Communication Institute for Online Scholarship
Communication Institute for Online
Scholarship Continous online service and innovation
since 1986
Site index
ComAbstracts Visual Communication Concept Explorer Tables of Contents Electronic Journal of Communication ComVista

Structural Characteristics of Computer-mediated Language: A Comparative Analysis of Interchange Discourse
EJC logo
The Electronic Journal of Communication / La Revue Electronique de Communication
******* KO ************** EJC/REC Vol. 6, No. 3, 1996 ******


Kwang-Kyu Ko
Kaya University, South Korea

        Abstract.  This study investigates the
     structural characteristics of computer-mediated
     language by comparing one form of synchronous
     computer-mediated communication (CMC), Daedalus
     InterChange, with analogous spoken and written
     corpora.  The analysis focuses on 28 linguistic
     features claimed in previous research to charac-
     terize prototypical attributes of speaking and
     writing.  Each feature was examined and quantified
     to determine the structural characteristics of
     InterChange in comparison with the two other

        Three distinctive feature occurrence patterns
     emerge as a result of the analysis.  In the first
     pattern, the linguistic features of InterChange
     occur with higher frequency than in written dis-
     course, but lower than in spoken.  This occurrence
     pattern reflects a general tendency for
     InterChange to pattern like spoken language,
     offset to varying degrees by written-like
     constraints on the medium.  In the second pattern,
     linguistic features occur with higher frequency in
     InterChange than in either spoken or written
     discourse.  This pattern is claimed to be the
     result of users compensating linguistically for
     electronically-determined communicative
     constraints, especially the lack of turn-taking
     and interactional cues.  In the third pattern,
     linguistic features occur less often in
     InterChange than in either spoken or written
     discourse.  The lower occurrence of these
     features, which tend to be associated with
     elaborateness in writing, is analyzed as a
     consequence of the heavy processing burden placed
     on InterChange users, who are under both physical
     demands imposed by the need to type, and cognitive
     demands imposed by the real- time communication

        These results reveal that the InterChange
     discourse mode is not merely intermediate between
     (or a "combination of") speaking and writing;
     rather the electronic medium uniquely fosters some
     behaviors and inhibits others, in support of the
     view that the physical mode of communication
     shapes language use.

      Introduction:  Computer-Mediated Discourse as a
                 Different Mode of Language

     A new communicative situation has arisen with the
development of electronic communication systems, in which
"the words appear as lights on the computer screen" (Daiute,
1985, p. xiii).  Computer-mediated writing is regarded by
some scholars as a great advancement of "writing technology"
(Ong, 1982, p. 81) achieved by changing the set of
conditions in which symbols appear, that which Heim (1987)
calls the "elements" of language.  Heim distinguishes
"element" from "medium" in characterizing word technology:

     Medium emphasizes the instrumental method for
     communicative interchange.  Element emphasizes the
     conditions of symbolic expression and the
     implications of the mode in which things are
     represented (p. 102).

We may thus see computer-mediated discourse as a written
(typed) medium, but with a different language element, the
electronic element.

     Research on variation across spoken and written modes
of language has shown that the characteristics of language
use are often closely related to the means of language
production (Chafe, 1982).  Similarly, it has been claimed
that when communication is electronically mediated through
computers, the computer-mediated communication system
reshapes the forms and functions of language (Ong, 1982;
Crook, 1985; Peyton, 1986).  The term "electronic language"
in the present study is used to distinguish the language
produced in such electronic environments from other language
types[1].  Electronic communication can be further divided
into "real time [synchronous] communication" and "non-real
time [asynchronous] communication" (Quinn et al., 1983, p.

     The present study analyzes one form of synchronous
electronic language, that produced in a university setting
using a chat-like protocol known as Daedalus InterChange[2].
One of the most significant electronic elements of this type
of computer-mediated communication is its "linkage" (Heim,
1987, p. 160) in a time-sharing communication system.  Users
log on to a computer network which consists of a work
station linked by cables to individual terminals and
mainframes.  The individuals in the time-sharing community
share processing time, programs, storage space, and
resources such as printers.  In the InterChange protocol,
texts/messages can be brought simultaneously onto all the
screens in the same network, allowing multiple users to
communicate with each other.  This kind of receptive
communication capacity of the computer network provides a
social context for writing.  An individual who does not feel
an obligation to respond can act as a "bystander" (Crook,
1985, p. 66).  Otherwise, the individual users of the
time-sharing system communicate with each other -- reading
messages from or answering or sending personal messages to
other people working on the computer.  Such interactive
writing occurs in "real time"[3], allowing users to
approximate, to a greater extent than in asynchronous CMC,
the speed and spontaneity of oral communication.

     It has been claimed that the InterChange system, like
other computer-mediated communication systems, changes the
form of written language; "it becomes more speechlike, more
like a talking text than we now know, but yet not 'speech
writ down'" (Horowitz & Samuels, 1987, p. 27), because the
electronic computer system cannot provide all the contextual
and non-verbal support of real conversation.  However,
InterChange discourse has yet to be seriously investigated
by linguists.

     The potential impact of this new communication system
on various aspects of language use is enormous; the way in
which written language is produced by computer-mediated
communication systems may dramatically "change not only the
nature of writing as a process, but also the nature of
language as an object" (Horowitz & Samuels, 1987, p. 26).
It is thus important to conduct research to discover the
nature of electronic discourse as it takes place in

                    Research Methodology

     The research methodology in the present investigation
is designed to reveal the structural characteristics of one
type of synchronous computer-mediated discourse, Daedalus
InterChange, in comparison with spoken and written discourse
data provided by Biber (1988).  The scope of this study is
limited to one dimension of language that distinguishes
basic discourse types in terms of features of interaction
versus informativity.

Data Selection

     Computer-Mediated Discourse Data.  The electronic
discourse data were obtained from a university computer-
mediated communication system using InterChange, a
synchronous communication protocol designed primarily for
classroom use.  With InterChange, students in a class can
exchange messages with fellow students and a teacher
on-line.  The interface allows them to ask questions to a
specific person or post their message to the group as a
whole.  Messages scroll down the computer screen as the
discourse progresses, in the order in which they are
received by the system.  The user's real name or a
pseudo-name, as registered at the beginning of each session,
is automatically attached to the beginning of each message,
making it possible for users to track the source of each

     Since InterChange is a synchronous protocol, it does
not allow students to spend much time producing language or
editing it; rather there is pressure to produce and respond
to messages directly, as one would in face-to-face
communication.  At the same time, students are strictly
confined to a written mode of production:  all expression is
made by typing on a computer keyboard.  The discourse
produced through the InterChange program is herein termed

     The raw data of the computer-mediated discourse were
obtained from the InterChange computer network in use at a
large Midwestern university in the United States, and
contained discussions from different classes archived during
the 1993-94 academic year.  These data were produced by
students and instructors in university computer labs,
carrying out computer-mediated discussion as part of
regularly- scheduled course work.  The large amounts of raw
data that were obtained from the backup system of the
computer network were rearranged according to each class
session in order to avoid overlapping of the topics that had
been discussed in different classes.  Subsequently, a sample
text of 2,000 words was selected in which to count
linguistic features corresponding to comparable features in
the other two sets of data.  The InterChange sample was made
up of 94 messages/ turns produced by 25 college students and
one instructor in an English 101 class.

