Header

THE MODAL COMPLEXITY OF SPEECH EVENTS IN A SOCIAL MUD

Paper to appear in Electronic Journal of Communication,
Summer 1995

                        1. The Site

     There is a growing literature about social phenomena
in text-based virtual realities, or MUDs ("multi-user
dimensions"); for instance, Curtis (1992), Reid (1994),
Bruckman (1992), Dibbell (1993), Cherny (1994a, 1994b),
Kendall (1995), McRae (1995).  This literature has explored
how users gender-swap, role play, build communities and
support groups, punish violators of emergent social norms,
indulge sexual fantasies where bodies are discursive
constructions, and often reinforce gender norms in subtle
ways interactionally.  This paper takes this literature as
background, and examines in detail how a community
functions in text.  In particular, I investigate the
categories of "action" that have evolved to delimit and
communicate availability, interest, and attention in
conversation, provide opportunity for collaborative play,
and open up multiple frames of reference in the MUD and "in
real life."

     My discussion is based on discourses gathered as a
participant-observer on a MOO (object-oriented MUD) over 11
months.  Some of my comments or conclusions may apply to
conversation in other social MUDs, but for the most part, my
observations will be specific to a group of well-practiced
users in a particular speech community (Gumperz, 1972).  The
names of the MUD and its participants have been changed in
this paper to protect their privacy.

     My data were collected during ethnographic observation
of a community called ElseMOO.[1]  The virtual geography
is modelled after a real house and the surrounding town in
Minnesota.  This community has a population of about 30 to
40 fairly regular users, with a core population of about 20.
The median age is 22.5; the gender ratio is about a third
female to two-thirds male.  "In real life" the users are
mostly students, many in computer science, and others are
computer professionals or in network related fields like
information technology.  The regulars are geographically
scattered around North America, with a few in the U.K. and
Australia; there are concentrations of regulars in Seattle
and Boston.  The original founders and many of their MOOing
friends started in Minnesota, but most have since moved to
Boston for school or work, joining other regulars there.

     Since most mudders are regulars on more than one MUD,
it helps to understand the relationship between the
population on EM and those on a few other MUDs.  The EM
population is historically related to the populations of
LambdaMOO (Dibbell, 1993; Curtis, 1993), DeepSeasMUSH, and
InfoPark.  The majority of the users of EM also have
characters on LambdaMOO, and often multi-MUD (splitting
their connection window between the two MOOs) in order to
talk about politics on LM with friends on EM.  Several of
the wizards from LM are regulars on EM.  There are also
several core regulars who frequent DSM, a historically
significant MUD with a community that dates back to the
original MUDs in the U.S. in the late eighties.  Most of
the regulars on EM have been mudding for at least a year,
and in many cases, as long as four years, hence their overlap
with the DSM crowd.  InfoPark (Evard, 1993) is a MUD for
system administrators at Northeastern University, which
several members of EM work on.  The administrator of IP
believes that there has been significant flow of
communicative conventions and culture between IP and EM.
The same is certainly true between EM and LambdaMOO and
DSM.  Membership in multiple communities renders the
conventions in one MUD community more likely to hold for
other, related, MUD communities as well.

     For a researcher interested in conversation and
community formation in MUDs, there are several attractive
qualities to the EM community.  The users are
well-practiced at MUD conversation, spending several hours
every day online, often while working.  The community views
the MUD as an extension of real life, rather than an escape
from it: they make efforts to meet one another in real life,
they talk on the phone, and many make available their real
names and email addresses in their character information.
They do not gender switch or role play in their MUD
personas.  Finally, the community was formed as a place for
hacking on networking and virtual reality, as well as
socializing; the users have a heavy investment in the
technology and many hope to put it to use in their
professional lives.  Although the MUD exists external to
their work contexts, professional work is accomplished on
it; many users report that it is a great place to find
answers to technical problems or to come for design
discussions.  EM is therefore interesting as a proto
computer-supported cooperative work environment, as well.

                 2. Communication Commands

     There are two main modes of communication in MUDs, the
"say" and the "emote," plus various programmatic ways of
producing communicative output that resembles them.   A
user (me, with character lynn, in this case) can "say"
something by typing at her prompt:

     >"hi there.
                
and the output in the virtual room she occupies will be:
        
     lynn says, "hi there."
                
     She may also direct her "say" to another character, to
make it unambiguous who she is responding to in a
multi-threaded conversation in a crowded room.

     >`Tom how are you?
     lynn [to Tom]: how are you?
                
     The other main mode of communication is the "emote,"
which allows body language or other forms of narration.  For
instance, using the emote command or its shorter
abbreviation ":", I can perform an action:

     >:scratches her head.
     lynn scratches her head.
                
Emotes may also be used in other ways: for example, to
reflect states of mind or belief, as an alternative to a
"say."

     >:wonders where that book went that she loaned
          paul.
     lynn wonders where that book went that she loaned
          paul.

     Conversation can also take place "long-distance"
between characters who are not in the same room.  Analogs of
the "say" and emote commands are available for long-distance
communication.  If paul is in another room, I can page him
(line 1) or "remote-emote" (line 4).

     1 >page paul how's it going?
     2 Your message has been sent to paul.
     3 paul pages, "not bad, you?"
     4 >+paul grins.
     5 (to paul) lynn grins.
     6 (from the sunroom) paul grins too.

No one but me sees his answering page in line 3.  Likewise
with his remote-emote response to me (showing which room he
occupies) in line 6.

     To simplify examples in this paper, I have removed
my prompt line from the logs unless the command I typed
was a page.  A table of command types is shown in Table 1.


==========================================================

Table 1.  Summary of Communication Command Types

Command        Example                          Public?
----------------------------------------------------------
Say            lynn says, "hi"                  Yes
Emote          lynn waves.                      Yes
Page           >page paul hi                    No
               paul pages, "hi"
Remote-emote   (from the sunroom) paul grins.   No
               (to Tom) lynn hugs.

==========================================================


     Sometimes communication also involves programmed
functions that can be called by users.  The programs are
commands (known as "verbs") which are either globally
available anywhere or reside on objects in the MOO, like a
chair or a dictionary.  The "Antisocial Commands" are
examples of global commands accessible to anyone anywhere
in EM.  These commands are records of in-jokes in the
community, or useful ritual phrases that communicate affect
responses in conversation.  For instance, if I want to
communicate that I think someone is behaving oddly, I may
"eye" them "warily" using an Antisocial verb:

     1 >eye tom
     2 lynn eyes Tom warily.

As shown in line 1, I type the Antisocial verb, followed by
the name of the person I want to "aim" it at, and the
message shown in line 2 is seen by the occupants of the
room I am in.  I could also type line 2 as an emote,
clearly, but the verb provides a shorthand input form.

     Objects frequently have commands on them that output
some information to the room they occupy when someone
interacts with them.  For instance, if I want to sit on a
chair:

     >sit on chair
     lynn sits on the folding chair.

Objects may only be interacted with by a user in the same
room as the object.  The Antisocial Commands, and the
Carpal Tunnel Syndrome commands I will discuss below, are
available anywhere.  The output from interactions with both
objects and the command set verbs most often appears as if
it were an emoted action.

     The flexibility of the emote, which is not available
in many other chat programs, makes the form of
communication in social MUDs particularly complex.  A
variation on the emote is in fact available in Internet
Relay Chat (apparently added since Reid, 1991), but IRC
does not have perseverative objects and geography (Danet,
Ruedenberg, & Rosenbaum-Tamari, 1994; Holmes & Dishman,
1994).  The emote command in IRC also differs from the
emote in MUDs in that it is not parallel to the "say"
utterance in IRC; it requires extra typing to label an
utterance as an action, while in MUDs the two require the
same effort.  The emote, combined with the social responses
to objects and geography, allows for a complex user culture
on MUDs.  In an indirect argument to support this claim, I
will suggest in section 4 that page conversations, which do
not observe room boundaries, are different from
conversations respecting room boundaries (the virtual
version of face-to-face) because of this emergent user
culture.  Various uses of emotes in conversation play the
major role in this account.

                    3. Types of Emotes

     I divide emote types into five broad categories which
are substantive in terms of their role in determining
content and structure of conversations.  I provide a
summary table for the five here:


===========================================================

Table 2.  Summary of Emote Types

Emote Type          Section Tense     Example
-----------------------------------------------------------
Conventional Action 3.1    present    lynn waves.
Back Channels       3.2    present    lynn nods.
Byplay              3.3    present    Mike pastes Tom's
                                                 lips...
Narration           3.4    present    ls packs for his
                                                 trip.
Exposition          3.5    any        lynn hated the film.

