20.11.1997

Seminar Paper

Linguistics

Submitted to

Prof. Richard J. Watts

Features of written and spoken language in Usenet newsgroup discourse

by

Matt Kimmich

Features of written and spoken language in Usenet newsgroup discourse

Table of Contents

1. Introduction 2

2. General considerations and definitions 3

2.1. Basic comparisons of written and spoken discourse 3

2.2. Relevant features of Usenet newsgroups 9

2.3. A brief description of RGCU-D (rec.games.computer.ultima-dragons) 11

3. Features of written and spoken language in newsgroup discourse 14

3.1. Formal features of newsgroup discourse 16

3.2. Discourse strategies in newsgroup conversations 25

4. Conclusion 31

5. Bibliography 33

Appendix 36

1. Results of polls on RGCU-D 36

2. ASCII (7-bit) characters 38

3. Some typical examples of signature files 38

1. Introduction⁽¹⁾

During the past few years the Internet has become increasingly accessible to the public as a means of international communication; buzzwords such as cyberspace or World Wide Web (WWW) are no longer known only to a selected group of students of computer sciences and readers of science fiction, and means of communication via the Internet are no longer restricted to large corporations, government organisations or universities.

Although detailed graphics and audio recordings of any kind can easily be transmitted over the Internet, most conversation over the net is still based on written text, mainly because visual and sound data is more difficult to produce, takes up more disk space (a word-processor file of an average 300-page paperback novel is smaller than most medium- or high-quality graphic files) and as a result takes more time to transmit as well as to receive. Text-based forms of Internet communication range from the IRC (Internet Relay Chat) which displays messages typed and sent by participants nearly in real-time, to e-mail and Usenet newsgroups that store messages which then are downloaded by the addressees at their convenience.

Although newsgroup discussion is basically a form of written communication it nevertheless displays many features that mirror spoken language. The aim of this paper is to discuss such features of written and spoken language found in Usenet newsgroup discourse. In order to do so, I will first compare mainly formal aspects of written and spoken discourse in general as well as give a brief description of Usenet newsgroups, concentrating on RGCU-D (rec.games.computer.ultima-dragons) which provides my data. Finally, I will discuss features of newsgroup discourse which are more common to either written or spoken language, features which imitate spoken language and features that apparently are common to neither. I will be using the collected data as evidence for my working hypotheses concerning both the communicative medium of Usenet and the validity of previous research on differences between written and spoken language. Many features traditionally attributed to spoken rather than written language can be found in Usenet newsgroup discourse, and Usenet functionally eliminates certain conversational conditions usually found in traditional forms of written communication - such as a substantial temporal distance between the discourse participants and, as a result, between the conversational turns, or as the often rather restricted group of participants. As a result, analysing such recently created means of written communication may shed new light on certain preconceived notions of inherent differences between written and spoken language.

2. General considerations and definitions

2.1. Basic comparisons of written and spoken discourse

In comparing written and spoken language, it is important to keep in mind that at least some of the traditional clear-cut distinctions between the two modes of discourse, as discussed below, no longer apply in all cases. This is mainly due to relatively recent technological advances; whereas writing in contrast to speech traditionally functioned as a way of permanently recording language, nowadays any kind of sound sequence can be recorded and replayed with devices such as a tape recorder. Furthermore, most forms of written communication take a comparably large amount of time to reach their respective addressees, whereas spoken communication usually is received practically immediately. This traditional distinction, like the previous one, is not inherent to the two modes of communication; for example, IRC channels transmit written messages almost instantaneously to all logged in on the respective channel, and telephone answering machines are specifically designed to record spoken messages that cannot be received immediately by the addressee. In the following paragraphs I will mainly focus on linguistic features generally attributed to either written or spoken discourse which are not directly or obviously connected to traditional rather than innate differences between the two modes.

A further misconception I will try to avoid in this paper is attributing levels of formality to either mode, as some researchers have explicitly or implicitly done,⁽²⁾ for example by comparing samples of written language taken from the highly formal end of the formal-informal continuum (e.g. scientific publications) to very informal samples of spoken language (e.g. leisure-time chat among friends or colleagues). In my opinion such a comparison is not very helpful or adequate as it presupposes a dichotomy between written and spoken discourse coinciding largely with the dichotomy between formal and informal discourse; this is stressed but in no way proven by using samples taken from two opposed kinds of discourse situations. Thankfully more recent researchers, such as Deborah Tannen, have abandoned this view of written language being inherently formal.

The basic formal distinction between written and spoken language is that whereas the former consists of "graphic substance" perceived visually, the latter is made up of "phonic substance" (Crystal 1987: 178-179), perceived aurally. This means that any spoken text is perceived by the audience primarily as a sequence in time, whereas a written text is seen as a sequence arranged spatially; from the moment a sample of writing has been produced, all parts of this sample are simultaneously accessible, for example over the space of a page or a computer screen, and as a result the text can generally be processed at a pace chosen by the addressee, whereas speech, even in recorded form, becomes distorted if it is replayed at a different pace from the one chosen by the addresser. It also appears that at least most people from western culture are more adapted to orienting themselves in spatially arranged recordings of language (i.e. written text) than in temporally arranged recordings (i.e. spoken text) as they will more readily thumb through a book and find a certain chapter or passage than they will find a passage in a recording of a conversation.

A further difference between the two modes lies in production and processing speeds. Whereas speech is produced by the speaker at the same speed as it is received by the audience, writing generally takes place at a much slower pace. Chafe puts typing at "about one-third the speed of speaking" (1982: 37), and it is likely that even trained typists on modern computer keyboards nowadays do not reach the average speed of speaking, whereas reading pace can reach more than twice that of speaking and listening; as Chafe puts it, "while speaking and listening necessarily proceed together at the same speed, writing and reading deviate from that spoken language baseline in opposite directions, writing being much slower and reading somewhat faster." This surely influences the way language is processed. Chafe assumes that so-called "idea units" appear in the text producer's mind on average every two seconds; writing thus would be too slow to reproduce these ideas in real-time. He then goes on to say:

As we write down one idea, our thoughts have plenty of time to move ahead to others. The result is that we have time to integrate a succession of ideas into a single linguistic whole in a way that is not available in speaking. [...] In writing we have time to mould a succession of ideas into a more complex, coherent, integrated whole, making use of devices we seldom use in speaking. (1982: 37)

In any case, most forms of oral communication are immediate, whereas practically all forms of written discourse are recorded. As a result, writing can quite easily be reread, analysed and changed by the addresser before it is transmitted to the addressee; as Crystal puts it, writing "promotes the development of careful organisation and more compact, intricately structured expression" (1987: 179), as it can be produced at a freely chosen pace in most cases.⁽³⁾ This, as well as the fact that a text is read faster than it is spoken, results in written conversational turns that are in general longer and cover more points than spoken turns, which often consist of single-sentence or even single-word responses. Also, as there usually is a considerable time span between written turns in mostly symmetric forms of written discourse (such as regular letter correspondence), participants will often paraphrase previous conversational moves on a topic in order to comment on them. In spoken discourse, the previous turn or turns are still in the participants' short-term memory and thus provide contextual effects that are important in generating meaning. Paraphrasing previous points in one's own words, however, can easily prompt different contextual effects since the paraphraser often selects and evaluates information by reformulating it. The only other way of providing the necessary conversational context is by quoting previous turns rather than paraphrasing them; this is the exception rather than the rule in most forms of written communication.

A reader can approach a written text with the same deliberateness as the writer, rereading single sentences and words as often as necessary to try to come to a thorough understanding of the text. Since written text in general can be reviewed, readers tend to expect the "careful organisation and more compact, intricately structured expression" mentioned above, especially when the text serves the purpose of conveying information rather than having a mainly phatic function, i.e. establishing and maintaining social contact. Speech, on the other hand, is usually seen as more spontaneous, featuring "looser construction, repetition, rephrasing, filler phrases" (Crystal 1987: 179). This is especially obvious to the addressee or addressees of any particular conversational turn; whereas written conversational turns are generally not transmitted as they have been produced - meaning that any corrections or hesitations are usually invisible to the reader -, listeners clearly hear exactly where speakers correct or repeat themselves in spoken conversation, or where they hesitate. Since this is standard in spoken communication, both speakers and hearers are tolerant to such self-correction in oral form. However, they are often surprised by how frequently these self-corrections, hesitations or other breaks in the steady flow of information actually occur, as can be seen by comments on accurate transcriptions of oral conversation.