     The sample size was determined based on procedures
followed by Biber.  Biber (1990) analyzed aspects of
internal variation within texts and concluded that texts of
from 2,000 to 5,000 words in the standard corpora were
reliably representative of discourse categories for purposes
of analysis (p. 285).

     Comparison Data.  For purposes of comparison, in this
study I selected prototypical spoken and written data from
the London-Lund (spoken) and Lancaster-Oslo-Bergen (written)
corpora, as reported by Biber (1988).  Prototypicality was
determined in relation to the dimension of interactivity and
informativity, as described below.

     Spoken Data.  The London-Lund corpus of Spoken English
is a collection of 87 texts of 5000 words each produced in a
variety of situations by speakers of British English (Biber,
1988, p. 66).  The set of spoken data I chose from the
London-Lund corpus was all face-to-face conversation.  Face-
to-face conversation is one of the most prototypical genres
of spoken language in that it is highly interactive.
Participants engage, not only in verbal communication, but
also in non-verbal communication that includes gestures and
facial expressions.

     Written Data.  The Lancaster-Oslo-Bergen corpus of
Written British English is drawn from printed sources
published in 1961, and comprises 500 text samples of 2000
words each (Biber, 1988, p. 66).  The set of written data I
chose from the Lancaster-Oslo-Bergen corpus was official
documents.  Official documents are a prototypical genre of
written texts in that they are highly informative, rather
than interactive.

Linguistic Feature Selection

     The selection of linguistic features was limited for
the purposes of this study.  The linguistic features
selected are closely related to the fundamental dimension of
language that distinguishes such basic discourse types as
spoken and written, as determined by previous research
(Biber, 1988; Chafe, 1985).

     By the application of factor analysis, Biber (1988)
sorted out factors representing a basic dimension of
variation among spoken and written texts.  This dimension
was interpreted in terms of functions shared by co-occurring
features (1988, p. 104), and labeled by Biber as the
dimension of "Involved versus Informational" production.
Biber describes the "involved" end of the dimension as

     ... All of [these linguistic features] can be
     associated in one way or another with an involved,
     non-informational focus, due to a primarily
     interactive or affective purpose and/or to highly
     constrained production circumstances.  These
     features can be characterized as verbal,
     interactional, affective, fragmented, reduced in
     form, and generalized in content (Biber, 1988,
     p. 105).

     A total of 28 linguistic features from the "Involved
versus Informational" dimension with a factor loading above
0.42 from Biber's data were selected as representing the
basic dimensions of different types of discourse[4].  These
features are described in detail in the presentation of
results below.


     The goal of this study is to reveal the inter-
relationships between language mode, the structural
characteristics of the selected texts, and their discourse
functions.  Data from three language modes were compared:
Spoken, Written and InterChange, as described above.  The
structural characteristics of the text samples were
determined by measuring the frequency of 28 linguistic
features along the dimension of interactivity versus
informativity.  The frequencies of all 28 features were
compared across the three text types and three general
occurrence patterns were identified.  In order to interpret
these patterns, I drew on previous research in linguistics
in order to determine the functional parameters underlying
the frequency of each feature in the three modes of
discourse.  The results of these analyses are presented in
the following section.

        Analysis of Feature Distribution:  Pattern I

     The results of the quantitative analysis fell into
three patterns.  These represent three groupings of the
relative frequencies of features in InterChange, Spoken and
Written discourse.  In the first pattern, the occurrence of
features in InterChange is higher than in written discourse
("Written"), but lower than in spoken discourse ("Spoken");
that is, S>I>W.  These results are summarized in Table 1:


Table 1.  Feature Distribution Pattern I

                        Spoken         CMC         Written
                    (Conversation) (InterChange) (Documents)

first person pronouns    57.9         50.0          10.0
pronoun-it               20.0         17.5           3.2
demonstrative pronouns   13.1         10.5           1.1
DO as pro-verb            9.0          4.0           0.6

word length               4.1          4.1           4.9
amplifiers                6.0          3.5           0.9
emphatics                12.2         11.0           4.0
possibility modals        7.9          7.0           5.0
private verbs            35.4         22.5           7.8

contractions             46.2          7.0           0.0
THAT deletion             9.6          8.0           0.8
stranded prepositions     4.8          3.5           0.3
present tense           128.4        102.0          59.1
non-phrasal coordination  9.5          7.5           1.2
adv. subordinator-cause   3.5          1.0           0.1


In Table 1 each column shows the frequency of features in
Spoken, InterChange and Written respectively.  All
frequencies represent number of occurrences per 1,000 words.

     All of the features in Table 1 are complementarily
distributed between spoken and written discourse; that is,
when a feature occurrence is high in Spoken, its occurrence
is systematically lower in Written.  Conversely, when a
feature's occurrence is high in Written, its occurrence is
systematically lower in Spoken.  This is to be expected,
given that the features were selected from Biber's analysis
based on their ability to distinguish prototypical Spoken
and Written genres.  What is more interesting is that in
InterChange discourse, the features in Table 1 occur at a
frequency that is intermediate between that for Spoken and
that for Written, although their occurrence is sometimes
closer to one than the other.  In what follows, I interpret
this systematic occurrence tendency in relation to the
situational components with which the three types of
discourse are produced.

     For the purposes of presentation, the 16 linguistic
features which represent Pattern I are further sub-grouped
into three categories according to structural criteria:  1)
pronouns and pro-verbs, 2) lexical items, and 3) morpho-
syntactic variations.

Pronouns and Pro-verbs

     Pronouns in Pattern I include first person pronouns,
the pronoun-it, and demonstrative pronouns.  The pro-verb DO
also belongs to this pattern.  Biber (1988) notes that
personal pronouns and pro-verbs often mark a relatively low
informational load, lesser precision in referential
identification, or a less formal style.

     First person pronouns.  First person pronouns have been
viewed as markers of ego-involvement in the discourse
process.  Chafe and Danielewicz (1987) point out that first
person pronouns indicate an interpersonal focus that is
explicitly concerned with oneself.  Thus, the occurrence of
first person pronouns is closely associated with the degree
of interaction among participants.  First person pronouns
are found in the electronic discourse with high frequency.
The following are examples from InterChange.  (Note:  all
underlining in the examples was added by the author.)

(1) (a)  _I_ thought _I_ felt a spider crawl up _my_ leg
         last night while _I_ was asleep.
    (b)  Did you know that _me_ and Joe broke up?
    (c)  Your name will be on _our_ guest list.
    (d)  Oh, _we_ really do tell each other, _we_ just don't
         really make a big deal about it.