===========================================================


     The first type consists of conventionalized "actions"
like waving when entering a room, poking someone to get
their attention, blinking to announce that one is "awake"
in the room again after a period of idleness.  The second
type are reactions during conversation that communicate
continued attention or understanding, similar to "back
channels" in face-to-face speech, like "lynn nods" or "lynn
grins." The third type comprises the joking by-play during
conversation, usually meta-discourse commentary, like
"Shelley hands some prepositions to Jon." The fourth type
includes emotes describing actions occurring in real life,
as a kind of narrative commentary; these often explain
periods of idleness.  Unlike the first four types
described, which all occur in simple present tense, the
fifth type are emotes of internal or factual background, in
which the tense may shift from the simple present to
another tense.  These may represent states of mind, for
instance: "lynn hated the film."

     There are two other sorts of emote, which are closer
to the "says" than the other emote types described above.
I will mention these here and then not return to them,
because they are more indicative of play with modality than
content.  In the first of these, a verb is formed out of
potential quoted material: e.g., instead of "Tom says,
`cool'" Tom expresses it as "Tom cools."  These
verb-creations are usually derived from short, frequently
produced utterances, like "uh oh," or "oh"; but there are
occasional examples of longer utterances (this was
collected for me by a user who found it odd herself):

     Pete actuallies, calls NES once he gets contact
          info from johnf

     The "say" gloss of this one might be something like:
"Pete says, `actually, i'll call NES once i get contact
info from johnf.'"  (Note the use of the simple present
tense "calls" for future intention.  I will return to this
phenomenon in section 5.)

     In another type of modality play, an exclamatory
utterance surfaces as an emote, but untensed, e.g., "Karen
eep."  This is probably intended to suggest urgency or fast
emotional response.

     Tom incidentally has probably broken `home' for
          people who live underground.
     Karen eep
     *******
     Tom aieee as he discovers how the secret tunnels
          connect to ken's tree

     Tom's "aieee" above simulates a wailing or screaming
sound of distress.  (Interestingly, I have only one case of
"aieee" appearing with the tense morpheme on it versus
dozens of untensed examples, which I would suggest is
because the exclamatory force is weakened by tensing it;
however, addition of complex modifiers apparently doesn't
have that effect, as indicated by the above example.  I do
find occasional examples of it as a say: "Mike says,
`aieee.'")

     These two types of emote represent variation in the
choice of modality, which appears to be systematically
meaningful for the community.  I will now discuss the other
five types in more detail.

3.1 Conventionalized Action

     The first sort of emote I will consider is the class of
what I call "conventionalized" or "ritual" actions.  As will
become clear, they are actions that are classifiable loosely
as conversation openers or closings (Schegloff, 1982).  This
category includes ritual greetings, like "hugging" or
"waving" to someone when they enter the room or leave,
"poking" someone to get their attention, "idling" when one
is leaving one's terminal and not able to participate in
conversation, "blinking" or "waking" when one returns to the
MUD window and becomes active again.

     I will focus mainly on the greeting and goodbye uses in
detail, since they are used most systematically.  I will
also expand my discussion to synonymous "say" greetings for
completeness' sake.   Examples of greeting waves follow.

     Bonny wanders in from the eastern path through
          the parsley trees.
     lynn waves bonny
     Bonny waves.
     Tom says, "this maroon shirt frightens me"
     Karen waves.

     Satire is one of the best ways of finding out what a
community recognizes as conventional behaviors or
categories.  George, one of the characters on EM, wrote
satiric regulations about proper etiquette in certain
situations.  These regulations are considered satiric
because they are exaggerated prescriptions of behavior that
has evolved over time.  No user is expected to follow the
regulations as stated.  Below is a script of me reading a
note containing his parody of departure and arrival rituals
(note that although reasons for departures are polite if
you are active in the conversation, they are not always
given; and there is no even close approximation to the
points under Leaving:2):

     >read 2131.g
     There appears to be some writing on the note ...

     Leaving:
     1) Say goodbye and/or wave.  A reason why is
     optional.
     2) Wait for at least 2 members of the social
     setting to acknoledge that you are leaving or 5
     mins.
     3) Leave.

     Entering:
     1) Wave
     2) Say hello or hi (optional)
     See Regulation 2131.GO for information about
     saying goodbye or greeting people entering the
     social setting.

     (You finish reading.)

(The "read" command typed at the ">" in the example above is
the command that prints the text on the note called "2131.g"
to my screen.  No other user sees the text when I "read"
the note.  The parenthesized "You finish reading." alerts me
that the output has stopped from the note object.)

     3.1.1 Contexts for Conversation.

     The importance of such conventional actions for the
community should not be underestimated.  They help define
and structure interactions, particularly at the limits.
They serve as signallers of contexts for potential
conversations.  Examination of greeting waves or other
forms of greeting reveals complex attitudes towards the
"geography" and idle behaviors.  The majority of entrances
to rooms on EM are accompanied by a wave or some greeting
from the enterer to the room's occupants, and then usually
by some reciprocal greeting to the enterer from the active
occupants.

     Although statistics are of only limited use when
behavior is complex and situated in a natural context
rather than a controlled laboratory setting, I present a
few numbers to illustrate the point.  In two days (16
hours) of logs, out of 71 cases of characters entering a
room I was in, in 54 of the cases (76%) the entering
character issued a greeting, while in 17 cases she/he
issued no greeting upon entry.  (I only considered cases of
regular community members who know the conventions in this
investigation.)

     Cases in which no wave or hello is issued upon arrival
can either be dismissed as simply "rude" or examined to
find a deeper organizing principle behind the use of
greetings (see Schegloff, 1968, for an example of this
strategy, examining cases of unusual speaker order in phone
greeting sequences).  I will attempt the latter, situating
the actions in their contexts.

     In a large sense, connecting to the MOO means entering
a potential conversational context.  Since many people use
login watchers, which alert them to connections and
disconnections, the possibility exists for greetings upon
connection rather than just upon entry to a room.  (The
login watcher is perhaps parallel with a ringing phone in
Schegloff, 1968, although the login watcher amounts to just
an announcement of presence; generally, though, greetings in
MOO differ from phone greetings in that no identification of
the person "calling" is necessary, since the name is
trivially available.  This situation parallels some examples
mentioned in Tang et. al., 1994, where caller id was
available, and greetings reduced to "Hi, Ellen" from "Hi,
this is Bob," for instance).

     Greetings that do occur immediately upon connection
happen in long-distance pages (since few people gather in
the room that serves as connection point).  The possibility
of long-distance communication expands the notion of
conversational context beyond the room one is in; as a
result, the borders of conversations are more flexible, and
this fact may explain deviation from the standard greeting
upon entry to a room.  An example of greeting immediately
on connection follows.  The first line was printed by the
login watcher.  After seeing the connection information, I
paged hello at the > prompt and received a hello in
response.  Penfold then joined the party, but did not
repeat a greeting.

     < connected: Penfold (#83) on Mon at 20:37.
          On-line: 27. >
     Tom dunno yet, is looking.
     >page penfold hi.
     Your message has been sent to Penfold.
     [at 8:38 p.m.]:
     Tom says, "his ongoing problem, he's now blaming
          Delta for it?"
     lynn says, "what prob?"
     Tom says, "he had a dispute with maryanne"
     Tom says, "the fish was selected to mediate"
     lynn says, "oh that, yeah."
     Tom Traceback says, "he was also mloses"
     [at 8:39 p.m.]:
     Penfold pages, "hi."
     Tom doesn't think he likes facial hair.
     Penfold arrives from the eastern end of the
          patio.
     Tom says, "except penfold's"
     Tom eyes himself warily.

     In fact, in several cases, joiners who didn't greet
seemed to be involved in paged conversations with subsets
of the group before joining them.  The general greeting on
entry to the room would have been somewhat redundant.
Other cases of non-greeters appeared to be people who were
not intending to participate, just intending to idle in the
midst of a group.  A few others were cases of people who
had joined other groups before, and greeted people there;
perhaps they felt that an initial greeting to people upon
arriving at the MOO was enough for the day.  Overlap in the
participants between the different groups probably was a
factor as well.  (Note that in ``real life'' it also
becomes awkward to repeat hellos to people one has already
greeted, in subsequent meetings.)