Since oral communication usually is instantaneous and in many cases face-to-face, speakers will readily use deictic expressions that refer "to the speaker's [and/or listener's] position in space and time", as Crystal (1987: 106) puts it. Because at least the speakers, often the audience as well, share their surroundings at the time of the conversation, they can allude to them without becoming unclear or ambiguous. Written discourse, on the other hand, often takes place over a significant temporal and spatial distance; as a result, writer and reader most likely do not have shared physical surroundings they can refer to in deictic form. To prevent ambiguity, all information pertaining to the addresser's physical environment generally has to be lexicalised, i.e. put explicitly into words, to be understood by the reader.

This tendency to lexicalise information can be found in most forms of written communication; as mentioned above, writers lexicalise to prevent ambiguity which in spoken form would be prevented by intonational patterns, indicating for example the use of irony, or by facial expression expressing attitudes such as disbelief. Especially when it comes to forms of humour (e.g. sarcasm, irony) speakers can rely on their voice, face and body to convey their intended meaning, whereas writers making the same statement without signalling their intent - for example by lexicalising it - run the risk of their addressees taking the statement literally, thus misunderstanding it. Sometimes, however, such explicit signalling can defy the addresser's purpose because it can, for example, make a joke obvious when it was intended to be subtle. To create the effect in the reader that a signal such as a grin or a rising intonational pattern can achieve in spoken discourse, writers cannot simply mention their intention explicitly.

There are other features of oral discourse that cannot be directly emulated or imitated to the same effect by a written text. Some of these concern the human voice and its diversity of expression. Whereas speakers can raise as well as lower the volume of their voices, writers can only inadequately imitate this, for example by using capital letters or different size fonts; whereas anything spoken in a very quiet voice forces the listener to concentrate more and focus only on that voice, or loud speeches can cover any other voices or sounds, text written in small or large letters can only signal to the reader how loud the respective texts should be in spoken form rather than let the addressee experience all the side effects of varying voice volume. The same can be said for distinct/indistinct speech or intentional changes in voice pitch, all of which writers cannot adequately imitate, replace or lexicalise to the same effect. Only vocal emphasis can be emulated satisfactorily in written language by using typographic means such as italics or underlining of words.

In addition, as a speech sequence is set in time, speakers have the option to change their speaking pace or even to pause, both of which cannot be influenced by the listener. The same effect cannot be achieved in most forms of written discourse as they transmit a full message only after it has been completed; intra-turn pauses can thus only be imitated by gaps in the text which the reader can easily ignore and jump over. Reading pace is also chosen by the reader, and as a result any written signs imitating or signalling change of pace in a dialogue between actual participants⁽⁴⁾ will most likely have a different function, for example, be meant ironically.

Furthermore, spoken communication in general is two- (or more-)way in that all participants at any time can speak as well as hear all the other participants. This means that although within one conversation there is generally only one speaker at any given time, the others can give instantaneous feedback, both verbal and non-verbal, or anyone can interrupt the current speaker. Such instantaneous signals of audience involvement and participation can rarely be seen in even the most immediate forms of written communication such as the aforementioned IRC. The reason for this could be that spontaneous feedback such as backchanneling becomes more or less automatic in spoken discourse consisting of relatively indistinct sounds rather than words. In written forms they would have to be deliberately imitated by using letters. Also, in spoken discourse such feedback can signal how involved the audience is, whereas written forms of communication are actively sought out by the actual participants, and as a result it is very likely that as audience and potential participants they will both be at least mentally involved in the discourse.

As far as interruptions are concerned, they mostly take place within a conversational turn, forcing the current speaker to relinquish or defend his turn.⁽⁵⁾ Since most turns in written communication are only transmitted after they have been completed, it is impossible to interrupt them. In general, any feature of intra-turn interaction between participants in spoken discourse cannot be found in most forms of written discourse for that same reason.

There are also features that are restricted to written language, mainly related to writing consisting of "graphic substance". Written texts can be displayed in any number of distinct ways, using handwriting or different fonts, typographical features or different spatial structuring, and writers can use all of these specific features to signal additional meaning that cannot adequately be conveyed by speech. A special font can, for example, suggest different contexts in which the text might be read,⁽⁶⁾ a text can be split into paragraphs that signal text-internal transitions or changes, or a list can be presented as a table. In general, writers can display their texts according to their needs, even to the point of presenting the texts also as graphics; a poem that on purpose ends as it begins in order to suggest circularity can, for instance, be printed in the shape of a circle. This graphic aspect of writing can in some cases also imitate or emulate especially features of spoken language related to intonation, as is mentioned above. Furthermore, there are certain lexical items that are generally only found in written discourse, such as abbreviations of single words or phrases.

Stylistically, differences between spoken and written discourse are optional rather than inherent in either mode of communication. However, there are certain stylistic features that are common to written texts. Chafe (1982: 38-40) mentions, for example, nominalisations that, according to his data, occur more than ten times as often in writing than in speech; as Chafe uses formal as well as informal samples of both modes, this feature does not appear to be restricted to aspects of formality. He also states that attributive adjectives occur four times as often in written texts. Finally, there are certain words and phrases related to change of topic that are distinctly more common to writing than to speech, such as "however" or "although", since written communication traditionally treats several topics within one turn - for example, in one letter - whereas in spoken discourse participants tend to stay with one topic until it has in some way been brought to an end. As a result one is not likely to see in written discourse typical features of topic shift over the course of several conversational turns by different participants, such as often occur in speech.⁽⁷⁾

One last difference I would like to point out between spoken and written discourse is that most forms of symmetrical written communication have a restricted number of participants and thus are not as open to new participants as spoken discourse is. Written messages in general are directed only at predefined addressees who participate in the discourse, and they often have a predefined purpose and/or topic of communication.⁽⁸⁾ Relatively spontaneous random conversations were traditionally restricted to speech; the only traditional written medium for anything approaching this kind of conversation are Letters to the Editor, and the number of participants as well as the conversational output in this manner of discourse generally are very limited.

2.2. Relevant features of Usenet newsgroups

Basically, Usenet newsgroups can be compared to public bulletin boards; they are "areas where [messages] on a topic of common interest are posted and stored for anyone to read", according to Shea (1984: 65). Newsgroups - also known as discussion groups - distribute such public messages, called posts or news items, via a service of the Internet called Usenet:

Usenet can be considered a "virtual" network. News items travel over a wide variety of physical networks to reach computers running Usenet software throughout the world. [...]

Usenet has no management; it exists because individual Usenet sites propagate news items to other sites. Consequently, an item often passes through many intermediate sites to get from one part of Usenet to another.

Because of this multi-site propagation, the methods of processing news on individual hosts, and the sheer enormity of Usenet, it can take several days until an item you have posted has propagated over all of Usenet.⁽⁹⁾

Additionally, due to the sometimes erratic Usenet propagation a response to a news item sometimes arrives at certain sites before the news item it responds to, which can lead to some confusion among participants.

Usenet newsgroups are ordered hierarchically according to topics; on the topmost level there are for example the rec.* (recreational), de.* (German) or soc.* (social interaction) hierarchies which branch off into more specific newsgroups such as rec.humor.tv.x-files (a group dedicated to parodic material on the tv series The X-Files) or as alt.fan.pop.genesis (a group mainly consisting of discussion on the pop group Genesis). There are newsgroups that are mainly frequented by scientists discussing new research findings, just as there are groups dedicated to 'serious' discussions of alien abductions and multinational conspiracies; as with most of the Internet, Usenet is populated by all sorts of groups and individuals. Newsgroups widely differing in topic usually differ considerably in language style as well as in the average amount of posts produced.⁽¹⁰⁾ Newsgroup discourse functions along the following lines: Any participant can begin what is called a thread by posting a message. This message in most cases consists of a text written in 7-bit ASCII code which includes small and capital letters, numbers and certain common symbols,⁽¹¹⁾ and it contains the addresser's e-mail address, the newsgroup it is addressed to and a title, the so-called subject header. Furthermore, many posters attach so-called sig files, signatures or sigs to their messages. These contain additional information on the author of the post, and sometimes distinct quotes, messages or graphics that serve as a kind of personal letterhead, intended to convey interests or personality traits. The post is then sent to other newsservers, and users subscribed to that specific newsgroup can download all new posts at their discretion. If they choose to reply, their post will be sent as a subordinate message to the first. As all posts - beginnings of threads as well as replies - are organised hierarchically, the threads are stored on the news servers in a tree structure, according to the example given below:

Illustration 1⁽¹²⁾

The exact content of any post can be freely chosen; some newsgroups are censored by supervisors who filter out obscene or otherwise offensive material, but most newsgroups are self-regulating, meaning that other participants will send complaints either to the group of directly to the writer of the original post if they consider the contents unsuitable in any way.