Table 1 above indicates the occurrence of first person
pronouns per thousand words in the Spoken (59.9),
InterChange (50.0), and Written (10.0) text samples.  Note
that the difference between InterChange and Written (40.0)
is far larger than the difference between InterChange and
Spoken (7.9).  As Chafe and Danielewicz (1987) point out,
the high occurrence of this feature in Spoken is closely
related to interaction with the addressee.  InterChange is a
highly interactive discourse type, like Spoken, but its
electronically-mediated interaction with the audience seems
to be somewhat less intense than in face-to-face
interaction, as indicated by the lower frequency of first
person pronouns.  The Written discourse in the sample has a
highly informative focus, rather than an interactive focus,
and this difference is reflected in the low frequency of
first person pronouns.

     Pronoun-it.  The pronoun _it_ can refer to a wide range
of entities from animate beings to abstract concepts, by
substituting for nouns, phrases, or whole clauses.  Biber
(1986) considers frequent use of this pronoun to be an
indication of relatively inexplicit lexical content due to
strict time constraints and a non-informational focus.
Similarly, the occurrence of pronoun-it may be associated
with the time constraints and contextual limitations of
on-line communication.  An InterChange example follows:

(2)  When you first have a child, _it_ takes a while for
     you to get used to having one.  _It_ is strange to
     have something that is part of you and totally your

Pronoun-it occurs less often in InterChange electronic
discourse (17.5) than in Spoken (20.0).  The reason seems to
be related to the contextual differences between Spoken and
InterChange.  Electronic discourse needs to be more explicit
to avoid referential ambiguity, because of the greater
contextual limitations than in face-to-face communication
(e.g., lack of intonation and gestures).  At the same time,
pronoun-it is much more frequent in InterChange than in
Written (3.2), suggesting lesser formality and a lesser
information focus in InterChange.

     Demonstrative pronouns.  Demonstrative pronouns such as
_this_ and _that_ can refer to an entity outside the text
(an exophoric referent), or a previous referent in the text
itself (an endophoric referent).  Halliday and Hasan (1976)
treat endophoric demonstratives as important devices for
achieving discourse cohesion.  The following InterChange
examples illustrate endophoric use of demonstrative

(3) (a)  How dare you talk about the woman I love _that_
    (b)  All _these_ people trying to help everyone else
         adjust to something new.
    (c)  All _this_ talk about _this_ poem isn't helping
    (d)  By presenting not just one study but several they
         trace the questions they have explored, the answers
         they have found and the questions _those_ answers
         have raised.

As shown in Table 1 above, demonstrative pronouns occur much
more in Spoken (13.1) and InterChange (10.5) than in Written
(1.1).  Apparently, InterChange discourse achieves discourse
cohesion, in part, by means of demonstrative pronouns, while
Written achieves discourse cohesion by means of other
grammatical devices.

     Pro-verb DO.  Biber (1988) notes that DO as pro-verb
substitutes for an entire clause, thereby reducing the
information density of a text.  It indicates a lesser
informational focus, due to processing constraints or a
greater concern with interpersonal matters.  The following
is an InterChange example:

(4)  You should look at people for who they are not for how
     neat or sloppy they are, but unfortunately lots of
     people _do_.

     [cf. ...unfortunately lots of people look at people for
     how neat or sloppy they are.]

As with the other pronouns, the occurrence of pro-verb DO
clearly distinguishes Spoken (9.0) from Written (0.6), and
its occurrence in InterChange (4.0) discourse falls
approximately in the middle.  In InterChange electronic
discourse, the occurrence of DO as a pro-verb is lower than
in Spoken, probably because InterChange discourse requires
more specific forms of reference to make meaning clear in
the absence of intonational and gestural cues.

Lexical Items

     Word length.  Haiman (1983) shows that longer words
convey more specific, specialized meanings than shorter
ones, while shorter words are more frequently used with more
general meanings.  Chafe and Danielewicz (1987) find a
relationship between lexical specificity and the production
differences associated with difference modes of discourse;
the faster spontaneous discourse is produced, the less
lexical specificity is found.  Thus, word length can be
interpreted as marking the relative degree of lexical
specificity in the discourse.  The following InterChange
example is typical in terms of word length:

(5)  Having class like this is great, we should motion
     to have it like this everyday! This makes more
     sense to me than poetry.

Words in the Written sample are comprised of an average of
4.9 letters, and words in Spoken average 4.1 letters, while
InterChange words also average 4.1 letters.  The reason why
words are more complex in Written than in Spoken seems to be
related to temporal constraints on discourse processing.
Because it is cognitively demanding to think and talk
simultaneously, speakers tend to use simpler language (less
complex grammar, shorter words, more repetition of the same
words) than writers.  In InterChange, cognitive demands on
production are even greater than in speaking, since one must
type (which requires more effort than speaking) at a near-
conversational pace.  At the same time, one has more time to
think, since typing, no matter how quickly one does it, is
still slower than speaking.  The fact that words are the
same length in both Spoken and InterChange seems to imply
that these two competing factors cancel each other out.

     Amplifiers.  Amplifiers are words which have a boosting
effect on the meaning of verbs, adjectives or adjective
phrases (Quirk et al., 1985).  Biber (1988) includes the
following words as amplifiers:  absolutely, altogether,
completely, enormously, entirely, extremely, fully, greatly,
highly, intensely, perfectly, thoroughly, totally, utterly,
and very.  Chafe (1985) claims that amplifiers indicate, in
positive terms, the reliability of propositions.  Holmes
(1984) notes that amplifiers can signal solidarity with the
listener, in addition to marking certainty or conviction
towards the proposition.  An InterChange example is the

(6)  It is strange to have something that is part of you and
     _totally_ your responsibility.

Amplifiers occur with very low frequency in Written (0.9),
compared to Spoken (12.2) and InterChange (11.0).  However,
InterChange manifests a slightly lower frequency than
Spoken.  These occurrence patterns may be related to degree
of solidarity among participants and the absence in CMC of a
direct audience; on the network, because one's interlocutors
are not visible and are often relative strangers, one may
feel less solidarity with one's audience than in
face-to-face conversation.

     Emphatics.  Biber (1988) notes that emphatics mark the
presence (versus absence) of certainty, in contrast with
amplifiers which indicate the degree of certainty towards a
proposition.  Chafe (1985) considers that, in speech,
emphatics indicate involvement with the topic.  Biber (1988)
includes the following words as emphatics:  for sure, a lot,
such a, just, really, most.  An example from InterChange is
the following:

(7)  That's _really_ selfish.  Oh, we _really_ do tell each
     other, we just don't _really_ make a big deal about it.

There is little difference between Spoken (12.2) and
InterChange (11.0) in terms of the occurrence of emphatics,
in spite of the difference in language mode, suggesting that
participants are only slightly less intensely involved with
discourse topics in the electronic network.  In general, the
distribution of amplifiers and emphatics suggests that the
participants become more personally involved with the
discourse when their discourse takes place interactively.

     Possibility modals.  Chafe (1985) notes that a speaker
or writer's assessment of the reliability of knowledge in a
proposition may be marked with possibility modals.  Thus,
the occurrence of possibility modals is a reflection of the
speaker's attitude toward the reliability of knowledge.
Biber (1988) includes can, may, might, and could as
possibility modals.  The examples in (8) are from the
InterChange sample.