     3.1.2 Departures as Closure.

     A close look at departures supports some of the basic
conclusions above about conversational contexts.  I showed
above that some people view connection as an opportunity
for greeting, even across room boundaries, which ordinarily
define the contexts for conversations.  Disconnection in a
room counts as a departure from conversation, and hence may
require a goodbye, as the following departure from Shawn
shows:

     Karen eeep a spider
     Tom eek a mouse
     Shawn [to Karen]: mmm nice juicy spiders
     lynn [to Karen]: keep it away from me....
     Tom says, "why are wohwait"
     Shawn bye
     [at 9:41 p.m.]
     Shawn has disconnected.

     However, not all disconnections are preceded by
goodbyes or waves; long idle folks may just disconnect
suddenly, may be kicked off by their machines logging them
out after a certain idle period, or bounced off by
call-waiting seizing their phone line.  When active people
leave a room with no wave or bye or statement of intent to
go, it is generally seen as odd (often on EM, the
departing character will be "eyed warily," signaling
unease); but if someone relatively idle walks out or
disconnects without any warning, it is not so unusual.  The
reaction to the departure is dependent on the
conversational involvement of the person.


===========================================================

Table 3. Departure Events in 16 Hours

Departures with goodbye from the departing    23
Departures with no goodbye from the departing 16
------------------------------------------------
Total departure events                        39

===========================================================


     Note that nearly half (16 out of 39) the leaving
events (of which 9 were disconnections) are not preceded by
a goodbye from the person leaving.  This contrasts with the
situation for arrivals, where 76% greeted on entry; idling
habits or connection problems appear to be the reason for
the disparity.

     Of the 16 who left with no goodbye, 10 were idle for
at least 5 minutes before their departure.  One or two
others appeared to be having connection problems.  Three
cases of departure from the room when less than a minute
idle appear to be instances of people trying to avoid
"spammy" (noisy) crowds; in two of those cases, paged
conversation continued with them after they left,
indicating that the conversation they were involved in was
not concluded by their departure.  Another one left quickly
to get something from an adjacent room and returned
immediately--with no goodbye or hello at each juncture,
probably because his participation was merely suspended
momentarily.  Out of the idlers, one entered without a
greeting and idled the entire time he was present in the
room, then left the room with no goodbye; his lack of
entrance hello may have been an indication of his intent to
just idle in the room.

     Leaving or joining a group normally requires an active
(non-idle) human agency which is expected to conform to
social norms when entering a conversational context.
Social norms usually involve greetings and closures.  On
the MUD, events are flexible, however.  Disregarding those
norms, by entering or leaving without appropriate greetings
or goodbyes, may signal a lack of interest in participating
in the conversation (and hence an interest in idling).  The
option of long-distance paged greetings expands the
conversational context from the room to the entire MUD, and
makes the lack of greetings upon entry to a room also a
viable social option.  Technical concerns like connection
problems may also excuse or explain flouting of the norms.

     In her discussion of computer messaging, Murray (1989)
suggests that closings don't occur in her data because
other media take precendence over the computer
conversation, neither party has anything new to contribute
to the topic, or the initial reason for opening
conversation has been resolved.  Phone interruptions or
someone entering an office can cause silence, and lack of
proper closings.  While silence does result from
interruptions in MUD conversations, it's not appropriate to
say that the other media take precedence, precisely;
generally someone interrupted ``in real life'' returns to
the MUD conversation afterwards.  Lack of new information
as a reason for concluding conversation presupposes a very
different, purely task-oriented use of the medium, which is
not true for the community on the MOO I observe.  People in
EM are in a social context, occasionally having technical,
task-oriented conversations there; the social framework is
not forgotten when the task is concluded, however.
Clearly, the motivation for use of a communications medium
has a large role to play in the way it is used, and the
results in Murray (1989) do not carry across to MUD
interactions.

3.2 Back Channel Reactions During Conversation

     The second class of emoted utterances consists of
simulated "back channels" during conversation, like "lynn
nods," "lynn giggles."  Again, I will also consider
synonymous say utterances in this discussion.

     Gumperz (1982) claims that back channels represent
"one common way in which conversational cooperation is
communicated and monitored," and may include nods or other
body movements, or interjections like "ok," "aha," "right."
Back channels, which include the class of "confirmation
feedback" discussed in Oviatt and Cohen (1988), are
important for determining the attention state of an
interlocutor, and establishing whether speaker intentions
have been understood.  In a text-based medium where no
physical cues are available, and interlocutors may be
called away from their desk or terminal at any moment,
these are particularly important; back channel emotes
therefore play a large role in establishing achievement of
mutual understanding and facilitating a sense of
co-presence.  Note that their existence contradicts
assumptions by early computer-mediated communication
researchers that the limited channel resulted in reduced
sense of social presence (Short, Williams, and Christie,
1976) or in less feedback (Daft & Lengel, 1984, 1986).
Users adapt to the medium or adapt the medium to their use,
as suggested by other CMC researchers (Hiltz & Turoff,
1981; Walther, 1992).

     Examples of back channel use in a conversation occur
in lines 3, 9, and 13 below.  (Conventional expressions of
puzzlement, which are actually examples of repair, as in
line 9, provide information about the status of mutual
understanding, so I include them in the class under
consideration here.)

     1 Tom says, "only in look self"
     2 Karen says, "cool"
     3 Karen nods.
     4 Karen thot so, but
     5 Karen says, "oh"
     6 Karen says, "there was another reason"
     7 Karen sigsh
     8 Karen wanted name j---- jdesc
     9 Tom says, "huh"
     10 Karen
     11 ----- --Karen
     12 Karen's description
     13 Tom nods.
     14 Karen says, "now i can't"

In this example, Karen is trying to describe how she wants
some text to be laid out: lines 10-12 are her attempt to
graphically represent the fields she wants, which consist of
a name, a line of hyphens, and a description underneath it.

     Evidence for the importance of these utterances can be
found in the fact that many of them have been translated
into "verbs," which can be typed at the prompt, for ease of
production.  For instance, the "Carpal Tunnel Syndrome" set
of commands includes "gg" which outputs "lynn giggles" (if
typed by me), "nd" for "lynn nods," "h" for "lynn hehs" (a
simulation of laughter), "gr" for "lynn grins," "si" for
"lynn sighs," "sm" for "lynn smiles," "/" for "lynn says,
`?'" (which approximates a questioning look, according to
users' interpretations).  (Other commands in the CTS set
include "waves" and "whuggles" which fit into the first
category of emotes discussed.  See Cherny, 1994a, for a
discussion of whuggling.)[2]

     The definition of a "back channel" in the literature
is vague; Duncan (1973) suggests that the difference
between turn and back channel becomes uncertain when
restatements of speaker position or requests for
clarification are considered.  I therefore focus on a
closed class of short utterances and "actions" which seem
to fulfil the definition in Gumperz (1982), plus allowing
in conventional expressions of hearer confusion.  Rather
than arbitrarily exclude says that apparently function the
way these emoted responses do, I expand my focus to include
some say responses like "yeah" or "yes" or "ok."

     I did pattern matching in two-person conversations for
the back channels "nods," "hsm," "hmm," "hrm," "ok," "oh,"
"yeah," "yes," "?," "giggles," "laughs," "grins," "smiles,"
and "hehs," which are conventional responses of either
puzzlement, understanding, or appreciation for an
interlocutor's remarks.


===========================================================

Table 4. Back Channels Against Utterance Rate

Tom:  time point   utts  words/utts lynn's back channels
--------------------------------------------------------
        10          35    6.8        5
        20          74    7.4        4
        31          52    7.5        0
        43          28    4.8        4
        53          33    8.1        5
        64          39    6.3        1
        78          22    7.1        6
        88          29    6.2        2
        98          44    6.0        1

lynn: time point   utts  words/utts  Tom's back channels
--------------------------------------------------------
        10          37    11.4       12
        20          49    10.1       5
        31          28     5.4       5
        43          17     7.8       4
        53          24     5.0       5
        64          26     6.3       1
        78          20     6.2       7
        88          32     7.2       5
        98          24     6.6       0


============================================================


     Table 4 illustrates two halves of the course of one
long conversation within a single room on June 20, 1994.
The left column corresponds to time points within the
conversation.  The "utts" column shows the number of
utterances in each time period by speaker, since the last
time period.  The third column, "words/utts," gives a
measure of the rate of speech, the number of words in the
time period divided by the number of utterances in the time
period.  The numbers in the last column correspond to the
number of back channels received in each time period from
the interlocutor.  (I.e., the table of Tom's utterances has
lynn's back channels in it.)