When replying to a post, participants will generally quote relevant passages from the original text, indicating any cuts with "<snip>" or variations thereof; newsreader programs offer quoting as a standard option, usually marking quoted lines by beginning them with the symbol > and noting author and date of the quoted post. Especially in larger, more intricate discussions posts sometimes contain quotes from four or more previous news items.

2.3. A brief description of RGCU-D (rec.games.computer.ultima-dragons)

Since my point of focus in this paper is discourse in Usenet newsgroups, a medium that has not been studied very often by linguists so far, and as my goal can in no way be to give an exhaustive study of linguistic features of newsgroup discourse, it was the logical choice to gather data from a highly productive newsgroup frequented by a diversified group of participants. For this purpose I have chosen rec.games.computer.ultima-dragons (short: RGCU-D) to provide my samples, as I am familiar with both the group and ist regular contributors myself. As discourse on the newsgroup is formed by a considerable number of participants, it is unlikely that my participation - as a member of the group rather than as an independent observer - has greatly influenced or distorted the data. Also, since Usenet discourse is stored by all participants over a certain length of time, the other participants were not aware of the exact period of time that provided my data, thus keeping the observer's paradox to a minimum. Nevertheless, I asked for and received permission from a number of contributors to use their posts as sample material. In the following passages I will show which features of RGCU-D and its frequent participants make the group a useful starting point for analysing newsgroup discourse.

Rec.games.computer.ultima-dragons⁽¹³⁾ is the official newsgroup of the Ultima Dragons Internet Chapter, a fan club dedicated to Ultima, a series of fantasy role-playing computer games produced and published by Origin, Inc. Due to the complex gameplay and storytelling of the Ultima games, the fan club does not consist of the (stereo-)typical players of computer games, namely male teenagers.⁽¹⁴⁾ As can be seen in the appendix, average RGCU-D regulars appear to be in their mid-twenties, and women are fairly well represented, constituting one quarter of the newsgroup population.⁽¹⁵⁾ English is the dominant first language among posters, as can be expected, and a high number of participants professionally work with computers. This suggests that participants in general are not tentative or inhibited in any way about using computers as a means of communication, being well-versed in handling them. Furthermore, most posters state that their behaviour on RGCU-D is mainly informal and personal.

Although UDIC is an Ultima fan club, there has not been a new game in the series for more than three years now. As a result, the games are not the predominant topic of discussion on RGCU-D. As the group's founding charter states,

Topics of discussion in RGCUD have ranged from Ultima hints, wild speculations about Ultima 9, wild speculations about what events in previous Ultimas meant, Ultima Dragons organisational issues, software piracy, religion, and what various Dragons had for lunch yesterday. [...]⁽¹⁶⁾

In general, many of the group's members share certain common interests such as fantasy or science fiction, and club members assume a fictional dragon persona and name for most interaction on the group. Other than that, interests vary considerably; this means that most actual conversation centres on the discourse participants rather than any fixed topic, as is obvious from the data.⁽¹⁷⁾ Out of 348 threads in total, 43 generally centred on Ultima, computer-related technical questions or advertising posted on the group, and twelve were so-called 'fantasy threads', continuing, mostly parodic narratives created by and featuring the Dragon alter egos, whereas the other 293 (»84%) would have to be labelled personal threads.

On average, RGCU-D features high numbers of posts - 2427 posts during the week recorded, on average more than 346 per day - any intensive discussion; several threads produced more than a hundred posts and were branched up to the eighteenth level. (Very likely the threads continued past those numbers in the following week or weeks.) Altogether, there were 103 participants, 53 of them contributing more than five posts; during the recorded week 20% of the participants accounted for roughly 90% of the posts.

As UDIC has been established for several years now, many of the members contributing to the newsgroup have established relationships within the group, meaning that there are friendships as well as animosities, and the tone of certain posts can become rather harsh. Nevertheless, the interpersonal relationships created and maintained on RGCU-D make it a dynamic and changing context for a discourse tone that is best described as talkative or even "chatty". Also, the members have often proved to be defensive of the newsgroup itself, protecting it against charges and verbally attacking anyone offering unfounded criticism, so-called 'trolls'. Older members even refer to events in the past of the group if they find them noteworthy, such as the warding-off of persistent 'trolls' or especially heated discussions, usually called 'flame wars'.⁽¹⁸⁾ All in all, this emphasis of the newsgroup as a communicative context shows that participants do not consider it to be a merely 'virtual', thus non-existent or unimportant context. Heated discussions on a possible renaming of the group showed that participants practically consider RGCU-D to be equivalent to an actual physical environment.

3. Features of written and spoken language in newsgroup discourse

In this section I will first describe certain features of Usenet discourse on the word or sentence level that are usually attributed to either the written or the spoken mode, as well as features that imitate or emulate spoken language; 'imitate' meaning that such features are consciously used by the addresser to reproduce in writing items of oral discourse, 'emulate' meaning that such features rather are intended to achieve the same effect as certain items of oral discourse by other means. Then I will discuss some sample conversations, concentrating on the conversational features and strategies used, and whether they are usually attributed to written or spoken language.

In order to illustrate my points, I will use two different ways of quoting my sample material; when discussing features on the word or sentence level, I will excerpt only the relevant lines from complete posts. To represent longer stretches of conversation, I will use a different approach, as illustrated below.

The following three posts are taken from the thread "U6 Help" and refer to the sixth game in the Ultima series. Headers and signature files have already been removed as they present information that is not specific or relevant to this conversation.

Post 1 (Daniele⁽¹⁹⁾):

I have a problem that makes me not to dream!

Who is Michael really?

What have I to do and is It linked with the Quenton murder?

Post 2 (Kender⁽²⁰⁾'s response to Daniele):

Geom. Daniele Cipollini wrote:

> I have a problem that makes me not to dream!

> Who is Michael really?

He is an assassin who worked for the evil Mondain; he lives between Skara Brae and Britain and will not say his name. He killed Quenton.

> What have I to do and is It linked with the Quenton murder?

There is nothing you can do here - you cannot solve the murder. The programmers forgot to put the solution in. You only find out that Michael is the murderer when you play Ultima 7.

Post 3 (Underworld's reply to Kender):

Kender wrote in article <33C4F715.4921@gate.net>...

>> Who is Michael really?

>He is an assassin who worked for the evil Mondain; he lives between
>Skara Brae and Britain and will not say his name. He killed Quenton.

There is no evidence that Michael worked for Mondain; he never mentions any such thing. Quenton's wife was killed by a minion of Mondain, but not Michael.

In order to represent this stretch of conversation, I will insert the additions made by Kender and Underworld in the appropriate places; furthermore, I will delete the quoted lines as well as all strictly technical information such as the lines indicating the author of the quoted post. As a result, the conversation now looks like this:

Taken from "U6Help" 10/07/97 12:53

The numbers after the participants' names indicate whether the participants have posted other messages in the relevant stretch of conversation, and which of their posts the conversational move is taken from.

One problem with this method of representation is that the moves are not necessarily printed in the order they are produced, as a follow-up post can insert replies at any point, thus fragmenting the turn; however, I will later show that in newsgroup discourse the single strands of conversation are processed as in the example shown above. Kender's first move - her description of the assassin - is seen in the course of the conversation as logically and content-wise preceding Daniele's second move asking about Quenton's murder. In other words, it is not very reasonable to look at posts strictly in the order in which they are produced; they do not always add to the discourse linearly. The semantic and structural cohesion between lines from an initial post and lines from a reply is often more obvious than the cohesion between lines in one post dealing with different topics, such as the two questions in Daniele's post.

3.1. Formal features of newsgroup discourse

First of all, as Usenet discourse is written, there are certain features that are close to traditional forms of written communication, some of them mentioned above. For example, all information has to be lexicalised or represented using the available characters, meaning that certain information will be put in the form of graphics or symbols rather than words. The most obvious examples of this are the so-called emoticons or smilies that are intended to represent attitudes or emotions; I will discuss them in detail below. Also, as with many forms of written discourse, newsgroup participants tend to expect correct orthography and grammar from posters and display a conservative attitude towards standardised English. When a post contains opinions questionable to participants as well as incorrect language according to standards - whether the deviations from the standard are deliberate or not - they will often use the poster's 'incorrect' use of language as a subject of their criticism or insults.

In general, Usenet regulars are rather conservative about language and its traditions or standards; they will readily accept established abbreviations of phrases or sentences, whether they stem from traditional written media - such as "i.e.", "e.g." and "etc." - or from Usenet tradition - such as "IMO" for "in my opinion" or "AFAIK" for "as far as I know".⁽²¹⁾ However, many newsgroup participants will criticise or even insult posters using single-word abbreviations that are common on IRC, such as the figure "2" for the words "to" or "too", or the letter "R" for "are". Sometimes these posters will not even have their questions answered, as the example below illustrates. The same can often be said about posters who leave out punctuation marks or write everything in lower-case letters.