(8) (a)  _Can_ this guy beat a dead horse?
    (b)  Kim, _could_ cow heavy refer to her breast being
         filled with milk ?
    (c)  I think this essay _may_ be kind of difficult to
         present, I really don't have any ideas so far.
    (d)  I agree that it _might_ be more comprehensible if
         it dealt with your field, but are all other works
         useless by definition?

The distribution of possibility modals shows that all three
modes of discourse signal the degree of reliability of
knowledge by means of this feature.  However, possibility
modals are used more often in Spoken (7.9) and InterChange
(7.0) than in Written (5.0).  The higher frequency of
possibility modals in Spoken and InterChange may reflect the
interlocutors' lesser certainty toward the statements they
are making under interactive real-time constraints.

     Private verbs.  A certain restricted class of verbs
describes mental activities.  Biber (1988) notes that
private verbs express intellectual states or non-observable
intellectual acts, e.g., assume, believe, conclude, decide,
estimate, feel, guess, hope, infer, know, learn, mean,
notice, prove, realize, seem, think, and understand.
InterChange examples are given in (9):

(9) (a)  Everything we did I can look back on and I _feel_
         bad for doing it, but I don't regret it.
    (b)  Well this is my first time using this computer.
         It _seems_ to be confusing.

InterChange (22.5) manifests fewer private verbs than Spoken
(35.4), but far more than Written (7.8).  From this we may
conclude that InterChange makes individuals feel relatively
free to share personal cognition and emotions with the
audience.  The fact that the InterChange sample is drawn
from class discussions may account for why it contains fewer
private verbs than the conversational Spoken sample.

Morpho-Syntactic Variation

     Contractions.  Biber (1988) defines contractions as
phonologically (orthographically) reduced surface structural
forms, and notes that they are usually dispreferred in edited
formal writing.  Frequent use of contractions is seen as a
consequence of a desire for fast and easy production under
time constraints.  Examples of contractions from the
InterChange sample are given in (10):

(10)(a)  Hi! everyone, _it's_ a beautiful day in the
    (b)  I think _we're_ OK now, but _we've_ been fighting a
         lot over stupid things.
    (c)  _I'm_ sick today and _don't_ want to be here.

Contractions are found only in Spoken (46.2) and InterChange
(7.0), although the occurrence of contractions is much
higher in Spoken.  Contractions do not occur at all in the
formal Written discourse included in this sample (0.0).

     The high frequency of contractions in Spoken is due to
a phonological principle of reduction governing real-time
speech production.  However, although InterChange discourse
is also produced under temporal constraints, the mechanical
mode of production is different.  Typing contractions does
not save as much time or effort as does saying them, since
in many cases the same number of characters is required.
This difference is reflected in the relatively low
occurrence of contractions in InterChange.  What
contractions do occur in InterChange seem to be used for
stylistic value, as a marker of familiarity or informality.
Conversely, contractions rarely appear in edited formal
writing, a tendency which is also reinforced by the
operation of prescriptive rules.

     Subordinator-that deletion.  Subordinator-that deletion
is a form of syntactic reduction which may be viewed as a
mismatch between surface form and underlying form, caused by
rapid production under real-time communicative constraints.
The following examples are from InterChange:

(11)(a)  I don't think ( ) I could live very well with
         someone like that.
    (b)  I thought ( ) I felt a spider crawl up my leg
         last night while I was asleep.

That-deletion occurs in Spoken (9.6) somewhat more often
than in InterChange (8.0).  However, the overall occurrence
of this feature is not as high as that of contractions,
because it is directly dependent on the frequency of
THAT-complement clauses, which are relatively uncommon.  As
this feature is regarded as a mismatch between surface form
and underling form, it tends to be avoided in edited formal
writing (0.8).  Biber (1988) asserts that a tendency toward
elaborated and explicit expression in edited writing is the
driving force preventing this reduction (p. 244).

     Stranded prepositions.  A stranded preposition is a
preposition that is not followed by any complements.
According to Biber (1988), stranded prepositions represent a
mismatch between surface and underlying representation as
well, since the relative pronoun and the preposition belong
to the same phrase in underlying structure.

(12)  My roommate is one of the people that I used to make
      fun _of_, and now he feels the same way that I do.

      [cf. 'of whom I used to make fun']

Final prepositions are relatively more frequent in both
Spoken (4.8) and InterChange (3.5) than in Written (0.3).
This tendency may be related to constraints in real-time
discourse production which leave one little time to pay
attention to prescriptive rules such as that regarding
stranded prepositions.  It seems important also to
acknowledge that both THAT-delation and stranded
prepositions have a conventional stylistic value, as markers
of informal discourse.  The prescriptive rule for
prepositions, in particular, results in stilted,
formal-sounding usage that would be inappropriate in a
casual interaction.

     Present Tense Verbs.  Biber (1988) notes that present
tense verbs are used with subjects of current importance.
Ochs (1979) observes that present tense verbs are more often
used in unplanned speech styles.  The following is an
InterChange example:

(13)  She _compares_ the child to a gold watch because gold
      _is_ one of the most expensive metals in the world.
      Everyone _knows_ children _are_ expensive.

The occurrence of present tense verbs is very high in Spoken
(128.4, as compared with only 59.1 for Written), indicating
a high degree of immediate contextual communication making
use of an unplanned discourse strategy.  InterChange (102.0)
is unplanned like spoken discourse, but it is a less
immediate discourse.  This difference may account for the
distribution of present tense verbs.

     Independent coordination.  Biber (1988) notes that
independent coordinations (or non-phrasal coordinations) can
mark many different logical relations between two clauses.
Chafe (1985) associates independent coordination with the
"fragmentation" of discourse, which results from the simple
chaining of idea units by means of juxtaposition or
coordinate conjunctions.  Thus, "the idea units in this
style of discourse tend to be not only shorter, but also
more independent of each other" (p. 111).  Some InterChange
examples follow:

(14)(a)  Everybody knows children are expensive.  _So_ who
         is everybody?
    (b)  Bert has class until 5 or so. _Then_ he has night
    (c)  I can look back on it _and_ I feel bad for doing
         it, _but_ I don't regret it.

The distribution of independently coordinated propositions
seems to be correlated to the degree of fragmentation of
each discourse type.  Spoken (9.5) and InterChange (7.5)
show relatively more frequent independent coordinations than
Written (1.2), because real-time discourse modes tend to be
fragmentary under timing constraints.

     Adverbial causal subordinator.  Beaman (1984) and
Tottie (1986) find more causationals in speech than in
writing, but the functional reasons for this distribution
are unclear.  Biber (1988) treats adverbial clauses as
linguistics devices that indicate informational relations in
a text.

(15)  She compares the child to a gold watch _because_ gold
      is one of the most expensive metals in the world.

The occurrence of this feature in InterChange (1.0) is lower
than in Spoken (3.5), but higher than in Written (0.1).
With regard to this feature, InterChange is closer to
Written than Spoken.  InterChange and Written indicate more
explicit informational relations between clauses, in
comparison with Spoken clauses which are often juxtaposed in
a fragmentary fashion.