     Periods with low utterance counts for both conversants
(43 minutes, 78 minutes) nevertheless show the presence of
several back channels.  If back channels were only
functioning as an indicator of comprehension of a speaker's
plans (as, for example, the treatment of some clue words in
Grosz & Sidner, 1986, implies) or of completion of a
discourse structure (Schegloff, 1982), there should be
fewer back channels during periods with little interaction.
The appearance of these signals in such periods probably
indicates that a potential interlocutor is attending and
may be available for more extended conversation.  The
conversants are in a "continual state of incipient
conversation" (Schegloff & Sacks, 1972), analogous to that
achieved by the two-party situations Goffman (1963)
describes: "communication arrangements that seem to lie
halfway between mere copresence and full-scale
co-participation."

     In the following excerpt from the period around 43
minutes, the conversation has trickled off the previous
topic which was a somewhat tense one, and the interlocutors
are registering their continued alertness, even while they
document their actions in real life.  In lines 1-2, I
illustrate that I am reading email in another window, and
paste a section from one message.  After a desultory
exchange on that topic, Tom begins playing with names in
thought bubbles (lines 7-9) and then reports singing, a
common practice while listening to music and mudding at the
same time.  I respond with back channel responses in lines
12 and 14, indicating I am still alert, before making
another desultory conversation attempt in line 15.

     1 lynn sees OJ all over the popcult list, of
          course.
     2 --------------------------------lynn-----------
     In any case, thanks to OJ, Al, and the LA chopper
     teams and reporters for providing all of us cult
     studs folx with yet another a perfect
     Baudrillardian moment...
     --------------------------- lynn stops pasting --
     Done @pasting.
     3 Tom says, "al?"
     4 lynn dunno.
     5 lynn says, "media coverage somehow."
     6 Tom says, "ah"
     7 Tom . o O ( oj et al )
     8 Tom . o O ( woj simpson )
     9 Tom . o O ( homer j simpson )
     10 Tom [sings]: who throwed lye on my dog?
     11 Tom wonders if he could fall asleep.
     12 lynn says, "hmm. go home and sleep."
     13 Tom is home.
     14 lynn says, "oh."
     15 lynn says, "I wonder why I keep dreaming about food."

     The context-sensitive nature of back channel
interpretation (which I can hardly do justice to here; see,
e.g., Heritage, 1984; Cherny, 1995) is supported by the
distribution of back channels during periods of rapid
conversation: the number of back channels given by speaker
A to an interlocutor B increases when B's number of
utterances increases, but stays low if A's utterance rate
increases in parallel (see times 21 minutes and 63 minutes
on Table 4).  The parallel rates of increase suggest that
the back channels' confirmatory function (or indication of
attention) is being fulfilled by other utterances which
more complexly express comprehension and attention.

3.3 Emoted Byplay
                
        These emotes of action usually occur in a multi-party,
joking context.  They serve a similar function to the back
channels described above, in that they signal attention to
conversation, but they may involve other characters in a
more phatic, often teasing manner.  Sometimes the imagery
is quite violent, or unrealistic (e.g., "Karen detonates a
low yield nuclear device over Penfold"), but it is
understood to be just play, given the conversational
context in which it occurs (see Cherny, 1994a).  The overall
impression these actions give is of a cartoonish unreality,
a rubberness of boundaries.[3]

        An example occurs in line 3 below:

        1 Tom says, "....can you speak, Mikey, without paste?"
        2 Kit [to Tom]: EAT MY PASTE
        3 Mike pastes Tom's lips together.

     Play in the MOO is a highly cooperative behavior,
usually happening during multi-party conversation and often
triggered by or directed at objects in the environment.
Objects and their modes for interaction form a partial
conversational context, which distinguishes MUDs from IRC or
other chat environments (Holmes & Dishman, 1994; Danet et
al., 1994; Evard, 1993).

     Interactions with objects in conversation may consist
of interactions with "verbs" (MOO code) on objects in the
room, references to objects in the MUD room without use of
coded verbs on the objects, or references to imaginary
objects conjured during the discourse (including bodies).
The real world may also be referred to in a joking manner
during conversation.  I discuss examples of each type in
the sections below.

     3.3.1 Interactions with Coded Objects.
                
     Objects serve as a record or embodiment of history,
giving a permanence to some events.  Stories are associated
with some objects: their reason for creation, their origin
on other MUDs, their original author.  The high5ing glove
on EM is a copy of an object made by the character lew on
a historically significant MUD (DreamScape).  The glove is
used in a congratulatory fashion.  Here, Mike uses it in
lines 4-7 because Henry comes up with a clever name
transformation to apply to Mike (the copyright notices were
attached when it was copied by a character on EM, with
lew's permission):

     1 Mike takes a AI from his chest and turns Henry
          into a hairy!
     2 Henry takes a FA from his chest and turns Mike
          into a fake!
     3 Delak takes a GA from his chest and turns Finder
          into a gander!
     4 Mike picks up the the special high-fiving glove
          and slips it onto his hand. (c) 1992 lew
          industries.
     5 Mike leaps up into the air and high5s Henry with
          a resounding >THWACK<.
     6 The high-fiving glove on Mike's hand glows gently
          for a moment, then returns to normal.
          (c) 1992, lew industries.
     7 Mike suddenly flings a large glove to the ground
          and shouts, "I CHALLENGE YOU, BENJAMIN J.
          TOAD!" You nod your head in solemn approval.
          (c) 1992, lew industries.

The entire interaction, from picking up the glove to
flinging it down, is a coded script on the object, which is
output to the room by typing a short command ("high5
 with glove").

     The vase in the EM living room dates back to another
historic joke from ChaosMUD, in which a vase object
transmitted everything said in the main party spot into
another room.  Apparently the original joking line for the
vase worked like this:

     thelfar speaks into the vase.  "lew suspects."

It evolved into a message referring to agents, possibly
because of a later joke on DreamScape (a MUD frequented by
the same crowd that hangs out on DeepSeasMUSH now).  Now,
on EM, discussion of agents in the living room is likely
to cause someone to type the command "identify  to
vase," resulting in the output message visible to all in
line 8 below:

     1 Mike says, "which one?"
     2 lynn highly recommends it for a giggle.
     3 lynn [to Mike]: "What's An Agent, Anyway?"
     4 Kit walks in from the sunroom.
     5 Shelley waveys Kit
     6 Ted |  lynn [to Mike]: "What's An Agent, Anyway?"
     7 Ted |  Kit walks in from the sunroom.
     8 Ted whispers into the vase, "I think Kit is an
          agent."

(In lines 6 and 7, the vertical bar represents a "paste,"
indicating that Ted is quoting the previous lines 3 and 4.
He does this to motivate his action in line 8.)

     3.3.2 Interactions with Uncoded Objects.

     The dual identify of character as user and character as
object to be played with is source of many joking exchanges
(see Cherny, 1994b).  Several commands on EM produce a
random list of characters and other objects in their output;
for instance, if I typed the right command, people in the
room with me might see a (nonsense) output utterance like
the following:

     lynn thinks we should have some kind of INTELLIGENCE
          TEST to get in here so people like Tom, the
          Library Browser, the Last Exit For The Lost, and
          the Generic Combination Lock can't BLUNDER ALONG
          and BOTHER ALL OF US AGAIN.

("The Library Browser," "the Last Exit for The Lost," and
"the Generic Combination Lock" are objects in the MOO,
randomly picked to parallel the name I gave to the command
I typed.)

     Objects in the environment which cannot be interacted
with via code on them can nevertheless feature in group
play.  Often they are targeted with the Antisocial commands,
which record community in-jokes with amusing (often
graphic) utterances.  One day's humor revolved around
vilifying the trees someone had put in the park.  Most of
the actions or quotes directed to the trees originated from
the Antisocial commands (some of which were invoked by me at
the prompt, e.g., "shake" in line 12).

     1  Jon stands up from the tree stump.
     2  Jon [to the trees]: Come to Perkins!
     3  Jon [to a tree]: Come to Perkins!
     4  Jon giggles
     5  >!The trees groan and pull their roots out of
          the ground; they advance on Jon threateningly.
     6  The trees groan and pull their roots out of the
          ground; they advance on Jon threateningly...
          --lynn
     7  >eye me
     8  lynn eyes herself warily.
     9  Ray giggles
     10 Ray nails a tree down.
     11 Jon detonates a low yield nuclear device over a
          tree.
     12 >shake trees
     13 lynn shakes the trees.
     14 Ray spraypaints "WAKE UP" on a tree in dayglo
          orange.
     15 Ray giggles
     16 Jon takes off and nukes a tree from orbit.  "It's
     the only way to be sure."

