Roy: hunh?
Marilyn: Man that’s a big hunk of fish.
Pete: Where do you want em put.
Marilyn: Shit,
it’s a huge…
Pete: Are they just going on that,
or, ← 244 | 245 →
you wanna put em in a colander,
and then wash em?
There’s a colander.

Face-to-face conversations SBC text 3

Example (20) contains no features with a bearing on Dimension 5;107 neither does example (21) from IRC, below, with interlocutors discussing an image file transfer, a concrete albeit virtual object.

<Genie500> Laughing Out Loud
<River>my hair is almost as long as yours
<Genie500>now ya know who to look for honking across the street
<Genie500>really?? lol
<River>well just in the back
<Genie500>Laughing Out Loud
<Genie500>and what color is yours??
<lookingforagirl> blue
<Genie500>oh river just a sec I gotta turn something off for
you to send okay
<River>this one is from 95 without the glasses.
<Genie500> okay try again
<River>but the hair is almost the same now as then
<River>plus a wee bit more grey in it
<Genie500>Laughing Out Loud ok

Internet relay chat text 4a (UCOW)

Examples (20) and (21) are texts from informal settings with highly personal content and interactive, loosely integrated discourse, i.e. typical texts with extremely high negative scores on Dimension 5. Similarly, the chatters in the split-window ICQ chats discuss non-abstract and non-impersonal, i.e. personal, matters. Example (22) is a passage of a very personal kind, which contains none of the features that carry weight on Dimension 5. ← 245 | 246 →

(22)<6>i know…he’s like one day “omg i like you sooo much etc..” and the next he’s all pissy
<6>he has like PMS
<6>yeah….i do…but sometimes i can’t take when he’s in a bad mood..
<6>like idk i’m one of those scarcastic girls…i can’t help it i make fun of him when i get the chance…and he gets so mad at me for that
<F>owch…do you like him?
<6>but he knows i’m just messin
<F>some guys take that stuff personally
<F>yeah i mean when he takes stuff so seriously, and he’s in the bad mood all the time..not to mention he’s a distance away, correct?
<6>i know…like i see him…even when i don’t plan to cuz he’s best friends w/ my cuz…so the distance isn’t even an issue..cuz me and my cuz r together like all teh time

Split-window ICQ chat text 5 (UCOW)

In figure 5.5b, the texts exemplified in (20) and (21) rank among the “top” texts at the non-abstract/non-impersonal end, which for the genres under discussion could be renamed the concrete/personal end. Judging from the spread of scores, however, texts from several genres touch the same non-abstract/non-impersonal end; 14 out of the 26 genres have texts with a minimum dimension score below (here “above”) -4 (i.e. a sum that is four standard deviations below the mean of all of Biber’s genres). Most notably, no genre has a text with a dimension score “above” the “top” IRC texts, but five genres reach the same “top” score of -4.8 (viz. general fiction, personal letters, split-window ICQ chat, face-to-face conversations SBC and telephone conversations). No other dimension shows an equally uniform distinct end to dimension scores, which naturally raises the intricate question whether there is a distinct far end to being non-abstract/non-impersonal (i.e. concrete and personal) – an initially mind-boggling question that, however, has a simple answer. Whenever a text displays no single occurrence of any of the features with a bearing on the dimension, the text will receive the “top” score. Two IRC texts, two split-window ICQ chat texts and one SBC text display such a distribution, along with an unknown number of texts from general fiction, personal letters and telephone conversations.

With regard to Dimension 5, we can conclude that conversational writing most closely resembles the genres of face-to-face and telephone conversation, but also that no text of conversational writing surpasses the most informal texts of face-to-face or telephone conversations – all four genres have texts that reach the same distinct non-abstract, non-impersonal end. ← 246 | 247 →

Figure 5.5b:  Spread of scores along Dimension 5 for all genres (capitalization denotes conversational writing). Dimension 5: “Abstract/Impersonal versus Non-Abstract/Non-Impersonal Information” (adaptation of Biber 1988: 172–177 and 122–125, supplemented with new genres).


5.2.6  Dimension 6: On-Line Informational Elaboration

Dimensions 6 and 7 have few linguistic features with important loadings and are thus difficult to interpret. Accordingly, most studies applying Biber’s 1988 methodology have not considered these dimensions; in fact, Biber’s (1988) study discards Dimension 7 on theoretical grounds on an a priori basis, finding its factorial structure too weak for further exploration. Dimension 6, nevertheless, will be considered here, for the sake of a complete analysis, as even tentative results might be worthwhile to study in the exploration of conversational writing.