     The findings for Pattern I can be summarized as
follows.  Electronic discourse tends to follow the Spoken
discourse pattern rather than that of Written with regard to
the distribution of pronouns and pro-verb DO, in spite of
the fact that InterChange shares some situational elements
with both speaking and writing.  This suggests that these
are features of interactive discourse more generally.
Lexical choice is influenced by more complex factors, such
as the user's degree of personal involvement in and
commitment to the truth of an utterance in the case of
amplifiers, modals and private verbs, and cognitive
processing constraints in the case of word length.  Finally,
the occurrence of morpho- syntactic variation seems to be
influenced by real-time communicative constraints on the
rapid production of discourse.  Reduced morpho-syntactic
variants also mark informality in InterChange.  All of these
features are generally related to interactive discourse, but
the fact that their occurrences are less than in
face-to-face conversation suggests that there are opposing
tendencies caused by the written (typed) nature of the
medium that partially cancel the interactive features out.

              Feature Distribution Pattern II

     In the second pattern, the occurrence of InterChange
electronic features is higher than that of both Spoken and
Written (SW).  Seven out of the 28 linguistic features
exhibit this pattern.


Table 2.  Feature Distribution Pattern II

                        Spoken         CMC         Written
                    (Conversation) (InterChange) (Documents)

WH-questions              0.7         11.0          0.0
indefinitive pronouns     3.9         11.5          0.2
BE as main verb          29.5         37.0         16.5
WH-clauses                1.3          2.5          0.2
discourse particles       3.9          5.5          0.0
analytic negation        18.5         20.5          3.4
second person pronouns   30.8         31.0          1.4


All of the features in Table 2 are "spoken features;" that
is, they are found more often in Spoken than in Written.  It
is notable that with regard to this set of features,
InterChange is more "speech-like" even than natural spoken

     This feature occurrence pattern suggests that the
situational elements normally associated with speaking and
writing do not cancel each other out in electronic
discourse, but rather heighten the occurrence of certain
features in a way that is unique to electronic discourse.
It has been observed that print writing compensates for loss
of information present in the oral mode by adding words or
altering wording (Jones, 1943, p. 207).  Similarly, it is my
contention that certain feature occurrences in InterChange
are enhanced as linguistic compensation for strategies that
are unavailable in the computer-mediated communication
situation.  That is, the absence of one communicative
strategy is compensated for with others when the discourse
is produced in a different mode.  This is discussed for each
individual feature below.

     WH-questions.  Biber (1986) treats direct WH-questions
as markers which indicate "a high degree of interpersonal
interaction and personal involvement" (p. 394) with the
addressee.  WH-questions are identified with the WH-element,
which is a phrase containing or consisting of a WH-word.
The following are examples from InterChange:

(16)(a)  _What_ is this poem talking about?
    (b)  _Why_ keep things from other people?
    (c)  So _who_ is everybody?
    (d)  _Where_ is Stacie?
    (f)  _When_ did that happen?

The occurrence of direct WH-questions is much higher in
InterChange electronic discourse (11.0) than in Spoken
(0.7), in spite of the fact that both are produced in an
interactive manner.  One reason for this relates to the
communicative situation in which the discourse is produced.
InterChange discourse is produced by multiple participants
who are physically separated and cannot see one another,
which makes it difficult to engage in coherent and orderly
patterns of turn-taking.  Questions are first parts of
adjacency pairs (Schegloff & Sacks, 1973) which
conventionally require a response.  The high frequency of
WH-questions in the InterChange sample may be a way of
creating structured interaction, in compensation for the
unavailability of other turn-taking cues such as intonation,
gesture, and gaze.  It should also be noted that discussion
on InterChange usually begins in reaction to a question that
invites speculation, triggering the use of further
questions, as is typical in a classroom situation.  Thus the
fact that the InterChange sample is "classroom" discourse
may also contribute to a higher frequency of WH-questions.
In contrast, no WH- questions were found in the
non-interactive Written sample (0.0).

     WH-clauses.  Beaman (1984) sees WH-clauses as
resembling questions and serving interpersonal functions in
discourse.  However, Chafe (1985) treats complement clauses,
including WH-clauses, as devices for "idea unit expansion"
(p. 108), as is more common in informational discourse.  The
following are some examples from InterChange:

(17)(a)  I think _what_ you said about the bonding of the
         mother and the newborn is a pretty good assumption.
    (b)  That's probably _why_ I don't talk to guys that talk
         about that shit.

The occurrence of WH-clauses is generally low in all three
types of texts.  Although Spoken (1.3) shows a higher
frequency of this feature than Written (0.2), the frequency
of this feature in InterChange (2.5) is higher than in
either Spoken or Written.

     The relatively high occurrence of WH-clauses seems to
be related to the high occurrence of WH-questions in
InterChange mentioned above; WH-questions are often embedded
into WH- clauses when users refer to WH-questions posted by
others in the electronic discourse.

     Indefinite pronouns.  Indefinite pronouns have not been
used frequently for register comparison in discourse
analysis.  Biber (1988) mentions them as markers of
unspecified pronominal reference, including as examples:
anybody, anyone, everybody, everyone, everything, nobody,
none, nothing, nowhere, somebody, someone, and something.
The following are InterChange examples:

(18)(a)  Hey!_ anybody_ else out there a devoted Christian?
    (b)  Is _anyone_ out there?
    (c)  Hey Listen if you don't have _anything_ to do on
         friday night come to the Sigma Alpha Epsilon House
         down on riverside.
    (d)  So who is _everybody_?
    (e)  _Everything_ we did I can look back on and I feel
         bad for doing it, but I don't regret it.
    (f)  _All these people_ trying to help _everyone_ else
         adjust to _something_ new.

The frequency of indefinite pronouns is very high in
InterChange (11.5), in comparison with Spoken (3.9) and
especially Written (0.2).  This high frequency is related to
a situational difference:  electronic discourse such as
InterChange is usually carried out with multiple,
unspecified interlocutors.  Users do not know for certain
who their audience is at any given moment.  This fact
appears to be reflected in the high occurrence of indefinite

     Discourse particles.  Schiffrin (1985, 1987) claims
that discourse particles are used to maintain conversational
coherence.  Chafe (1985) regards discourse particles as
"flow- monitoring devices" which control the "information
flow" in involved discourse (p. 112).  Biber (1988) includes
as discourse particles the words well, now, anyway, anyhow,
and anyways.  The following are InterChange examples:

(19)(a)  _Now_ we know why Aristotle and Quintillion were so
         hung up on the ethical appeal.
    (b)  I've never enjoyed being told how to write _anyway_
         ...But I've yet to find a text for a class.

InterChange (5.5) shows higher use of discourse particles
than Spoken (3.9).  This suggests that InterChange achieves
overall coherence by means of discourse particles, rather
than by other means such as the global organization of
discourse (common in writing) or lexical and intonational
repetition (common in speaking).  The higher occurrence of
discourse particles in InterChange may be the result of an
increased need to monitor the flow of information in a
situational context where there are multiple participants
and no simultaneous feedback cues available to show
listenership.  That is, users must explicitly construct
discourse relations that are otherwise implicit when
participants share a physical context.  Discourse particles
are entirely absent in the formal writing in the Written
sample (0.0), suggesting that their use is stylistically
informal as well.