>there R 6 levers
>I have to pull which 2 open door?

<Vulcan reaches out of the monitor.>

SLAP!!

BTW - the words you're looking for are "are" and "to".

Another feature of newsgroup posts are signatures; as mentioned above, these are mainly used to personalise posts with graphics or quotes specific to the poster. They can be compared to letterheads in their function, although sigs tend to be more informal.⁽²²⁾ They are clearly a feature restricted to written language, and they are intended to make the posts look more personal, as post text is shown on-screen in standard fonts that give it an impersonal appearance. Thus, they could be said to compensate for personality features such as handwriting or even voice qualities that distinguish discourse participants from each other.

Another feature typical for the written mode which can be found in newsgroup discourse are long enumerative lists, for example, humorous top ten lists, which are rather common in Usenet messages. Such lists are rarely found in spoken discourse as they are difficult to memorise, and it generally takes a long time to create them and produce them as speech. Such lists, as an example of a relatively long and structured conversational turn, are thus better suited for the more permanent written mode.

On the other hand, one can also find many features in newsgroup discourse that, depending on interpretation, either imitate or emulate speech. Brown and Yule (1983: 19) quote from Goody's The Domestication of the Savage Mind (1972: 124): "Who, except the most obsessive academic, reads a book as he hears speech? Who except the most avant-garde of modern dramatists, attempts to write as they speak?" One answer to these two questions - meant to be rhetorical - is, at least to a certain extent, many newsgroup regulars. Of course certain of the features I am about to discuss are not wholly equivalent to their oral counterparts, especially as they are not transmitted and received as immediately as backchanneling features, for instance, and as they are more fixed and unequivocal than corresponding oral features. Nevertheless, some are arguably more than imitations of speech.

One of the features of speech that can hardly be reproduced to the same effect in writing are pauses. Since they are "temporary break[s] in the flow of speech" (Crystal 1992: 293), they contain the element of time that can hardly be reproduced on newsgroups on the level of seconds or split seconds. However, there are several types or functions of pauses, some of which can be imitated or even emulated in writing. Keseling (1992: 33) mentions the following five reasons for pauses:

1) psychological aspects such as personality, anxiety or emotions

2) hesitation phenomena that signal a sort of mental reorientation to the listener

3) breathing

4) the text producer's realisation of juncture

5) the possibility of transition points.

Obviously the first two of these might apply to the production of a written text, but will in most cases not influence the actual product as the reader sees it. Only the last two points can be approximated to a similar effect in newsgroup discourse. One example of this is passage breaks, common to most forms of writing; they could be described as "spatial breaks in the flow of text" and usually indicate transition or juncture points, often indicating a change or shit of topic. Another attempt to reproduce one possible effect of a pause, particular to Usenet and e-mail discourse, is the so-called spoiler space. This feature is always marked and explained, and it precedes textual information that some people might want to avoid, such as the punchline to a joke or solution to a puzzle, as in the following two illustrations taken from RGCU-D:

SPOILER SPACE
[26 blank lines deleted]
Go back to the Goblin Caves (if you've never been there, you shouldn't be worrying about the traitor yet) and look for a locked side passage

-----⁽²³⁾

.spoiler alert
[10 blank lines deleted]
It's the Mirror of Truth that you need, which you get back in Moonshade.

Spoiler spaces have the same function as the type of pause that is intended to give an audience the choice not to listen, also because the following information might be considered a kind of 'spoiler'. However, there is a difference between the written and spoken variant in the audience's reaction. In speech, a listener actively has to avoid hearing the next conversational move, whereas in writing the reader actively has to proceed to the next move. The general purpose, however, is the same.

There also are other instances of posters trying to represent pauses, as in the following example:

Ta-dum!

[Frenetic Dragon pauses momentarily, eyeing for signs of incoming cream
pies from the food fighting Dragons.]

In this passage, the author of the post switches to a narrative form more common to writing, indicated by the brackets, the use of a third person singular, and the imaginary setting. The passage is written as fictional narration, a deliberate instance of make-believe. In this mode, the use of the verb "pauses" does not have the same effect as a pause itself, nor is it intended to.

The next example is similar:

(deep breath) THAT IS WHAT WE ARE DOING!!!!!!!!!!!!!!!!!!!!!!!

(pause)

Everyone on this group is attacking the Christian Church's policies of
proselytization, exclusivity, oppression, etc etc ad infinitum. In

The author of this passage uses the word "pause" - as well as the "deep breath" - in a way that can be compared to actor directions in a play, which is indicated by the brackets. Clearly, she uses the two directions to show how she would vocalise her command, implying that she indeed might write as she speaks. Additionally, the use of capital letters traditionally represents shouting - another oral feature - in newsgroup discourse, as Angell and Heslop (1994: 11) mention. However, both 'dramatic' directions are conscious, deliberate and non-traditional in the Usenet medium, and as a result will not have an effect identical to the actions they represent. The pause in speech has its effect in real-time, for instance, meaning that the audience is left guessing whether or not the speaker will calm down after her outburst. Also, the twenty-two exclamation marks can only suggest how loud the exclamation was intended to be rather than convey any exact information.

One feature that appears to emulate speech in newsgroup discourse is the use of hesitators. Although many of them are difficult to reproduce adequately in normal 26-letter writing, using phonetic or phonological features such as the sound represented by [schwa symbol] or elongations, there are often conventional representations in that their orthography has been established in written representation of speech, such as "Erm." or "Er." When they occur in newsgroup posts, they are often in positions equivalent to their oral versions, as the following quotes illustrate:

>Where does the creative song come from? What reactions to stimuli led
>to that made you type the exact words(not some other words) in the
>above post?

Erm... I fail to see your problem. Creativity comes from the same place
Reason comes from -- your brain. You seem to place some special [...]

-----

>troubles as the Babylon 5 TV series. They've both defeated the 'Big
>Bad Guys' and now they don't know what to do.

Er.. maybe you don't know this, but Babylon 5 knows *exactly* what it's
Going to do. The whole basic five year story was scripted out in advance [...]

-----

> I'm very sorry that you are so lacking in imaginative abilities
> Dave... that's really sad.

No, what is sad is that you didn't even bother to refute anything I've
said. Er... grownups call that "debate", Silverlock.

As with some types of hesitators in speech, the ones in the examples above stand just before a statement that disagrees with what has been mentioned before. In the first two passages, the hesitators imply that the author of the reply does have some reservations about disagreeing - possibly stemming from politeness, doubt or simply surprise about the other's intentions or meaning. The author quoted first also expresses these reservations by continuing, "I fail to see your problems", implying that he might be the one at fault here, while the second one weakens his disagreeing statement by writing, "maybe you don't know this". In the third example the hesitator is not used to weaken a disagreement; after all, the author of the passage is replying to an insult in kind. Instead, the "Er..." followed by the statement about "grownups" - thus suggesting that Silverlock does not belong to this group - implies that the author hesitates to convey his doubts that the other participant will understand his statement. The intent behind the hesitators in all three cases corresponds closely to the intent behind similar hesitators in spoken language when they are used as discourse markers.

One feature traditionally used in newsgroup discourse is bracketing a - generally short - phrase or single words with asterisks in order to emphasise them. This corresponds very closely to intonational emphasis in that it carries no unequivocal meaning or implicature, as the passages below show:

>[...] if you want to discuss it, think about it scientifically...

Oh yes. Please can we *scientifically* discuss the possibility of
Alien life forms so that we can *sound* intellectual while still
Maintaining the flavor of yet another bull shooting session.

-----

> > That is irrelevant. The point is that none of the countries labeled
> >communist by westerners actually called themselves communist.

> What about China, the name of the government party their is the
> Communist party of China.

China has a centralized government.
By definition, a communist country cannot have a centralized government.
Therefore, China is not a communist country.

They are *lying*.

-----

Christianity is pretty tame. Open up the Koran sometime, and read where
it mentions Christians and Jews *by name*, and says "don't associate with
these people! They are your enemy!". [...]

The author of the first passage, in response to the previous post, obviously adapts an ironic style, indicated for instance by the initial "Oh yes" and the "Please". This is enhanced by stressing the "scientifically" as well as the "sound" in "so that we can sound intellectual"; especially the second emphasis implies that the discussion participants might sound intellectual without actually being intellectual or discussing intellectually, which contradicts the "scientifically" to a certain degree. The two emphases are thus juxtaposed to convey the author's ironic intentions.