Dimension 6 has only features with positive weights; see table 5.4: three types of dependent clause (THAT complement clauses on verbs and adjectives and THAT relative clauses on object position) and demonstratives (that, this, these, those preceding nominals, not to be confused with demonstrative pronouns, which load on Dimension 1). Texts with high scores of the features are informationally elaborate, yet display relatively unplanned discourse, i.e. a type of discourse produced under real-time constraints, and the dimension is therefore ← 247 | 248 → labeled “On-Line Informational Elaboration.” By contrast, texts lacking the same features are regarded as containing no on-line informational elaboration. The positions on Dimension 6 of the genres studied are shown in figure 5.6a.

Genres with high dimension scores are the less involved (cf. Dimension 1), i.e. less interactive, spoken genres spontaneous speeches, interviews and prepared speeches. They recurrently have an informational focus and often convey the speaker’s attitude or beliefs, but are produced in strict real time.108 Example (23) is part of a spontaneous oration given in the House of Commons, exemplifying real-time informational elaboration.

(23)I do not think that it would be helpful # [pause]
to to engage #
in sort of in name calling #
against the opponents # [pause]
of the Concorde project #
and certainly #
neither I #
nor my right honourable friend #
intend to follow the honourable gentleman in that regard #
Mr Speaker #
will my honourable friend accept #
that many people in this House #
think that Concorde is going to be a gigantic financial disaster # [pause]
will he ensure that in any cuts in public expenditure #
education and social services take priority over this huge pit into which money is being poured #

Spontaneous speeches LLC 11: text 4

The pauses in example (23), here retained from LLC’s original prosodic transcription, reflect the speaker’s planning time required to further elaborate the subject matter. THAT complement clauses are used ad hoc on verbs to add pieces of information (I do not think that…, will my honourable friend accept that…, many people in this House think that…, will he ensure that…). In this way, information is tacked on as the speaker progresses, rather than integrated tightly into the text (Biber 1988: 157). In combination with demonstrative determiners, which are typically thought to be informal (that regard, this House, this huge pit), the tacked-on information renders evident the production constraints of time and situation. In unplanned discourse, like that in example (23), demonstrative determiners are often preferred to articles (Ochs 1979). ← 248 | 249 →

Figure 5.6a:  Mean scores on Dimension 6 for all genres (capitalization denotes conversational writing). Dimension 6: “On-Line Informational Elaboration” (adapted from Biber 1988: 155).


← 249 | 250 →

As can be seen in figure 5.6a, face-to-face conversations from both SBC and LLC are unmarked with respect to the features on this dimension (the genres with signs of on-line informational elaboration appear in the upper right end of the scale in figure 5.6a). On the other hand, figure 5.6b reveals that both genres of face-to-face conversations show an extensive spread of scores, ranging from no informational elaboration to highly on-line informational. On-line informational elaboration is typically found in personal communication where speakers elaborate on information while at the same time indicating their stance on subject matters. That is to say, it occurs where the linguistic features counted on Dimension 6 enable the encoding of attitude towards propositions, as in example (24) from a business conversation between board members.

(24)Phil:…Uh I would prefer that,
that you were there on one hand,
because I think that it would be most expedient.
Phil:But I think,
..what was ..felt,
was that at this point,
rather than ha-
than create
…I don’t really f- find it to be, know,
..a …confrontation,
by any means,
Phil:I just think,
..they wanna be able to just kind of …figure out,
I think our board eh,
…quite frankly we have more …problems to resolve interior, than we do ..outside of it

Face-to-face conversations SBC text 10

On-line informational elaboration is often found in communication such as example (24) where a speaker’s turn is fairly long, i.e. more monologic than in typical face-to-face interactions. The text from which example (24) derives scores high on Dimension 6 (6.9), this brief example indicating two THAT complements on verbs109 (would prefer that, that…, I think that…) and one demonstrative ← 250 | 251 → (this point). Here again, pauses enable the speaker’s mental redrafting and THAT complements are added on ad hoc to elaborate the discourse. Face-to-face conversations, on average, contain more on-line informational elaboration than do telephone conversations and conversational writing.

What about in conversational writing, then? Internet relay chat and ICQ chat are indeed both carried out in real time, on-line. Firstly, Biber’s term “on-line” was coined before the advent of CMC among the general public, and must be interpreted as “live” in current situational analyses. Secondly, before we address this question we must recall what was learned in the analysis of Dimensions 1 and 3 about the nature of conversational writing; computer chat is not informational, and it is not explicit/elaborated. Rather, the discourse displayed in computer chat is highly involved and interactive, with abundant situation-dependent reference. From Dimension 6 we learn that hardly any live informational elaboration takes place in conversational writing, or rather, informational elaboration is not carried out live in conversational writing. We saw in examples (3), (9) and (15), above, indications of how the interlocutors in IRC are mainly concerned with finding conversational partners, that greetings abound and conversations in public channels rarely evolve beyond superficiality. The IRC participants rarely share personal information in their typically brief turns, producing fewer private verbs (think, know, feel, etc.) than for instance face-to-face conversationalists. Decisive for the low scores on Dimension 6, however, is the fact that both modes of conversational writing allow interlocutors to edit their contributions, in IRC before sending, and in split-window ICQ by real-time erasure and replacement, which minimizes the need to add complement clauses ad hoc. With this in mind, the low dimension scores of conversational writing on Dimension 6 come as no surprise. Example (25) illustrates a few carefully edited turns in IRC (judging from turn length, word length and complex nominal constructions), a passage in which a participant (_oups) is asking for help with his/her business assignment.