     BE as main verb.  Biber (1988) notes that BE as a main
verb is frequently found in non-complex constructions with
reduced informational load, and is a marker of a static,
informational style of discourse.  Predicative constructions
are regarded as alternatives to more integrated attributive
adjective constructions.  For instance, the attributive
construction "the big boy" is an alternative to the
predicative construction "the boy is big."  Thus, the
frequency of BE verbs with predicative adjectives is
associated with fragmentation in discourse.  InterChange

(20)(a)  You know, this conversation _is_ both enjoyable and
         relevant to the assignment as well. I'm proud of us!
    (b)  The descriptions in the last stanza _are_ very vivid
         and descriptive.

BE as a main verb occurs with high frequency in all three
types of discourse.  However, InterChange (37.0) and Spoken
(29.5) show a much higher BE verb use than Written (16.5).
This pattern is the result of a more frequent use of
predicative constructions in InterChange than attributive
constructions, which are more integrated.  Under the heavy
production constraints of InterChange, it appears that users
have less time to construct the more cognitively demanding
attributive constructions than even in Spoken.  A social
explanation may play a role here as well.  Stative
predications are frequently used for description and
evaluation, both of which are important activities in
college writing classes like those from which these
InterChange data were obtained.  This could also contribute
to the higher use of BE verbs in the sample.

     Analytic negation.  Tottie (1991) distinguishes between
synthetic negation (e.g., He is neither smart, nor diligent)
and analytic negation (e.g., He is not smart).  Analytic
negation is treated as more colloquial and more fragmented
than syntactic negation; two independent clauses are needed
to express what is expressed by one clause in the syntactic
type.  For the purposes of this analysis, analytic negation
was counted for all sentences which make use of the analytic
negation adverb 'not' and its contracted forms.  InterChange
examples follow:

(21)  _Not_ a whole lot. I am sick today and I _don't_ want
      to be here. All this talk about this poem _isn't_
      helping either.  I _won't_ have anything to write about
      in my journal because I _don't_ understand.

Analytic negation is much more frequent in InterChange
(20.5) and in Spoken (16.6) than in Written (3.4).  Analytic
negation is strongly preferred to syntactic negation in the
interactive modes because of the fragmentary nature of
interactive discourse.  That is, the temporal constraints of
real-time discourse production favor relatively simple
structures, as noted elsewhere in this discussion.  However,
this does not explain why InterChange contains more analytic
negation than Spoken.  The explanation for this seems to
reside in the fact that most negation is analytic negation;
thus InterChange has the most negation (of any type) of the
three samples.  This is consistent with the tendency
observed by Kim and Raja (1991) for computer-mediated
communication to contain a high degree of interpersonal

     Second person pronouns.  Second person pronouns refer
to an addressee in the discourse.  Chafe and Danielewicz
(1987) link the occurrence of second person pronouns to
involvement with the addressee and the degree of interaction
among participants.  The InterChange electronic discourse
exhibits a high degree of occurrence of these features, as
illustrated in the following example:

(22)  As a matter a fact, _you_ opened up to me about _your_
      girlfriend problems, and told me some of the emotions
      that were going through _your_ head at the time.

Second person pronouns occur with high frequency in Spoken
(30.8) in comparison with Written (1.4), owing to the
interactive nature of the spoken discourse.  In InterChange,
second person pronouns occur with an even higher frequency
(31.0) than in Spoken.  In face-to-face conversation,
interlocutors may be selected with non-verbal cues such as
gestures, facial expressions and eye contact, which requires
that participants be visible to one another.  The lack of
visibility in on-line communication is compensated for in
part with a higher frequency of second person pronouns.

     In summary, the high occurrence of Pattern II
linguistic features in InterChange reflects the
characteristics of the electronically determined
communicative situation.  In order to maintain
conversational coherence in a situation where the number and
identities of participants are unknown, communicators ask
more questions and use more second-person and indefinite
pronouns to evoke unspecified participants, and control the
flow of information by means of discourse particles.  The
higher occurrence of these linguistic features compensates
for constraints in the electronic discourse situation.
Other behaviors, such as frequent use of main verb BE,
appear to reflect the classroom context.

              Feature Distribution Pattern III

     In the third feature pattern, the frequency of the
InterChange electronic discourse features is lower than in
either Written or Spoken.  There are six features out of the
selected 28 linguistic features which show this pattern.


Table 3.  Feature Distribution Pattern III

                       Spoken         CMC         Written
                   (Conversation) (InterChange) (Documents)
type/token ratio        46.1         33.7          47.8
nouns                  137.4        118.5         206.5
prepositions            85.0         68.5         150.9
attributive adjectives  40.0         12.5          77.9
hedges                   2.1          0.0           0.0
sentence relatives       0.7          0.0           0.1


This pattern results from a mechanism that is in some sense
opposite to that in Pattern II.  Rather than being enhanced
by situational factors, the occurrence of linguistic features
in this pattern is disfavored in interactive real-time
communication, because of cognitive and mechanical constraints
that exist in the electronically-mediated communicative

     Type-Token ratio.  Type-token ratio (TTR) is the ratio
between the total number of different lexical items (types)
and the total number of words (tokens).  The type-token
ratio is a basic means for measuring the lexical diversity
in a text sample.  The range of vocabulary used reflects the
way one chooses words in a given communicative situation.

     As Table 3 above illustrates, TTR in Written (47.8) is
higher than in Spoken (46.1), that is, the range of
vocabulary in the written discourse is "richer" than that of
spoken discourse.  TTR in InterChange (33.7) is lower than
in either Spoken or Written.

     It is generally held that lexical diversity reflects
the production efforts of discourse participants.  Chafe and
Danielewicz (1987) provide a plausible account of this
phenomenon.  Written discourse is produced more slowly than
Spoken because of mechanical constraints on the written mode
of language, whether it is typing on a keyboard or writing
with pen and paper.  The slow process of language production
allows the writer to access more lexical resources in long-
term memory.  In contrast, spoken discourse is produced with
fewer mechanical constraints, relatively rapidly.  However,
in this process of rapid production, speakers have little
time to sift through all the possible choices they might
make, and often settle on the first words that occur to

     InterChange, in a sense, inherits the worst of both
worlds.  InterChange discourse is produced with the physical
constraints of the written mode of language, in this case
typing.  In the real-time communicative situation, however,
InterChange is expected to be rapid, almost as rapid as
speaking (although it cannot be; not quite).  This puts
tremendous production pressures on the user, to "write" as
fast as one normally "speaks".  The reduced lexical
diversity of InterChange appears to be a consequence of
these competing demands.

     Total nouns.  Biber (1988) notes that a high nominal
content in a text indicates a high (abstract) informational
focus, as opposed to a primarily interpersonal focus.  The
total number of nouns in a text provides an overall nominal
assessment of the discourse text (p. 228).  The nominal
assessment is associated with the relative degree of
information focus, while interpersonal discourse is more
concerned with actions of immediate relevance.  The
following sentence illustrates the use of nouns in the
InterChange sample.