(In line 5, I "spoofed" the behavior of the trees, which
resulted in the line shown in 6, with my name appended as
the "author" of the spoof.  Spoofs allow production of
utterances that don't begin with the author's character
name.  They are ordinarily used in playful interactions.)

3.3.3 Interactions with Imaginary Objects.

Contextual suggestions may briefly evoke imaginary
objects in the discourse, producing a cartoon-like
atmosphere at times.  The play with these unreal objects is
often collaborative, like in this case, where Marie responds
to the gift of imaginary flowers:

     Marie [to Patrick]: caller is "this" from the calling
          verb.  You sure you want caller and not caller
          perms?
     Patrick ose!
     Patrick duhs.
     [at 2:49 P.M.]:
     Patrick YAYS.
     [at 2:50 P.M.]:
     Patrick pulls a handful of roses out of the air and
          presents them to Marie in appreciation of her
          help.
     Patrick feel st00pi, to.
     Patrick kicks his silly S1 keyboard.
     [at 2:51 P.M.]:
     Patrick [to Marie]: thanks.
     Marie stops and smells the roses.

     It is out of the ordinary for characters to indulge in
extended, solitary interactions with imaginary objects, as
occurred in this example, when a guest on the MOO enacted
a biker persona fantasy at length.

     The Canadian guest lights smoke:
     Tom eyes the Canadian guest warily.
     lynn backs away from smoking.
     [at 2:10 P.M.]:
     Karen too!
     Karen is allergic
     The Canadian guest removes whisky bottle from pocket
          and takes a long hit:
     The Canadian guest throws smoke out window and returns
          bottle to pocket:
     Robin says, "hey how was the brunch thingy?"
     [at 2:11 P.M.]:
     Tom [to Robin]: ask carrot
     The Canadian guest runs hands threw greay hair:
     Robin [to Tom]: okay.
     [at 2:12 P.M.]:
     Karen whispers to lynn, "what a dea guest"
     Robin [to lynn]: gil is pretty normal...
     Karen whispers to lynn, "what a dear guest, that is"
     The Canadian guest removes biker jacket:
     Tom [to the Canadian guest]: out of curiosity, why are
          you using : to end sentences?
     Karen too Tom!
     Karen was just typing it
     lynn grins.
     The Canadian guest drops jacket on floor
     Robin hehs.
     [at 2:13 P.M.]:
     Tom says, "what floor?"
     Karen [to Canadian guest]: did you notice we're in a
          ditch?
     The Canadian guest bottle in jacket breaks
     Karen says, "there's no floor or window"
     Tom eyes the Canadian guest warily.
     lynn eyes the Canadian guest warily.
     Robin peers at the Canadian guest suspiciously.
     The Canadian guest says, "when you drink and smoke as
          much as me, you loose track of things""
     [at 2:14 P.M.]:
     Tom feels like he's witnessing a really clumsy strip
          tease.

     Tom later told me that the annoying aspect of that
guest's interaction was his lack of interest in or reaction
to the other participants in the room.  It also appears to
be significant to the regulars that the guest does not pay
attention to the characteristics of the room they are in,
which is a ditch at the side of a road.  This kind of
imaginary play usually occurs within a larger discourse
context, involves multiple parties in the creation of the
fantasy, and often involves the characteristics of the room
as well.

     3.3.4 Real Life Imaginary Actions.

     In another subcategory of byplay, a character
describes what she could be doing in real life.  These
actions don't involve the virtual environment.  The most
common, perhaps, is "LOL," which stands for "laughs out
loud," but frequently the speaker is not in fact laughing
audibly in real life.  It is certainly difficult to tell to
what extent reported real actions are occurring or not; but
in at least one instance, Tom's typist was revealed by
another mudder in the room with him in "real life" to be
quite silent despite the reports of "Tom LOL" in the course
of MOO conversation.  Another possible case:

     Damon writes down "keep lights on when with
          inga"

     These emotes are often distinguishable from the next
category of emotes by the fact that they are responses to
the on-going MOO conversation, although affect responses
like "LOL" remain ambiguous.

3.4 Narration of Real Life Actions

     The next major category of emotes is the Narration
category.  It is quite common for users to document in real
time their actions in real life, while they MUD.  In this
example, ls packs his computers for a trip while mudding.

     1 ls starts packing for his argonne trip.
     ...
     2 ls pulls out his second duo.
     3 lynn says, "you have two?"
     4 paul [to ls]: show-off
     5 ls says, "uh, yeah."

     Users of this narration category of emotes may
occasionally be motivated by need to explain distraction,
lack of timely response, or departure from the MUD
conversation; in other cases, they could be providing
conversational openers; or they could be motivated by a
desire for autobiography in the text window and in the
memories of the people experiencing the MUD conversation.
Some users seem to connect solely to document their daily
work actions, hardly noticing remarks directed to them or
conversation around them.

     Interestingly, emotes about real life actions still
occur in simple present tense, as if the actions are going
on simultaneous with the typing.  In some cases there is
overlap, as in the case of ongoing activities in real life.
For instance, evidence that the activity of talking to
Brock continues beyond the typed narration in line 1 below
is found in line 2, where Shelley describes how Brock
sounds (present tense).

     1 Shelley talks to Brock.
     2 Shelley says, "he sounds, um, i dunno...disappointed
          or something."

     Perhaps oddly, simple present tense can also be used
to describe events that have already occurred in real life.

     Marie wakes, scrolls, giggles
          "$code_utils:boot_player_if_needed"

In this case, Marie describes waking after being idle,
reading her scrollback in the MOO window, and giggling at a
programming line quoted in scrollback.  Her narration of
the entire sequence of events occurs in one line.  We
interpret these as sequential actions because narrative
time is understood to progress with each verb that reports
an achievement or accomplishment in a sequence (Dowty,
1986).[4]  (Achievements and accomplishments describe
punctual events, like reaching the top of a mountain, or
waking.  For further discussion of aspectual classes and
use of present tense in MUD discourse, see Cherny, 1995.)

     These emotes can also describe intentions before they
become actions.  In the next example, Tom leaves for a
meeting.  He reports his intention to "really disconnect
this time" in simple present tense, before doing it.  He
even says goodbye before disconnecting.

     Tom says, "welp"
     Tom says, "time to go back to dealer"
     Tom really disconnects this time.
     Tom says, "bye"
     Tom heads for the eastern end of the patio.
     < disconnected: Tom (#73) on Wed at 16:51.
          On-line: 21. >

     In the analogous case below, note the similarity to
the stylized use of the first person present tense in the
formulaic spoken "And I quote:" --which indicates
immediately upcoming quoted material.

     Tom quotes the rest of the paragraph about Robin
          (whose name rings a bell, but he doesn't
          think he knew her):
     Tom | Robin Weiland continues to work on her
          writing in Minneapolis.  In the midst of a
          financial scandal, half of the Paydirt staff
          was laid off, giving her time for writing
          and career planning.

     The oddity of the utterance "Pete actuallies, calls
NES once he gets contact info from johnf" is that the event
intended (the calling) is nowhere near about to happen; but
I suggest that it is in simple present tense to communicate
Pete's strength of intention.  He indicates that he really
will do it, so it can almost be considered an accomplished
fact already.   This is similar to some uses of the present
to denote future in non-MUD speech, like "I leave for Utah
on Monday."

3.5 Emotes of Background or Exposition

     Emotes in this category differ from the other types in
that these need not be in the simple present tense; they
are usually statements about the attitudes, beliefs, or
relevant background of a user in a conversational
context.[5]

     These emotes seamlessly fit into conversation as
if they were uttered as says: they can be responded to as
if they were "spoken" as says, and they can occur in
response to says. Often they show first person speech-like
properties, like being directed to another speaker,
containing second person pronouns, and even showing
dysfluencies similar to those frequently (intentionally)
introduced in says.  E.g., line 5 below:

     1 lynn [to Damon]: so Kit thinks you and I would
          be a cute couple.
     2 Damon says, "um"
     3 Damon says, "how nice"
     4 lynn laughs.
     5 Damon hasn't, well, met you, lynn

     The choice of emote modality over say modality for an
utterance is no doubt significant, but a complete analysis
of the meaning difference will be deferred for now.  I
suggest that it has something to do with tone, via
manipulation of perspective.  The omniscient narrator's
voice suggested by third person is more distanced, and
perhaps feels more authoritative.  However, for some users,
the emote modality is used for long stretches of
conversation.