In the second passage, the emphasis markers are used in a more straightforward manner, simply stressing the word that summarises the statement and interprets its content, thus bringing it to a point; whereas in the third paragraph, the author stresses one part of his instructions to make them as unequivocal as possible.

Although in certain types of written discourse features such as underlining a passage or putting it in italics emphasise this passage, these features are traditionally used for very specific types of emphasis, such as indicating titles or marking technical terms. Furthermore, they are often applied to whole sentences, whereas asterisks are in effect mostly used for single words or short phrases, as mentioned above, perhaps because they only mark the beginning and end of the stressed text rather than the whole passage, which is easier for readers to process. In any case, they are used very frequently in newsgroup discourse, and the text they stress, as in examples two and three, would generally be unequivocal even without the added emphasis markers. As a result it seems that the use of emphasis in Usenet resembles speech emphasis more closely than written emphasis.

Similarly to other types and media of communication, Usenet also develops discourse traditions and strategies of its own; the most prominent one being emoticons, which I will discuss below. One of the more frequent traditions stems from the fact that many Usenet users tend to be computer-literate, and it is clearly a feature that employs written strategies rather than oral ones. It is taken from computer codes such as the aforementioned HTML which assign a certain attribute to a section by beginning it with "<[attribute]" in order to activate the attribute, and ending it with "</[attribute]>" to deactivate it. The following quotes illustrate its use:

<typical American>
(Not sure why we'd want to conquer Canada anyway, it's so *small* ...)
</typical American>

-----

<HOMER>
It's funny 'cause it's true.
</HOMER>

In both cases, the attributes indicate vocal qualities or a certain conversational style; they can also be used to set a certain scene. In general, their function can be compared to quotation marks, in that they relativise a statement, showing that the author quotes another person, or assumes a fictional character. In other words, by assuming and signalling a stance different from his or her usual stance, the poster relinquishes some responsibility for the argument. Interestingly, this feature is similar to one relatively new feature found in oral discourse, namely marking a passage of speech by gesturing "quotation marks" in order to distance oneself from the utterance in some way.

One of the better known features of newsgroup discourse is the so-called emoticons or smileys. They exist in many varieties, most of which have distinct functions, and they normally represent either emotional states and moods or facial expressions. The most common emoticon is the basic smiley, representing a smiling face: :) or :-).⁽²⁴⁾ The following examples illustrate its use:

Actually, I believe that one of Mars' polar ice caps is frozen water,
and the other is frozen CO2. I do hope that that author did their
research now, or I'll feel silly... :) [...]

-----

As for me, I just vote for a party I know haven't a chance of being
Elected. :)

-----

>And, of course, there is only one reason to watch a Spice Girls video,
>and quite frankly the video's a lot better with the sound *off*! >:D

:)

In all three cases, the producers of the smileys indicate their general amusement, either with their own statement or with the previous one. Especially the two first authors use the emoticon to indicate their attitude to what they have written, making clear that they do not take the discussion overly serious, whereas the third smiley obviously expresses a reaction to the previous post. It is interesting that the entire turn consists only of the emoticon - suggesting that the author considers his reaction to be relevant to the original poster and the group. As a result, it appears to have a function similar to backchanneling in that it supports the previous author and conveys a reaction or mood the same way a smile would in spoken, face-to-face conversation. Such a minimal reaction will most probably not be found in traditional forms of written discourse.

One other emoticon frequently used in Usenet is ;) or ;-), signifying the combination of smiling and winking. It is used in the following ways:

Now clearly the high pressure this water would be under and the proximity
to the thermal would make its bottom section tremendously hot, but the
top of the liquid bubble might be at a more inhabitable temperature. Who
is to say that from the vastness of Jupiter's atmosphere, the right
chemicals for carbon-based life haven't dissolved in the water...? Just a
thought.

BTW, I haven't studied physics for far too long. Can you tell? ;)

-----

Yeah, witch hunts, I'm sure that'll be fun. Why they should be starting
any minute now, what with all those evil fundamentalists restricting free
speech and all. Let's see, where are they, they should be right around...
oops. Can't find any. Let's see, maybe hiding under the bed... nope, not
there. Behind the sofa? No... darn, everywhere I look, I just see
reasonable people. Those DARN fundamentalists! They keep spitefully
supporting free speech just to fool me! ;-)

-----

You all drink beer, wrestle gators, and shag roos. For those in NZ
Substitute sheep for roos. You also have accents about as wacky as the
Texan accent. Yup. ;-)

The passages quoted above all contain irony in that their intended meaning differs from their literal meaning. The author of the first passage, for instance, formulates several concepts in idioms taken from physics and then comments on his own statement by writing "[By the way], I haven't studied physics for far too long. Can you tell?" and signalling irony. The second and third passage both use the emoticon to the same effect, reinforcing the ironic intention. However, in all three cases there are other irony markers as well; in the second passage it is for example the emphasised "DARN" in combination with "fundamentalists" as well as the exaggeratedly naive tone of the previous sentences that make the poster's irony obvious, and in the third post it is the mention of stereotypes followed by the laconic confirmation "Yup." Thus, the emoticons are not necessary in order to express unequivocally the author's ironic intention; they are not merely lexicalising irony in order to make the posts unambiguous, as could be assumed from literature on written discourse. This, as well as the frequency of ;) or ;-) in newsgroup posts, suggests that the feature has a mainly phatic function and is intended to convey an emotional response. Its function is thus similar to the corresponding oral or metalingual features.

I will briefly mention a few other types of emoticon in order to show their variety and function:

Yes. Right at the beginning there was also Zak's tender relationship
With his goldfish: "Sorry, Sushi, but you're fertiliser now." >:)

-----

Hey, there's a lot to be said for older women.

Mostly, "Nice cellulite!" >:)

In both cases the >:) - sometimes also >:-) - indicates mischievous humour such as sarcasm or cynicism. However, similar to ;) it does not add to or clarify the previous statement; instead, it simply signals a reaction, similar to its corresponding facial expression.

My mom doesn't understand why I don't watch Babylon 5 (she loves it).
I'd like to, but back a few months after it started, there was a period when I was unable to watch it for a while, and just lost the whole plot
Thread. Now I wouldn't know what's going on. :(

-----

BTW, is it just me, or has anyone else had trouble keeping from
colliding with other fighters in dogfights? It's the way I get damaged
the most. :(

The :( or :-( basically expresses the opposite of :); it is sometimes called frownie. It indicates that the poster is unhappy about what he or she has just stated, or that he dislikes it. Again, in most instances in which this emoticon is used, the writer's intention is already clear without the additional symbol, which suggests that it has a similar function in the conversational turn to a frown or an unhappy intonational pattern or speech rhythm in oral discourse.

Corporal punishment ended in my family when my father slapped me and I
promptly punched him in the kidneys. We're not a closely knit family.

:P

-----

I get that in reverse all the time. Even when I'm in "pretend to be
femme for relatives" mode, which really pisses me off ... :P

The emoticon :P or :-P represents someone sticking his or her tongue out. This can signify both disgust or cheekiness, similar to sticking out one's tongue in face-to-face conversation, and it generally expresses a mixture of humour and dislike. As with the previous emoticons, it emphasises or reinforces meaning rather than adding to it; again, it emulates the function of the corresponding feature of face-to-face discourse.

There are also other types, or variations, of emoticon; the ones illustrated above are the most typical and traditional ones, and as a result they also fall into a certain tradition of use. As mentioned, they are used regularly in discourse and usually approximate the function of corresponding oral or metalingual features, meaning that they are emulations rather than imitations of those features.

3.2. Discourse strategies in newsgroup conversations

In this subsection I will quote and discuss two excerpts from threads on RGCU-D, both of which illustrate the use of oral and literate strategies in the written Usenet medium. However, one problem with this approach is that literature on the topic mainly discusses formal or stylistic differences between the two modes, whereas discourse strategies are rarely compared; discourse strategies meaning the "functionally related moves [through which] participants try to realize optimally their purposes or goals" (van Dijk 1985: 4).⁽²⁵⁾ For this reason, some of the following assumptions on strategies traditionally attributed to written or spoken discourse are based on private experiences and data rather than representative analyses.