(23)  I was a _member_ of a _clique_. I hung out with the
      _jocks_, since I used to run _track_.

Nouns are found in all three types of discourse with
considerable frequency.  As Table 3 above indicates, nouns
occur most frequently in Written (206.5), perhaps because
the Written sample has a highly informational focus.  Their
lower frequency in Spoken (137.4) and InterChange (118.5)
suggests that these modes contain proportionately more
verbs, consistent with a concern for actions required for
immediate relevance in interactive discourse.  The lower
occurrence of nouns in InterChange suggests an even higher
frequency of verbs in the computer-mediated discourse than
in Spoken, which is surprising, given the lack of (actual)
physical interaction in CMC.  It may be that users attempt
to create a sense of activity in real-time CMC by talking
more about actions.

     Prepositional phrases.  Biber (1988) notes that
prepositions are important devices for packing information
into written discourse.  Chafe (1982, 1985; Chafe &
Danielewicz, 1987) treat prepositions as a device for
integrating information into idea units.  Some InterChange
examples are given in (24) below.

(24)  I have a question not relevant _to_ this essay.  Did
      anyone have their room bombed _for_ bugs?  I heard that
      all rooms are supposed to be done.  I thought I felt a
      spider crawl _up_ my leg last night while I was asleep.
      I freaked out.

Prepositional phrases occur with considerable frequency in
all three types of discourse.  However, the occurrence of
prepositional phrases is much higher in Written (150.0) than
in Spoken (85.0).  In InterChange (68.5), the occurrence of
this feature is even lower than in Spoken, indicating that
InterChange discourse contains more "light idea units"
(Chafe, 1982).  Under the combined time/production
constraints, InterChange clauses tend to be stripped down to
their obligatory core, minus optional adjuncts such as
prepositional phrases.

     Attributive Adjectives.  Attributive adjectives are
used in discourse to expand and elaborate on the information
presented in a text.  When a concept is modified with
attributives, it usually becomes clearer or more elaborated.
Thus, the distribution of attributive adjectives is closely
related to the elaborateness of a text.  Chafe (1985) treats
attributive constructions as devices used in idea unit
integration and expansion.

(25)  I don't think _messy_ people and _neat_ people can
      live together very well.  The _neat_ person would
      always be trying to clean up for the _messy _ person.

The distribution of attributive adjectives is lower in
Spoken (40.0) than in Written (77.9); however, the
occurrence of attributive adjectives in InterChange (12.5)
is much lower than even in Spoken.  Such a low occurrence in
InterChange can be accounted for in terms of cognitive
constraints imposed by the real-time communication, as
observed by Chafe and Danielewicz (1987).  Attributive
adjectives are a means of increasing the information density
of an utterance by embedding a proposition within a noun
phrase.  To the extent that embedding requires extra
processing effort, it is disfavored in a mode where
communicators already must struggle with processing burdens
imposed by time and the physical demands of typing.

     Hedges.  Biber (1988) notes that hedges are informal,
less specific markers of probability or uncertainty.  Chafe
and Danielwicz (1987) find that the time constraints of
speech are related to the use of hedges; speakers use more
hedges to indicate the uncertainty of the words chosen under
the temporal constraints of real-time communication.  Hedges
also buy the speaker processing time, and they hold the
floor (i.e., prevent interruption) so that the speaker gets
a chance to finish his turn.  Biber (1988) includes _about_,
_something like_, _more or less_, _almost_, _maybe_, _a sort
of_, and _a kind of_ as hedges.

     Interestingly, there are no hedges found in the Inter-
Change discourse.  It is not necessary to hold the floor via
hedges in InterChange -- turns are sent in their entirety,
so interruption is technically impossible.  Thus,
InterChange users may omit hedges because they are not worth
the trouble of writing/typing, given the time constraints of
the medium.

     Sentence relatives.  Winter (1982) points out that
relative clauses provide a way to talk about nouns, either
for identification or simply to provide additional
information.  Sentence relatives do not have a nominal
antecedent, but rather refer to the entire predication of a
class (Quirk et al., 1985, pp. 1118-20), functioning as a
type of comment clause.  The occurrence of this feature is
extremely low in Spoken (0.7) and in Written (0.1).  There
are no occurrences at all of this feature in InterChange

     It is not a coincidence that many of the features in
Pattern III occur with highest frequency in Written; these
feature occurrences are closely associated with information
focus and elaborateness of discourse, and require processing
time and effort to produce.  Since this is opposite to the
informal, interactive tendencies of InterChange discourse,
the InterChange data tend to pattern closer to Spoken with
regard to these features.  The fact that the features are
found least often in InterChange appears to be the result of
the heavy cognitive demands imposed by the requirements of
real-time discourse processing, in combination with the
mechanical constraints on production in the electronic mode.

                   Summary and Conclusion

     In this investigation, the structural characteristics
of InterChange were measured in comparison with Spoken and
Written text samples in terms of the frequency of 28
linguistic features.  Each feature of the electronic
discourse was described in terms of its discourse functions,
and categorized as belonging to one of three basic
occurrence patterns on the basis of its frequency relative
to Spoken and Written.  The features in Pattern I occur with
higher frequency than in Written, but lower than in Spoken.
The linguistic features in Pattern II occur with higher
frequency than in either Spoken or Written, and the
linguistic features in Pattern III occur less often than in
either Spoken or Written.

     Previous research has tended to situate computer-
mediated language as intermediate between speaking and
writing (Collot & Belmore, 1996; Yates, 1996), or else as
more speech-like (Werry, 1996).  The difference seems to
have to do with whether the electronic discourse takes place
in non-real time or real time, with the latter being more
"speech-like" than the former.  The present study of a
real-time genre of electronic discourse also finds that it
is more similar to Spoken than Written language overall.
That is, InterChange discourse tends to be interpersonally
involved, syntactically fragmented, and to have a relatively
low degree of information focus and elaborateness.

     However, my findings show that InterChange discourse is
not "speech-like" in any simple way.  Depending on the
discourse function of the particular linguistic feature, it
may approximate speech more closely than writing, or it may
split the difference between the two, as in Pattern I. More
surprising, it may be more "speech-like" than actual speech,
as in Pattern II, especially when the features involved
compensate for the lack of turn-taking and other discourse
management strategies in the electronically-mediated
situation.  Alternatively, InterChange discourse may be
relatively linguistically impoverished compared with
speaking and writing, as in Pattern III.  This pattern is
found especially with features characteristic of information
focus and elaborateness, and appears to be a consequence of
production constraints imposed by the InterChange medium
which are heavier than those for either speaking or writing.

     In addition to providing a linguistic description of
InterChange discourse, these results support the general
claim that the electronic mode of production shapes
computer- mediated language, just as writing and speaking
modalities shape the nature of written and spoken language.
Further, the study shows how occurrence patterns of
linguistic features are influenced by underlying functional
factors, confirming Biber's assertion that, "if certain
features consistently co-occur, then it is reasonable to
look for an underlying functional influence that encourages
their use" (Biber, 1988, p. 13).