     Marie wants chocolate and M&M's make her teeth
          hurt.
     Marie's house mate is allegedly buying instant
          chocolate pudding at the store.
     lynn says, "switch to fudge."
     Marie nods.
     Marie has been eating hot fudge topping straight
          out of the jar :-)
     lynn says, "aieee, wow."
     Marie might be out of it now tho.
     Marie is like hardcore.

     To give a general feel for the regularity of this sort
of emote in conversation, I include breakdowns of 3
two-party conversations showing the ratio of says to emotes
in Table 5.  The emotes are broken down into emoted back
channels, which include utterances like "lynn hsm"
(indication of thought) and "Shelley nods," and the
remainder of the emotes.  Interestingly, in each case, the
emoted back channels amount to almost half of the total
emotes.  Playful action emotes (section 3.3) are most
common in multi-person parties, rather than the more sedate
conversation that occurs between two people alone, and
emotes narrating real life actions are most common in
semi-idle cases, so it is likely that the remaining
non-back channel emotes are of this expository background
category.


============================================================

     Table 5.  Emoted Back Channels, Other Emotes, and
               Says in Two-Party Conversations

     back channel emotes:  3       3
     other emotes:         5       8
     says:                25      17
                        ---- -------
                        lynn Shelley (Nov 30, 14 mins.)

     back channel emotes:  30      23
     other emotes:         29      69
     says:                102     261
                          ----    ---
                          lynn    Tom (July 13, 109 mins.)

     back channel emotes:  17      33
     other emotes:         17      52
     says:                142     192
                         ----   -----
                         lynn   Karen (Nov 28, 55 mins.)

==========================================================


         4. Page Conversations and Virtual Geography

     The term "page conversation" generally refers to a
private conversation with both pages and remote-emotes
occurring in it.  Page conversations often occur when a
user has a quick topic to discuss with someone in another
room and neither wants to leave their current party to join
the other.  Since page conversations are private, and not
detectable by other users, they may be used for private
topics of discussion as well; even the fact of an
interaction between two characters may be hidden if the
page modality is used, since the two characters never need
be in the same room together to converse.  (Private
conversations can obviously occur between two characters in
a room alone together, but their activity may be noted and
provide a source for comment or speculation.  Two
characters alone in a room are generally assumed to be a
private party for two, in fact; it is common for a third
person to page and ask one or both if the party is private
before attempting to join.)

     Page conversations are interesting because they do not
observe the room/geography boundaries that normally
determine contexts for conversations.  In effect, they
become dyadic "channels," like found in IRC or other chat
programs.  Some discussion of how page conversation is used
may throw some light on how the geography affects users'
experience of MUDding.  The following comments are based on
my own page conversations and informal questions to other
users about their page conversations.

     With respect to greetings, paged greetings do occur,
but not commonly in my experience.  Usually the paged
greeting occurs when someone is particularly looking for
another user, either because of close friendship or a need
to discuss some topic.  Closures in page conversations
generally occur when one person is explicity withdrawing
from the MUD, to idle, or to disconnect.  Otherwise, page
conversations tend to dangle without closings. This
contrasts with the use of closures when people leave rooms,
but do not become idle or disconnect.

     A statistical comparison of the ratio of back channels
to other utterances in several (7 page, 10 non-page)
conversations between myself and four other people revealed
that back channels are significantly absent from
long-distance page conversations (T-test, p < .05).[6] This
finding is non-intuitive at first glance, because emoted
commands are available in long-distance conversation mode
via the "remote-emote" command.  However, a closer look at
conversation within the same room reveals that the duration
of the conversation is usually longer compared to that of
page conversations,[7] whereas page conversations usually
cover only one or two topics (one user reports preferring
to keep page conversations to short question-answer
sequences, rather than prolonged exchanges).  Often a pager
joins her interlocutor and the conversation then continues
within the same room.  One conversational partner in my
logs claims to feel more co-presence in non-paged
conversation.  The tendency to use back channels probably
creates this feeling.

     Since page conversations don't allow interactions with
objects or room characteristics, and dyadic conversation is
much less prone to collaborative play, there appear to be
fewer byplay emotes in page conversation.  A possible
exception to this may occur in "netsex" (McRae, 1995)
conducted in pages; because a page conversation is
private and dyadic, other users cannot "walk in on" netsex
encounters conducted in pages and remote-emotes.  (One user
on EM used the page modality for netsex, reportedly.)

     Likewise there appear to be fewer narrative emotes of
the type in section 3.4 during page conversations.
Narrative emotes often occur between periods of idleness,
as a way of reinitiating conversation; people in page
conversations do not "idle together," apparently: they lose
contact when one or both becomes idle.  Expository emotes,
however, are quite common in page conversations.  These
emotes, and back channel responses, are the main uses of
the emote in page conversation, I believe.

     When asked about their feelings towards page
conversation vis a vis the virtual geography metaphor that
normally governs conversations in MUDs, users reported that
they generally preferred room conversation, for a variety
of reasons.  Having a page conversation with someone while
also conducting public room conversation can be confusing,
and may lead to mistakes in modality (saying something "out
loud" that was intended for the page conversation, for
instance; particularly worrying during netsex).  Room
conversations give users access to objects, like furniture
to "sit" on, or toy objects to play with.  The geography
also gives a sense of "connected" or "local" space... users
often idle in rooms that are adjacent to party spots, for
instance.

     The virtual reality (and geography) metaphor matters
more to some users than others.  Bothered by the violation
of metaphor that pages cause, one user (George) created a
radio object that "transmits" pages to another user who has
a radio.  The resulting page message looks like this:

     On your radio, Mike hails, "hey, your boss is here"

Not all users have radios or even know how to work them,
though.  Some users claim they are quite comfortable paging
people when they are in a room alone, on the other hand,
and not likely to be distracted by public conversation;
this is common on LambdaMOO, for instance, where many
people avoid public rooms because of the noise level
("spam").

     In conclusion, as Holmes and Dishman (1994) report,
MUDs provide "an explicit spatial context" for conversation
which chat programs do not have; similarly, in MUDs there
is richer narrative play within the spatial context than
found in most chat programs, and I would argue, a broader
range of conventionalized action types and more flexible
self-presentation in conversation.[8]  User preferences for
room interactions over page conversations and differences
in the types of emotes that occur in the two sorts of
conversation support this claim, although a fuller
investigation of actions in IRC and page conversations
other than my own are surely called for.

              5. Verb Tenses and Speech Acts

     To summarize so far: I have illustrated the five main
categories of emotes, and how they play a role in
conversation and in structuring interactions.  Emotes,
combined with social responses to persistent geography and
objects, can be used to signal alertness and willingness to
participate in conversation, supply feedback of the sort
normally found in back channels, to play, to narrate real
life activity which might cause distraction from the MUD,
to provide background exposition in place of says.

     The many uses of emotes appear to reflect a rich
culture of practice and use on MUDs, but there remain a few
unanswered questions raised in the preceding sections.  Why
are almost all emotes in simple present tense?  How can a
user describe events in any complexity given only one
tense?

     One possible explanation for the prevalence of present
tense is that the convention is inspired or otherwise
historically based on the fact that MUD system responses to
user commands (such as a navigation command, like go north)
are messages in present tense.  The following examples
are output presented when users type commands to interact
with the virtual geography or objects:

     Guest steps off the catwalk.
     *********
     Guest goes east.
     *********
     Ferris heads for the eastern end of the patio.
     *********
     Plaid_Guest wanders up the street to the west.
     *********
     Tom picks up TomTraceback's Business Card.

     A diachronic explanation is not enough, however, to explain
how the present tense is used now in the discourse.

     The persistence of present tense suggests a similarity
to special speech registers like sportscaster commentary,
in which present tense is adhered to while describing
ongoing activity (Ferguson, 1983).  Oral narrative often
shows regular use of the present tense, also known as
conversational historical present in this context (cf.
Schiffrin, 1981; Wolfson, 1982); for instance: "So I go
to the store, and as I'm walking in, this guy says to
me..."  Supposition as to the reason for the use of present
tense in narrative usually involves hypotheses about the
story teller being so wrapped up in her story that she
recreates it as she lived it.  The historical present in
narrative usually alternates with past tense, however, and
the events described are past.  Emoted actions in at least
the first three categories discussed are not descriptions
of past events, but of actions co-temporaneous with their
appearance on the screen.