The first conversation is taken from a thread on the eighth and most disputed game in the Ultima series. The thread was named "Why is U8 considered a bad game?":

In line 2 - his first post - Ophidian begins to enumerate contradictions to the statements made in the previous post. As mentioned above, enumerative lists are a literate⁽²⁷⁾ discourse strategy as it takes effort and time to produce them, and they are likely to be interrupted in oral conversations. In his reply, Vulcan contradicts him and asks him a direct question which Ophidian answers; the questions in lines 4/5 and the reply in line 6 make up an adjancency pair that is common in spoken discourse, as traditional written conversations are made up of longer turns containing several different moves and thus usually do not contain typical adjacency pairs⁽²⁸⁾ such as question-reply, request-permission, etc. In replying, Ophidian also uses a discourse strategy common to speech in that his reply is highly context-dependent; he does not state what he remembers, and he refers to the dagger he stole in the game as "it". In order for the audience to understand that exchange, they have to be aware of the previous turns, which in traditional written media usually is achieved by the writer oaraphrasing what he or she considers to be necessary context. In Ophidian's second post he quotes Vulcan; thus, the immediate discourse context is repeated, not reformulated. However, only by fragmenting Vulcan's previous turn does Ophidian create the adjacency pair. Lines 7, 8 and 9 illustrate the same strategies. The pronouns referring to words in the other's move serve as strong cohesive links between the moves; as mentioned in the introduction to this section, structural and semantic cohesion is thus often stronger between adjoining moves from different posts than within a single replying post. Furthermore, it is noteworthy that the exchange consists mostly of main clauses, which also is a feature more common to speech.

In line 11, Vulcan replies to the third item on Ophidian's original list of points, utilising another oral strategy; he begins with a feature Deborah Schiffrin (1985b: 43) refers to as a "ritual display of cooperation",acknowledging the other's opinion but immediately questioning it ("Ah, but [...]"). Again, as can be seen throughout the conversation, subsequent moves refer to previous ones, as is typical in oral conversation, mainly by using pronouns that semantically connect to previous moves. However, in lines 13/14, Vulcan unequivocally gives one of his phrases a lower priority than the others by putting it into brackets, which is clearly a feature particular to writing. Similar oral strategies - varying volume, for example - are not as distinct in their effect. Ophidian, in his reply (line 15), derives an implicature from Vulcan's statement without explicitly stating how he arrives at this implicature; this is also a strategy that only functions if the exact context is given, as in immediate speech exchanges. Vulcan then attempts to invalidate Ophidian's analogy. Strictly speaking, his first sentence, "Entirely different situation.", is syntactically incomplete. Such syntactic minimal structures (also the one in line 11 - "The Avatar?") are common to speech rather than writing as most written media traditionally adhere to standards and conventions of orthography and grammar. As mentioned earlier, by using asterisks in order to emphasise single words, Vulcan emulates stress in speech, and the paragraph break between lines 19 and 20 could be compared to a short pause in oral discourse. Lines 23-27 feature discourse strategies mentioned before, such as brackets or minimal reference and semantic connections to previous moves. Since this specific dialogue is mainly an argument, the transitions from 27 to 28 and from 31 to 32 are interesting; in both of them, Ophidian does not counter Vulcan's argument and even seems to ignore them, although he obviously replies to other statements made in the same post (Vulcan 2). In most written traditions, this would not be unusual, as conversational turns are far enough from one another for other participants not to remember exactly what previous moves there were. In Usenet discourse, however, the immediate conversational context is available; as a result, Ophidian's lack of reaction will have an effect on Vulcan and the other newsgroup participants.

A further discourse strategy traditionally considered to be oral is the repetition of "[the Guardian threatening] to destroy the universe" (line 28) in Vulcan's reply in line 30. Deborah Tannen writes:

Repeating the words, phrases, or sentences of other speakers (a) accomplishes a conversation, (b) shows one's response to another utterance, (c) shows acceptance of other's utterances, their participation and them, and (d) gives evidence of one's own participation. [...] In terms of mutual participation in sensemaking, each time a word or phrase is repeated, its meaning is altered. The audience reinterprets the meaning of the word or phrase in light of the accretion, juxtaposition, or expansion. (1989: 52)

Repetition of "words, phrases, or sentences" in traditional written media would not have the same effect as previous turns are not immediately accessible; in other words, by the time addressees receive a reply, they most likely will not remember the exact words or phrases of the previous turn, and as a result might not recognise the repetition.

In line 34 Vulcan again uses a minimal syntactic structure. Interestingly he refers to his opinion by writing "I'm just saying [...]", implying that - to a certain degree - he equates his written statements to utterances, something that is common in newsgroup discourse.⁽²⁹⁾

Altogether, we can state that both Ophidian and Vulcan in their turns rely heavily on the conversational context provided by quoting previous moves. In general, exact quoting is possible in all written media; however, it is more common that authors of written turns paraphrase necessary context, thus changing this context. Both participants in the dialogue above refer to previous moves by using pronouns, thus linking their own moves to the previous ones as is done in speech. The same effect is achieved by creating adjacency pairs such as lines 4/5 and 6, or 15/16 and 17. Main written strategies applied by the two participants are Ophidian's use of a numbered list and Vulcan's use of brackets.

The next conversation is taken from a fantasy thread named "[SILLY] Daermonestroer Dragon Is The Anti-Avatar!",⁽³⁰⁾ and it illustrates how easily some participants switch from clearly written strategies to oral strategies between turns, sometimes between moves:

Throughout his initial post, Lost adapts style and strategies of written narration; he uses the third person singular to refer to his fictional character, and his sentence structures and choice of words are more usual in writing. Also, he demonstrates that he is aware of the written medium, as he indicates uncertainty about his spelling of "Zarathustra" in line 1, indicated by the "(sp)".⁽³¹⁾ Samurai, on the other hand, imitates an oral strategy in that he interrupts or 'fragments' Lost's first sentence in order to give immediate verbal and non-verbal feedback. In response to the joking criticism provided by Samurai - "Showoff! ;)" - Lost digresses from his main narrative by telling an explanatory story, this time using oral narrative strategies such as first person narrator and stress on his emotional attitude towards the story.⁽³²⁾ However, even in this brief tale he switches to third person narrative for a brief interjection that is more closely connected to the main narrative in content (cf. lines 16 and 22-25).

Samurai's second reply to the main narrative (line 11) is again a commentary on the story. Also, it refers to another discussion on the framework of the newsgroup; Samurai's use of the phrase "elsewhere in the NG" implies that he treats RGCU-D as a physical conversational context,⁽³³⁾ whereas traditional written discourse is not usually set in such a shared physical (albeit imagined) context. This time, Lost answers to Samurai's commentary by extending his main narrative, again according to the written narrative strategies mentioned above, which is taken up in lines 14 and 15. Both of Samurai's subsequent moves in lines 17 and 20/21 are commentaries in the form of quotes that refer to the previous moves; only in lines 22 and 23 does he switch back to mainly oral strategies, this time commenting on the main narrative as a whole and reacting to it by imitating direct speech from face-to-face conversation. The "Oh, BTW, Lost?" is intended to get the other's attention - within the fictional framework, as Lost cannot really react to it - and the "*SPLUT*" is the in-group term used for signifying a thrown custard pie. Differently from Lost, Samurai builds on the previous narrative - which used written strategies - in a style more common to oral discourse.

This passage illustrates that participants in newsgroup discourse can and will switch between traditionally oral discourse strategies and written discourse strategies according to personal style; however, they obviously are not restricted by the fact that the Usenet medium is based on writing.

Of course both dialogue passages discussed above cannot representatively show whether strategies from the written or spoken mode are predominant in newsgroup discourse; however, they clearly indicate that both kinds of strategies are used, and that conversations on Usenet in many ways more closely resemble oral conversations than traditional types of written communication such as exchange of letters, however personal.