     Last but not least, the results of this study have
practical implications for the use of text-based synchronous
computer-mediated interaction in educational settings.  The
finding that the interactive medium disfavors elaborated
forms of expression calls into question its suitableness for
use in high-level academic discussions.  At the same time,
the interactive nature of the medium may make it well-suited
for class discussion in which widespread participation is
the desired outcome (Colomb & Simutis, 1996).  Future
research could test these hypotheses by applying methods of
discourse analysis which focus on these and other discourse
functions to comparisons of face-to-face and
computer-mediated classroom activities.


     [1] The distinction between "electronic language" and
other media is based on Heim.  He sees spoken language and
written language as two different media, and electronic
language as a written medium with electronically shifted
elements -- shifted conditions of symbolic expression (Heim,
1987, p. 101).

     [2] The Daedalus Instructional System.  Version 3.0
from the Daedalus Group, Inc.

     [3] Strictly speaking, simultaneous exchanges are not
possible with InterChange -- the software distributes each
message in its entirety, one after the other.  There may
also be some "lag" in the transmission of messages due to
technical difficulties with the system.  However, for
convenience, I refer to InterChange as a "real-time" system
in this article.

     [4] Factor loadings indicate the degree to which one
can generalize from a given factor to an individual
linguistic feature.  Biber excluded factor loadings below
0.42 in his computation of factor scores for Factor 1
(Biber, 1988, pp. 81-89).


Beaman, K. (1984).  Coordination and subordination
     revisited:  Syntactic complexity in spoken and written
     narrative discourse.  In D. Tannen (Ed.), Coherence in
     spoken and written discourse (pp. 45-80).  Norwood, NJ:

Biber, D. (1986).  Spoken and written textual dimensions in
     English:  Resolving the contradictory findings.
     Language 62, 384-414.

--------- (1988).  Variation across speech and writing.
     Cambridge:  Cambridge University Press.

--------- (1990).  Methodological issues regarding corpus-
     based analyses of linguistic variation.  Literary and
     Linguistic Computing, 5 (4), 257-269.

Chafe, W.L.  (1982).  Integration and involvement in
     speaking, writing, and oral literature.  In D. Tannen
     (Ed.), Spoken and written language:  Exploring orality
     and literacy (pp. 35-53).  Norwood, NJ:  Ablex.

--------- (1985).  Linguistic differences produced by
     differences between speaking and writing.  In D.R.
     Olson, N. Torrance, & A. Hildyary (Eds.), Literacy,
     language, and learning (pp. 105-123).  London:
     Cambridge University Press.

---------, & Danielewicz, J. (1987).  Properties of spoken
     and written language.  In R. Horowitz & S. J. Samuels
     (Eds.), Comprehending oral and written language_ (pp.
     83-113).  New York:  Academic Press.

Collot, M., & Belmore, N. (1996).  Electronic language:  A
     new variety of English.  In S. Herring (Ed.), pp. i

Colomb, G.G., & Simutis, J.A.  (1996).  Visible conversation
     and academic inquiry:  CMC in a culturally diverse
     classroom.  In S. Herring (Ed.), pp. 203-222.

Crook, C. (1985).  Electronic messaging and the social
     organization of information.  The Quarterly Newsletter
     of the Laboratory of Comparative Human Cognition, 7
     (3), 65-69.

Daiute, C. (1985).  Writing & computers.  Reading, MA:

Haiman, J. (1983).  Iconic and economic motivation.
     Language 59(4), 781-819.

Halliday, M.A.K., & Hasan, R. (1976).  Cohesion in English.
     London:  Longman.

Heim, M. (1987).  Electric Language:  A philosophical study
     of word processing.  New Haven:  Yale University Press.

Herring, S.C.  (Ed.)  (1996).  Computer-mediated
     communication:  Linguistic, social and cross-cultural
     perspectives.  Amsterdam:  John Benjamins.

Holmes, J. (1984).  Hedging your bets and sitting on the
     fence:  Some evidence for hedges as support structures.
     Te Reo, 27, 47-62.

Horowitz, R., & Samuels, J.S.  (1987).  Comprehending oral
     and written language:  Critical contrasts for literacy
     and schooling.  In R. Horowitz & J.S.  Samuels (Eds.),
     Comprehending oral and written language (pp. 1-52).
     New York:  Academic Press.

Jones, D. (1943).  Differences between spoken and written
     language.  Journal of Education (London) 70, 207-208.

Kim, M-S., & Raja, N.S.  (1991).  Verbal aggression and
     self-disclosure on computer bulletin boards.
     Washington, D.C.:  ERIC Clearinghouse on Languages and
     Linguistics.  ED334620.

Ochs, E. (1979).  Planned and unplanned discourse.  In T.
     Givon (Ed.), Syntax and semantics 12:  Discourse and
     syntax (pp. 51-80).  New York:  Academic Press.

Ong, W.J.  (1982).  Orality and literacy:  The
     technologizing of the word.  New York:  Methuen.

Peyton, J.K., & Batson, T. (1986).  Computer networking:
     Making connections between speech and writing.
     ERIC/CLL News Bulletin 10 (1), 1-7.

Quinn, C.N., Mehan, H., Levin, J.A., & Black, S.D.  (1983).
     Real education in non-real time:  The use of electronic
     message systems for instruction.  Instructional
     Science, 11, 313-327.

Quirk, R., Greenbaum, S., Leech, G., & Svartvik, J. (1985).
     A comprehensive grammar of the English language.
     London:  Longman.

Schegloff, E., & Sacks, H. (1973).  Opening up closings.
     Semiotica 8, 289-327.

Schiffrin, D. (1985).  Conversational coherence:  The role
     of Well.  Language, 61, 640-67.

--------- (1987).  Discourse markers.  Cambridge:  Cambridge
     University Press.

Tottie, G. (1986).  The importance of being adverbial:
     Focusing and contingency adverbials in spoken and
     written English.  In English in Speech and Writing:  A
     Symposium.  (pp. 93-118) ed. by G. Tottie and I.
     Backlund.  Studia Anglistica Upsaliensia 60.
     Stockholm:  Almquivist Wiksell.

Tottie, G. (1991).  Negation in English speech and writing:
     A study in variation.  New York:  Academic Press.

Werry, C.C.  (1996).  Linguistic and interactional features
     of Internet Relay Chat.  In S. Herring (Ed.), pp.

Winter, E.O.  (1982).  Towards a contextual grammar of
     English:  The clause and its place in the definition of
     sentence.  London:  George Allen & Unwin.

Yates, S.J.  (1996).  Oral and written linguistic aspects of
     computer conferencing.  In S. Herring (Ed.), pp. 29-46.
                      Copyright 1996
   Communication Institute for Online Scholarship, Inc.

     This file may not be publicly distributed or reproduced
without written permission of the Communication Institute
for Online Scholarship, P.O.  Box 57, Rotterdam Jct., NY
12150 USA (phone:  518-887-2443).