     The emoted actions in section 3.1 to 3.3 suggest the
sort of speech acts that Searle (1969, 1989) and Austin
(1962) discussed, called "performatives."  Performative
speech acts are utterances that usually occur in the first
person, simple present tense, indicative, and via their
utterance perform some action: e.g., "I hereby pronounce
you man and wife," "I promise I won't spend that dollar."
Despite being in the simple present, they do not describe
habitual or generic actions; they describe an event that
occurs at their time of utterance, by virtue of their
utterance.  Emoted actions seem to be related to this class
of speech acts, albeit from third person, being conjured as
"events" solely by their utterance.  (I follow Searle,
1989, in not including indirect speech acts in my
discussion here.)

     Intentionality is a major component in Searle's (1989)
account.  He argues that a class of verbs with
intentionality as a component must be recognized; that one
cannot perform a performative action without intention to
do so.  "Manifestation of the intention to perform the
action, in an appropriate context, is sufficient for the
performance of the action." (Searle, 1989, p. 551).
Furthermore, the performative verbs are self-referential,
in that they describe their own actions and execute them at
the same time.  If I say "I promise to come home at noon,"
I both describe the content of my promise and make it.

     Searle concludes that the explanation for performatives
does not lie in the meaning of the verbs themselves, but in
the world itself.  "If God decides to fry an egg by saying
`I hereby fry an egg,' or to fix the roof by saying `I
hereby fix the roof,' He is not misusing English.  It is
just a fact about how the world works, and not part of the
semantics of English verbs, that we humans are unable to
perform these acts by declaration" (Searle, 1989, p. 554).

     It is tempting to conclude that the world of MUD
conversation has the status of the imaginary world Searle
proposes for performatives.  Every uttered action is
understood to occur at utterance, in the context of the MUD
conversation.  However, the participants understand that no
real action has occurred, that in a sense these are just
descriptions of actions, carrying social significance
within the MUD discourse.

     There is some evidence that users view such emotes as
events that "happen" (even if only as a communicative
action) as soon as they are emoted, and are thus
nondeniable, like performatives.  One woman described to me
her upset at being hugged by a guest she didn't know and
then her attempt to negate the event:

     1 The guest hugs Karen.
     2 Karen is NOT hugged by Guest.

Despite her attempt in line 2 to retract the hug, another
character later referred to "the guest who hugged her,"
indicating that he perceived it as not deniable, or at
least, nondeniable by her.  In some sense, the action
occurred as soon as the message showed up on people's
screens.

     One problem Searle (1989) and Verschueren (1994)
tackle is why there are no performatives for verbs like
"hint," "boast," or "lie" which sound as if they ought to
parallel verbs like "promise" and "order."  Searle
concludes that they cannot be used performatively because
they imply that their actions are not performed explicity
and overtly, which is required of performatives.
Verscheuren expands this into a theory about distance
between the description of the action and the action the
utterance performs.  In the utterance "I lie to you that I
am done with my PhD," the action performed is the lying
about a proposition, and the description is the statement
that it is a lie.  He claims there must be no "distance"
between the two in order to be performative.  To claim to
"lie," for instance, is to describe an action as negatively
valued and insincere; the choice of the word "lie" creates
an evaluative distance between the description and the
action intended, causing the performative attempt to fail.
"I order you to go," on the other hand, succeeds because
the action is an order, and the description is of an order
as well, with no evaluative distance or insincerity
implied.

     Interestingly, there are occasional uses of verbs like
"lie" in the discourse of the EM community:

     Ray lies, "I'm awake, I'm awake!"

     In fact, this utterance probably succeeds partly
because in fact it is not uttered by the user of the
character Ray, it is a programmed response that is
triggered when the name "Ray" is mentioned in a room Ray is
idling in (an "idle twitch").  In fact, his character is
lying, as the utterance states.  Any negative evaluative
distance in this case is confined to the third person
description of action, rather than the utterance "I'm
awake" itself.  That description is important for
communicating that the utterance is not an ordinary one
typed by an active user.

     Note that this utterance would not succeed as a simple
"say" from Ray's character: "Ray says, `I lie, I'm awake.'"
The third person emote allows a distance between the
description and the action which is not allowed in first
person performatives, where the two must coincide.  This
perspectival distance allowed by emotes complicates and
enriches the sorts of actions that can be undertaken in the
text conversation.

     Verschueren (1989) discusses some speech acts which
are not first person simple present tense, which he calls
semi-performatives:

     1. Le porteur declare etre majeur.
          [The holder (of this ticket) declares
          that he/she is over 18.]
     2. You are dismissed.
     3. Passengers are warned not to lean out the
          window.
     4. I am asking that you to do this for me, Henry,
          I am asking for you to do it for me and
          Cynthia and the children. (Searle, 1989)

     Number (4) he dismisses as not containing the
"content" of the request and therefore calls it a
description of the request rather than a making of the
request itself.

     Emotes in category 3.3 and higher suggest that a
simple performative analysis for all emotes is not
sufficient.  The events described can be more complex than
simple co-temporaneous actions.  The MOO example below
consists of a description of action:

          Ted says one thing, then the opposite, and
               agressively asserts that both are true.
               He then annoys you for a while, just
               before boasting of his SPARCbook like
               you care.

     Ted describes himself as "boasting," but he is in fact
describing an action he is not explicitly performing (since
one cannot boast solely by saying that one boasts), which
puts this example in the category of semi-performatives
like Verscheuren's (4) above.  Emotes like this are a
common occurrence, some of them more explicitly
performative than others.  The more performative types
include "note," "observe," "ask," "tell," "wonder," which
usually take as subordinate clauses a proposition, unlike
the boasting case above.  The proposition appears to
accomplish the action, unlike the subordinating verb on its
own.

     Sandy notes that the power elite IS NOT linked...
     *********
     Shelley observes that she's about conferenced out
          and would really like to go to the espresso
          bar across the street and talk about
          something else.
     *********
     Marie thanks lynn for signing

     The "boasting" example is not alone in its
meta-discourse character; in the following example, Tom
describes the conversation in line 9.

     1  You see a car scream by on Hwy 169, you hear a
          SLAM, and Mike zings into the ditch and
          lands with a painful THUD.
     2  Ray laughs
     3  Tom says, "zing"
     4  Ray says, "THUD"
     5  lynn says, "THUD"
     6  Mike says, "WHAM!"
     7  lynn [to Ray]: hmmph
     8  Shelley says, "THUD"
     9  Tom and Ray and lynn reenact the event.
     10 Ray wins by virtue of only being a few yards
          away from em.ccs
     11 lynn [to Tom]: using Mike.
     12 Tom [to Ray]: It must be a real pain for you
          not to be able to blame lag when you lose
          anymore.

(em.ccs is the machine that the MOO resides on, and Tom
refers to the lag of the network in line 12.)  Note that
Tom describes it in present tense, suggesting the
reenactment is not over. Note also that Ray claims he
"wins" in reference to getting his "THUD" out before mine
in line 5.  He has already won, but he describes the event
in present tense.  Present tense seems to be fine for
actions that have just ended as well, as seen under 3.3.

     Considering the range of "action" types that are
performed in third person simple present tense emotes, many
of which describe either ongoing or concluded events, as
well as imaginary worlds' events, it seems that a simple
performative analysis of emoted actions is not going to
hold up.  Whatever "semi-performatives" amount to, these
emotes may indeed be of that type, however.

             6. The Frame Contexts for Events

     The first two categories of emotes discussed
(Conventional Action, Back Channels) simply appear to
be describing punctual events: the action is evoked
co-temporaneous with the utterance.  However, the emotes in
other categories accomplish more sophisticated event
description.  Emoted byplay sometimes contains
meta-discourse description, describing the action of a
conversation with respect to some imaginary world.
Narration of actions in real life can either refer to
activities ongoing, to events just concluded, or events
about to occur.

     The context for an emote determines with what frame of
reference it is to be interpreted, both with regard to
temporal structure of the event being described, and
whether it is intended to represent real action in the real
world, imaginary action in the real world, action with
objects in the MOO, or imaginary actions in the MOO (cf.
Chayko, 1993; Goffman, 1974).

     The way conversants determine the boundaries of events
is partly pragmatic, based on contextual clues.  We know
that a description of an event that is punctual (like an
accomplishment or achievement) cannot be co-temporaneous
with typing, so we interpret it according to the most
plausible interpretation in context.  If Tom announces "Tom
gets off the phone" we can guess he has just done it in
real life, because the event is punctual, there are no
phones in the MOO, and he was probably idle just before the
utterance.  On the other hand, punctual play utterances are
seen as accomplished with their transmission, even if
events that might take a long time in the real world are
described: "Robin takes off and nukes Davey from orbit."