(25)<_oups>without getting into details…what would be best… takeing a depreciation allowance and use your retained profit, or taking a bank loan, with favourble interest, and inviting new shareholders (its Ltd)
<AdamSxy35>can i take both?
<_oups>well yeah…but that would be a disadvantage..
<livinboy> americans rule ← 251 | 252 →
<AdamSxy35>oups why dont you try a business chat room on yahoo?
<_oups>hm…well do they have that..
<AdamSxy35>it works for me when i cant fall asleep ;)
<CityWoman>My guess would be to take a bank loan and show positive cash flow and reserves.. it is alwasy a good idea to use someone elses money.
<Zerj> 15/m/canada with webcam and netmeeting – MSG me me to chat
<AdamSxy35>can i have a loan CW?
<^sarah^xx^>me back again
<CityWoman>24% AdamSxy35.. ha!
<CityWoman>MY positive cash flow.. !
<Jivie>AdamSxy35, can you share it? :p
<AdamSxy35>i get better rates from guido on the street corner ;)
<CityWoman>Yah but Guido breaks legs.. ha!

Internet relay chat text 5b (UCOW)

Example (25) has no instance of the features counted on Dimension 6. The example derives from a text with an overall dimension score of -3.5 and is thus typical for Internet relay chat with regard to the dimension. The text in question, Internet relay chat text 5b, is further typical of IRC in that it scores high on Dimension 1, meaning it is not informational, and it scores fairly low on Dimension 3, meaning it is not particularly elaborated in reference. The seemingly elaborated turns of this IRC example are thus the result of careful editing, prior to posting, rather than real-time elaboration. No THAT clauses are added on to objects, or brought in as complements to verbs and adjectives, and no demonstratives are used to replace articles.

While participants in IRC may take advantage of editing options, before sending, to elaborate their turns, interlocutors in split-window ICQ can edit their contributions at any time in their full semi-window. In addition to the text format, several of the texts in the split-window ICQ component of UCOW are found as video clips of the chat screens (as mentioned in section 3.3), which render chatters’ redrafting explicit. Upon studying these, it becomes evident that ICQ chatters indeed use their editing options frequently, but mostly to erase passages, to replace strings of text and to correct their spelling, and very rarely to elaborate their propositions. The text format split-window ICQ corpus consists of the resulting texts after the chat session ended, and if there were any elaboration by complement THAT clauses, for instance, it would be evident in these. As can be ← 252 | 253 → inferred from the dimension score of split-window ICQ chat on Dimension 6, however, such evidence is very rarely found. Example (26) from ICQ contains one demonstrative (this), but no other features that mark added-on informational elaboration. From the clip of the interaction displayed in example (26), it is evident that both interlocutors frequently scroll back to edit their spelling, but not to elaborate their turns.

(26)<7>hahahh did you get the lift on your truck
<7>oh yeah
<G>na, i did that when i was eating cinamon toast crunch this morning, i think im just going to get rimes and tires, i already lifted the from of my truck 3 inches
<7>if you go off roading with it why would you put rimes on it
<G>so i can have bigger tires, im not getting like 22” or nething
<G>im getting 15” american racing rims, and 35 in tires, like the rimes on mike crowells jeep, or jake mitchells truck, 1100, i just don’t know if i feel like paying for that cause i have the money but i want to take off a month during summer and yeah…….bye
<7>how much
<7>yeah later

Split-window ICQ chat text 6 (UCOW)

The low score of split-window ICQ chat on Dimension 6 is thus the result of an over-all lack of elaboration, though not necessarily a lack of editing. ICQ chatters simply edit their texts by scrolling back and replacing letters, words and phrases, not by elaborating their propositions post hoc.

Biber (1995: 167) tentatively labels the negative end of Dimension 6 “Edited or Not Informational.” With regard to conversational writing, both parts of the label are corroborated, even though “edited” in the live conversational writing context bears little resemblance to published traditional writing. In IRC, the rare instances of elaboration are composed prior to sending, and in split-window ICQ through live editing, and in neither case is informational elaboration added on at the end of existing turns. Figure 5.6b, finally, displays the spread of scores on Dimension 6, uncovering considerable ranges of variation in most genres. Evidently, features of on-line informational elaboration (THAT dependent clauses and demonstratives) are produced, or tolerated, in most genres, although less so in conversational writing than in conversational speech. ← 253 | 254 →

Figure 5.6b:  Spread of scores along Dimension 6 for all genres (capitalization denotes conversational writing). Dimension 6: “On-Line Informational Elaboration” (adaptation of Biber 1988: 172–177 and 122–125, supplemented with new genres).