4. Conclusion

Although the discussions in the previous sections of this paper are too limited in their scope to be seen as representative, as I have stressed above, my data suggests several working hypotheses that should be followed up by research, the first one being that the Usenet tradition of quoting relevant passages from previous posts results in the discourse participants accessing and using that conversational context similarly to speakers in oral discourse; the relative temporal immediacy of speech is translated into spatial immediacy, providing the same context.⁽³⁴⁾ In traditional written media, on the other hand, writers give the context of previous turns by paraphrasing these turns, which can result in different contextual effects.⁽³⁵⁾ The spatial immediacy of Usenet is reinforced by the fact that newsgroup exchanges in general are significantly more frequent than in other written media.⁽³⁶⁾

Secondly, the fact that newsgroup regulars easily switch between what is traditionally perceived as 'oral' and 'literate' discourse strategies and styles suggests that many preconceived notions about written and spoken language - as they are often found in literature on the subject - might have to be abandoned in favour of a more valid system. It seems to me that discourse strategies are applied to what could be called specific discourse situations, such as chat, scientific discussion, etc. Although those discourse situations in some cases might be more conducive to one of the two modes - writing or speech - they are not necessarily restricted to either the one or the other; for instance, newsgroup conversation can just as well be described as 'chat' as conversations in the local pub, in style, content, formality and other linguistic features. So, instead of assigning these discourse situations a position in a not very appropriate continuum between speech and writing,⁽³⁷⁾ I consider it to be more valid to position the aforementioned discourse situations along continua such as:

Fast Production -- Slow production

Immediate exchange -- Not immediate

Informal/Personal -- Formal/Impersonal

Phatic communication -- Communication as exchange of information

Although traditional written media often tend to be situated on the right-hand side of these continua, new mediums such as Usenet, IRC or e-mail show that attributes such as formality or non-immediacy are in no way inherent to the written mode. Articles like "Some of My Favorite Writers are Literate: The Mingling of Oral and Literate Strategies in Written Communication" by Robin Lakoff (1982) suggest that some researchers simply use preconceived notions of differences between written and spoken mode as a basis for hypotheses such as the following: "[T]here is evidence all around us that as a culture we are contemplating - if we have not taken already - a leap from being written-oriented to spoken-oriented. Our stylistic preferences naturally are shifting along with our values [...]" (1982: 256). This implies a direct link between choice of mode and style, which is clearly refuted by newsgroup participants who readily switch between traditionally oral and literate strategies, as shown above.

As new media such as the Internet remove many traditional functional constraints on written discourse, they should be studies further in order to allow us to arrive at valid conclusions on differences between the written and the spoken mode. Also, researchers should consider examining not only the product of writing but also the process of producing text,⁽³⁸⁾ as is done in oral discourse analysis. Especially highly productive media such as newsgroups justify an approach in which text production is considered to be a dynamic process just like speech production.

5. Bibliography

Angell, David and Brent Heslop
1994 The Elements of E-Mail Style. Massachusetts: Addison-Wesley Publishing Company.

Beaman, Karen
1984 "Coordination and subordination revisited: Syntactic complexity in spoken and written narrative discourse." In Deborah Tannen (ed.), Coherence in Spoken and Written Discourse. New Jersey: Ablex Publishing.

Brown, Gillian and George Yule
1983 Discourse analysis. Cambridge: Cambridge University Press.

Chafe, Wallace L.
1982 "Integration and involvement in speaking, writing, and oral literature." In Deborah Tannen (ed.), Spoken and Written Language. New Jersey: Ablex.

1984 Linguistic differences produced by differences between speaking and writing. In David R. Olson, Nancy Torrance and Angela Hildyard (eds.), Literacy, Language, and Learning. The nature and consequences of reading and writing. Cambridge: Cambridge University Press.

Crystal, David
1987 The Cambridge Encyclopedia of Language. Cambridge: Cambridge University Press.

1992 An Encyclopedic Dictionary of Language and Languages. Oxford: Blackwell Publishers.

De Beaugrande, Robert
1985 Text Production. New Jersey: Ablex Publishing.

Goody, J.
1977 The Domestication of the Savage Mind. Cambridge: Cambridge University Press.

Gumperz, John J.
1982 Discourse Strategies. Cambridge: Cambridge University Press.

Keseling, Gisbert
1992 "Pause and intonation contours in written and oral discourse." In Dieter Stein (ed.), Cooperating with Written Texts. The Pragmatics and Comprehension of Written Texts. New York: Mouton de Gruyter.

Lakoff, Robin
1982 "Some of My Favorite Writers are Literate: The mingling of oral and literate strategies in written communication." In Deborah Tannen (ed.), Coherence in Spoken and Written Discourse. New Jersey: Ablex Publishing.

Oreström, Bengt
1983 Turn-Taking in English Conversation. Malmö: CWK Gleerup.

Shea, Virginia
1993 NETiquette. San Francisco: Albion Books.

Schiffrin, Deborah
1984 "Everyday argument: the organisation of diversity in talk." In Teun A. van Dijk, Handbook of Discourse Analysis. Volume 3. Discourse and Dialogue. London: Academic Press.

Sperber, Dan and Deirdre Wilson
1985 Relevance. Communication and Cognition. Oxford: Blackwell Publishers.

Tannen, Deborah
1980 "A comparative analysis of oral narrative strategies." In Wallace L. Chafe (ed.), The Pear stories: Cognitive, cultural and linguistic aspects of narrative production. Norwood: Ablex Publishing.

1983 "Spoken and written narrative in English and Greek." In Deborah Tannen, Coherence in Spoken and Written Discourse. New Jersey: Ablex Publishing.

1988 Talking Voices. Cambridge: Cambridge University Press.

Van Dijk, Teun A.
1984 "Introduction: Dialogue as discourse and interaction." In Teun A. van Dijk, Handbook of Discourse Analysis. Volume 3. Discourse and Dialogue. London, Academic Press.

Appendix

Results of polls on RGCU-D

Do people revise their posts before sending them?

Yes: 14 No: 6 Depends/unsure: 4 Total: 24

à hypothesis: it is unlikely that the 'spontaneous' attitude of most posts comes about because they are quickly written and sent; in this (not necessarily representative) poll, 58.3% of posters regularly revise their posts, 25% of posters never/almost never revise, and 16.7% revise depending on what kind of post it is.

How old are the Dragons, what gender are they, where do they live, and what is their first language?

Dragon name	Age	Gender>	Residence	First language
Angelina	23	female	USA	English
BlackMage	34	male	USA	English
Daermonestroer	23	male	USA	Russian
Destard	21	male	Romania	Romanian
Disoriented	50	female	USA	English
DragonSpawn	25	male	New Zealand	English
Dreamworld	22	male	Canada	English
Earendil	36	female	Netherlands	Dutch
Erratic	22	male	New Zealand	English
Erynnes	41	female	USA	English
Exanter	21	male	USA	English
Faramir	16	female	South Africa	English
Fortran	33	male	USA	English
Gazza	19	male	USA	English
Granny Weatherwax	25	male	Austria	German
Indecisive	19	male	Canada	English
Kender	21	female	USA	English
Kilted Ghost	23	male	USA	English
Kyote	21	male	USA	English
Lost	23	male	USA	English
Min Joon	42	female	USA	English
MVD	23	male	Germany	German
Ophidian	15	male	USA	English
Origin-al	20	male	USA	English
Polychromic	33	male	USA	English
Red Ferret	19	male	England	English
Reknaw	19	male	USA	English
Samurai	21	male	England	English
Shadow Wraith	16	male	USA	English
Schole	23	male	USA	English
Silly	23	male	Canada	English
Telemachus	25	male	USA	English
Tenacious	17	male	USA	English
Torpid	62	female	England	English
Unicorn	43	female	Canada	English
Vulcan	28	male	USA	English

36 Dragons listed

Relevant statistics	Average age in years	Gender	L1
	26.3±10.4(39)	75% male/25% female	86% English

Do people frequent other newsgroups, do they behave differently on them, and do they professionally use computers?

Dragon name	Other NGs?	Behaviour on other NGs	Professional use of computers?
Daermonestroer	yes	Less discussion	yes
Disoriented	yes	More serious	no
Dragonspawn	yes	Less personal	yes
Dreamworld	no	---	yes
Erratic	yes	More timid	yes
Exanter	yes	Less personal	yes
Faramir	yes	Same	no
Fortran	no	---	yes
Gazza	no	---	no
Granny Weatherwax	yes	Same	yes
Joe Hendrix	yes	Same	yes
Kilted Ghost	yes	Same	no
Lost	yes	Same	yes
Min Joon	yes	Same	no
MVD	yes	Same	yes
Octal	yes	Less personal/more professional	yes
Ophidian	no	---	no
Pesslock	yes	Same	no
Polychromic	yes	Same	yes
Reknaw	no	---	no
Samurai	yes	Same	no
Telemachus	no	---	no
Tenacious	yes	Same	yes
Unicorn	yes	Same	yes
Vulcan	yes	Same	yes

25 Dragons listed

à hypothesis: For one thing, the relatively high number of people who professionally use computers (56%) suggests that they are used to working with computers as tools; furthermore, the answers in column three show that it is unlikely posters will generally use an impersonal style of conversation on the newsgroup. Finally, the fact that most of the people who answered to the polls on RGCU-D were - at that time, at least - regular posters suggests that they are at ease with communicating over the computer.