5.3  Chapter summary

In the present chapter, one of the primary aims of the study was accomplished; the conversational writing genres Internet relay chat and split-window ICQ chat were positioned on Biber’s (1988) dimensions of linguistic variation, alongside Biber’s multiple written and spoken genres from LOB and LLC. The focus of the chapter was to elucidate the nature of conversational writing by discussing the lexico-grammatical features that contribute to the positions of the conversational writing genres, neighboring genres and contrastive genres, on the dimensions. The distribution and functions of the features were explored, inter alia, by way of contrasting numerous textual examples from the genres. This chapter dealt with each dimension on its own terms; in the next chapter, a more overarching approach will be taken, as the results from the full investigation now may be brought together and discussed in combination.

92 This standpoint was taken after a trial inclusion of UCOW and the SBC subset in the calculation of new means for an amalgamated corpus consisting of UCOW, the SBC subset and Biber’s 23 genres, which yielded essentially comparable scores. The UCOW and the SBC subset together consist of approximately 30,000 words (see table 3.1), whereas Biber’s established genres total approximately 960,000 words (see Appendix I). The resulting scales altered the zero point marginally, but in all significant respects genres kept their ordinal positions and relative distances on the dimensions.

93 As mentioned in footnote 92, the size of the corpora representing the new genres (UCOW and SBC subset) is small, meaning that the inclusion of the corpora in a new factor analysis of the spoken and written texts studied here might only very marginally alter the layout of dimensions. This notwithstanding, such an effort is naturally both feasible and commendable if ventured into, in future studies, with regard to new corpora of writing and speech.

94 Of the most salient features in conversational writing explored in section 4.4 (see table 4.6), eight are “positive” features on Dimension 1 (first and second person pronouns, direct WH-questions, analytic negation, demonstrative and indefinite pronouns, present tense verbs, predicative adjectives and contractions) and one is “negative” (prepositional phrases). The tenth most salient feature, predicative adjectives, did not load significantly on any dimension in Biber’s (1988) study.

95 Identified by Biber’s algorithm as containing a subordinator-THAT deletion (Biber 1988: 244), despite the possible alternative reading as two asyndetically coordinated main clauses. To attain results comparable with Biber’s, his algorithms were closely observed at all times (Biber 1988: 222–245). Setting Biber’s algorithm aside would have rendered incomparable results.

96 Actually, there is one (1) contraction in the 28,000-word official documents component studied.

97 Although see Tannen (1982b) for a vigorous exception.

98 The y-axes in Dimensions 2, 3 and 5 are reversed to facilitate comparison across dimensions and for the dimensions’ interpretive names to read correctly from left to right across genres.

99 Pronoun it is not a feature on Dimension 2.

100 Biber’s algorithm detects only present participial clauses preceded by punctuation or a tone unit boundary (Biber 1988: 233).

101 Biber’s algorithm for the automatic detection of nominalizations includes all words ending in –tion, -ment, -ness or –ity, whether with or without verbal origin or English stems.

102 “Split auxiliary” means an auxiliary split from the main verb by adverb(s). Biber’s (1988: 244) algorithm for the detection of split auxiliaries was interpreted aux+adv+(adv)+v, the last element being v (any verb), instead of his suggested vb (base form of verb).

103 The R2 value “indicates the percentage of variation in the dimension scores of texts that can be accounted for by knowing the genre category of the texts” (Biber 1988: 126). In our case less than 16.9% of the variation in dimension scores along Dimension 4 can be accounted for by knowing the genre categories.

104 Example (16) contains several extensively overlapping turns, which the logging software fails to record in logical sequence. The video clip of the example, however, shows the turns of <5> and <E> dialogically juxtaposed.

105 Only core modals are counted, not semi-modals; see section 4.2.

106 Only to-infinitives are included in the calculation of the Dimension 4 scores.

107 The verb put in the construction want them put does not qualify as an agentless passive since Biber’s (1988: 228) algorithm for agentless passives only detects those preceded by a form of the verb BE+noun/pronoun/adverb.

108 Prepared speeches in LLC retain “some spontaneity in not being read from a script,” “therefore allowing for improvisation” (Greenbaum & Svartvik 1990: 12).

109 Biber’s (1988: 230) algortihms for “THAT verb complements” detects instances of that preceded by a tone unit boundary, which is the case for the second that in would prefer that, that… The first instance in the example is not counted, as it is not preceded by a public, private, or suasive verb, a seem/appear, or any other identifying item.