ASCII (7-bit) characters⁽⁴⁰⁾

Non printing characters, including the space character, are given their designation in multiple characters. Thus SP is the space character. On different display screens, some of the non-printing characters may give special display symbols. The character numerical values are given in decimal / hexadecimal

Column (first hexadecimal "digit" of character byte - decimal value)

0 1 2 3 4 5 6 7

Row Number

Dec Hex

0 0 00/00 NUL 16/10 DLE 32/20 48/30 0 64/40 @ 80/50 P 96/60 ' 112/70 p
1 1 01/01 SOH 17/11 DC1 33/21 ! 49/31 1 65/41 A 81/51 Q 97/61 a 113/71 q
2 2 02/02 STX 18/12 DC2 34/22 " 50/32 2 66/42 B 82/52 R 98/62 b 114/72 r
3 3 03/03 ETX 19/13 DC3 35/23 # 51/33 3 67/43 C 83/53 S 99/63 c 115/73 s
4 4 04/04 EOT 20/14 DC4 36/24 $ 52/34 4 68/44 D 84/54 T 100/64 d 116/74 t
5 5 05/05 ENQ 21/15 NAK 37/25 % 53/35 5 69/45 E 85/55 U 101/65 e 117/75 u
6 6 06/06 ACK 22/16 SYN 38/26 & 54/36 6 70/46 F 86/56 V 102/66 f 118/76 v
7 7 07/07 BEL 23/17 ETB 39/27 ' 55/37 7 71/47 G 87/57 W 103/67 g 119/77 w
8 8 08/08 BS 24/18 CAN 40/28 ( 56/38 8 72/48 H 88/58 X 104/68 h 120/78 x
9 9 09/09 HT 25/19 EM 41/29 ) 57/39 9 73/49 I 89/59 Y 105/69 i 121/79 y
10 A 10/0A LF 26/1A SUB 42/2A * 58/3A : 74/4A J 89/5A Z 106/6A j 122/7A z
11 B 11/0B VT 27/1B ESC 43/2B + 59/3B ; 75/4B K 90/5B [ 107/6B k 123/7B {
12 C 12/0C FF 28/1C FS 44/2C , 60/3C < 76/4C L 91/5C \ 108/6C l 124/7C |
13 D 13/0D CR 29/1D GS 45/2D - 61/3D = 77/4D M 92/5D ] 109/6D m 125/7D }
14 E 14/0E LS1 30/1E RS 46/2E . 62/3E > 78/4E N 93/5E ^ 110/6E n 126/7E ~
15 F 15/0F LS0 31/1F US 47/2F / 63/3F ? 79/4F O 94/5F _ 111/6F o 127/7F DEL

Some typical examples of signature files

/| .oo__. .-----.------Lost Dragon Software------.-----. .__oo. |\
{ \| ,-'' | _O_ | Member: UDIC lostdrgn@cris.com | _O_ | ``-, |/ }
`,_/,/_)\_ | | | http://www.cris.com/~lostdrgn/ | | | _/(_),\_,'
<...{_)_)_''`-----^-Official Dungeon Bane Web Page-'-----'^^_(_(_}...>

-----

_____/\___ Erraticus
____/__\__) -==(UDIC)==-
(__/ \__ \\//
/ \ \/

-----

Matt Kimmich, a.k.a. Thirith Dragon
"With writing, I find, you can have all the right ingredients, give plenty of time
and care, and still get nothing. Also true of love. Cooking, therefore, can keep a
person who tries hard sane." - John Irving, The World According to Garp

Footnotes

1. I would like to thank all Ultima Dragons and others for providing data, especially Abstract, Daermonestroer, Dave Erazmus, Dragonsbane, Erraticus, Frenetic, Kender, Lorien, Lost, Ophidian, Polychromic, Samurai, Singing, Silverlock, Sith, St. George's, Terasen, Underworld and Vulcan, for allowing me to quote them. Thanks, guys!

2. Cf. Beamann (1984), Chafe (1985) or Lakoff (1982).

3. There are exceptional situations that pose time limits on the production of a written text, such as exams. Nevertheless, even such situations allow a writer to reread and edit the text.

4. I.e., as opposed to written representation of dialogue, for instance in a work of fiction.

5. For further information on turn-taking, see Oreström (1983: 135-143).

6. E.g. English or Federation hull. Both of these fonts suggest very different contexts; the first one presents a rounder, more flowing image and suggests notions of romance, whereas the second one is more angular, suggesting perhaps a military context.

7. Cf. Brown and Yule (1983: 94-106)

8. E.g. writing personal letters serves the purpose of staying in touch.

9. http://bkfug.kfunigraz.ac.at/HELP/@local/FNEWS/USENET/PROPAGATION

10. There are also hierarchies in Usenet that carry mainly large data files; these are rarely used for any stretches of conversation and are not pertinent to this paper.

11. See the appendix for a list of all 7-bit ASCII characters. Some people or programs favour the so-called HTML (hypertext meta language) format that allows for typographical features such as different fonts, typographical effects or even graphics. However, since this format is not supported by all newsreaders, participants usually refrain from using it and criticise posts written in HTML.

12. This example thread contains eleven posts and branches up to the fourth level. Message 1.3 is on the second level, 1.3.2 on the third.

13. By the time this paper was finished, the newsgroup RGCU-D had been renamed and split into four topic-specific groups. As a result, certain statements in this section no longer apply; however, this does not change or invalidate the points made in this paper.

14. Other newsgroups that feature discussion on action games, for example, represent this average 'gamer' much more obviously; e.g. rec.games.computer.quake.

15. I am aware that such polls cannot be representative, but as a long-time regular on RGCU-D I can confirm that the statistics give a fairly accurate impression of regular discussion participants on the group.

16. Quoted from http://crash.ihug.co.nz/~charlton/andrew/. This open-topic policy is strongly reinforced by most regulars. Occasionally new members complain about more obscure or intense discussion, and it has been suggested that the newsgroup might be more suited for the soc.* hierarchy.

17. Posts were collected over the course of seven days, starting on Monday, 2.6.1997. The statistics used in the following only serve to give a general indication of traffic on RGCU-D.

18. Flame war is the term used generally for heated Usenet discussions that degenerate into an exchange of insults.

19. Since this post obviously was not written by a native speaker, the mistakes are not commented on by the other participants. In general, however, most English-language newsgroup participants expect a high level of 'correctness' and display a rather conservative attitude towards language.

20. In order to maintain the anonymity of the UDIC members, I will refer to them by their chosen Dragon names.

21. Cf. Angel and Heslop (1994: 93-94)

22. Some examples of sigs are given in the appendix.

23. Breaks between quotes from different posts will be indicated by a line of five dashes.

24. All emoticons represent stylised faces turned 90 degrees counterclockwise.

25. Also cf. Tannen (1989: 17-29) and Gumperz (1982).

26. Since Ultima V the player character is the so-called "Avatar".

27. I.e., as opposed to oral.

28. For further information on adjacency pairs, cf. Oreström (1983: 33).

29. In my data, several participants use expressions such as "You sound as if [...]" or "Did she just say what I think she said?".

30. This thread was intended as a kind of counterbalance to animosities and heavy flamewars during the weeks before.

31. "(sp)" and other variations of it - e.g. "<sp?>" - indicate that the author of a post or mail is not sure of his or her spelling. It is traditional in Usenet and is probably a reaction to the oversensitivity towards orthography mentioned earlier.

32. Cf. Tannen 1984.

33. In general, newsgroup participants often refer to RGCU-D in terms that are usually used for physical places; e.g. "Get out of here!", "That's it I'm leaving [this thread]!", etc.

34. It might be argued that interruptions thus can also be translated into the written medium. However, discussing this would exceed the frame of this paper.

35. Contextual effects are created by combining previous knowledge or information with current linguistic input; cf. Sperber and Wilson 1986.

36. I do not wish to suggest that spatial and temporal immediacy are equivalent or exchangeable; however, they seem to share certain properties. The exact properties of the two types of immediacy in discourse might prove a fruitful topic for further studies.

37. As Deborah Tannen (1980) suggests. She has since abandoned her model and now speaks of "strategies reflecting relative focus on involvement" (in: Tannen 1984: 21). In doing so, however, she only side-steps the discussion of oral vs. written strategies, in my opinion.

38. Computers would be ideal for examining features such as pauses or corrections in written text production.

39. It is interesting that all five posters whose age falls outside the average are female. However, as I do not discuss gender-specific communicative features in Usenet newsgroups, and as the quantitative data I have gathered is far from representative, I will not elaborate on it in this paper.

40. Quoted from http://macnash.admin.uottawa.ca/dmd00.htm. All numbers followed by a two-or three letter code are control characters; these apply only for printer data. In practice this means that only characters 32 to 126 as well as Carriage Return (#13; equivalent to line feed) are used as a standard. Certain newsreaders also support 8-bit code which includes 'Umlaute' and other nationality-specific symbols.