     The expression "lynn waves" is also interpreted
according to the context it occurs in: if it occurs upon
entry to a MOO room, it must be intended to be interpreted
as a greeting action within the context of the MUD, rather
than a description of an action external to the MUD.  The
utterance "lynn waves at Terry Winograd" however, would be
interpreted as a real life action being reported, since
Terry Winograd is not a character on the MUD.  Similarly,
in line 7 below, Tom is not interpreted as waving to MOO
compatriots on entry to a room, because he has not just
entered the room, and he is not using the formulaic short
form of the utterance.

     1 lynn [to Tom]: we could make sure we are home
          on sunday to watch TV.
     2 lynn says, "B5, specifically."
     3 Tom says, "ok"
     4 Tom says, "are we supposed to be somewhere on
          sunday?"
     5 Tom and lynn make public appearances at
          fund-raisers throughout the state.
     6 lynn says, "no. I need to work, but."
     7 Tom smiles and waves to the crowd.
     8 lynn eyes Tom warily.

     Tom's flight of fancy in line 5 is interpretable in
light of his question about being somewhere on Sunday.  If
he had said "Tom and lynn made public appearances at
fund-raisers throughout the state" we would expect that
this had really occurred in real life, given the past
tense.

     This informal discussion of how emotes are
interpretable in context should not imply that people are
never confused about frame reference; it's not uncommon for
people to be confused about whether events are happening in
real life ("irl") or on-line:

     Kelly talks for awhile with soime guy who talks
          baout sex with no introduction. Or rather,
          safe sex.
     Eva peers at Kelly suspiciously.
     lynn eyes Kelly askance.
     Kelly [to Eva]: At MIT. I don't even know his name.
     lynn [to Kelly]: irl?

     Tom actually should probably just leave her out
          of the "people" section; she doesn't seem to
          mud anymore since the breakup.
     lynn says, "breakup?"
     Tom says, "with Willem"
     [at 2:40 p.m.]
     lynn says, "oh."
     Karen tries to remember waht the relationship was
     Tom [to Karen]: capital r
     Tom eyes himself warily.
     Karen says, "irl?"
     Tom nods.

     Most requests for clarification about whether
something happened in real life or in a MUD ("in vr")
usually happen in context of discussion about conversations
or relationships... either of which could be occurring
on-line or in real life.

     As Goffman (1974) points out (and Danet et al., 1994,
describe for IRC), in fanciful activities like games,
the frame of the play activity contains by necessity some
reference anchoring it to the real world.

          The understanding that players and
     nonplayers have of where the claims of the
     ongoing world leave off and and where the claims
     of play take over is part of what the players
     bring to their playing from the outside world,
     and yet is a necessary constituent of play.  The
     very points at which the internal activity leaves
     off and the external activity takes over---the
     rim of the frame itself---become generalized by
     the individual and taken into his frame of
     interpretation, thus becoming, recursively, an
     additional part of the frame.  In general, then,
     the assumptions that cut an activity off from the
     external surround also mark the ways in which
     this activity is inevitably bound to the
     surrounding world" (Goffman, 1974, p. 249).

Just so are the worlds created by the MOO conversation a
part of the real world, and vice versa, facts which the
members of the community fully appreciate and plan for in
their conversation.

                       6. Conclusion

     The different types of emotes discussed are summarized
again in the table below:


============================================================

Table 6.  Summary of Emote Types

Emote Type          Section Tense    Example
-----------------------------------------------------------
Conventional Action 3.1    present   lynn waves.
Back Channels       3.2    present   lynn nods.
Byplay              3.3    present   Mike pastes Tom's
                                                lips...
Narration           3.4    present   ls packs for his
                                                trip.
Exposition          3.5    any       lynn hated the film.

============================================================


     The use of the emote command creates a structured and
complex communicative environment, in which many different
sorts of actions are possible.  Emotes help define contexts
for conversations, establish responsiveness and
attentiveness, communicate understanding, initiate play,
describe actions in real life that may affect the
involvement in the MUD conversation.

     Page conversations provide an interesting contrast to
room conversations, given that the emote command is
available in page conversations as well.  However, users do
not provide as many back channel responses in page
conversations, and page conversation tends to be shorter
and more focused.  Differences between page and room
conversations suggest that social responses to the
geography and to idling behavior within the geography help
in making room conversations different.  The persistent
room metaphor is meaningful in practice, differentiating
MUDs as communication environments from other chat programs
without persistent geography or emotes.

     Emoted actions appear similar to performative speech
acts, which are executed upon utterance.  However,
existence of complex emote types (meta-conversational
description, imaginary world actions, narration of real
life events) makes the parallel with performatives
difficult to sustain; they are probably semi-performative
at best.  The third person distance from the event provided
by the emote creates interesting options for utterances
that cannot be said in the first person.

     Finally, context is important for determining frame of
interpretation for the myriad simple present tense emotes.
Emotes describing actions in the MOO and in the real world
depend a lot on conversational context for their correct
interpretation.  Frames for communication interpenetrate,
allowing users to refer to real life events in the MUD and
make them a part of their regular discourse.

                           Notes

[1] "ElseMOO" is a locative expression among MOOers meaning
roughly "on another MOO."  In other papers I used various
other names for the MOO, but the site is the same.

[2] Interestingly, the CTS command set was constructed by
two users on EM, but is used by a little more than a third
of the regular users of EM now.  This is a good example of
how MUDs evolve over time.

[3] Sexual interactions, which usually take place between
two characters alone or in pages, do not occur with
frequency on the MOO I observe (as far as I am aware).
Emoted sexual or otherwise physical exchanges may count as
"play" as well (McRae, 1995), but this may be a problematic
categorization.

[4] The giggle here is probably an achievement since she
giggled at something in particular.

[5] Note that the emote command makes the character's name the
subject of every utterance.

[6] As a participant-observer in the MUD community, I have
no access to private page conversations aside from my own,
so my inclusion in the logs was inevitable at this point.

[7] This is difficult to establish statistically, since a
minimal conversation in either case might consist of one
utterance exchanged; and it's unclear how to take pauses or
changes in topic into account.

[8] IRC may be evolving somewhat in parallel with MUDs,
however; Reid (1991) did not mention the emote command at
all, which seems to have been added since.  Reports on how
often it is used in IRC vary a lot, however.

                        References

Austin, J. (1962). How to do things with words. Cowford:
     Cowford University Press.

Bruckman, A. (1992). Identity workshop: Emergent social and
     psychological phenomena in text-based virtual reality.
     Electronic manuscript, M.I.T., Media Lab, Boston.

Chayko, M. (1993). What is real in the age of virtual
     reality? ``Reframing'' frame analysis for a
     technological world.  Symbolic Interaction, 16(2),
     171-181.

Cherny, L. (1994a). Gender differences in text-based virtual
     reality.  To appear in Proceedings of the Berkeley
     Women and Language Conference.

Cherny, L. (1994b). ``Objectifying'' the body in the
     discourse of an object-oriented MUD.  Paper presented
     at Midwest Popular Culture Assocation, Pittsburgh,
     PA.

Cherny, L. (1995).  Conversational modes of action in a
     text-based virtual reality.  In progress doctoral
     dissertation, Stanford University, Palo Alto.

Curtis, P. (1992). Mudding: Social phenomena in text-based
     virtual realities.  Intertek, 3(3), 26-34.

Daft, R.L., & Lengel, R.H. (1984). Information richness: A
     new approach to managerial behavior and organization
     design.  In B.M. Staw & L.L. Cummings (Eds.),
     Research in Organizational Behavior, vol. 6 (pp.
     191-233).  Greenwich, CT: JAI Press.

Daft, R.L., & Lengel, R.H. (1986).  Organizational
     information requirements, media richness, structural
     determinants.  Management Science, 32, 554-571.

Danet, B., Ruedenberg, L., & Rosenbaum-Tamari, Y. (1994).
     `Smoking dope' at a virtual party: writing, play, and
     performance on Internet Relay Chat.  To appear in S.
     Rafaeli, F. Sudweeks, & M. McLaughlin (Eds.), Network
     and netplay: virtual groups on the Internet.
     Cambridge, MA: MIT Press.

Dibbell, J. (1993).  Rape in cyberspace or how an evil
     clown, a haitian trickster spirit, two wizards, and a
     cast of dozens turned a database into a society.
     Village Voice, 38(51).

Dowty, D. (1986).  The effects of aspectual class on the
     temporal structure of discourse: semantics or
     pragmatics? Linguistics and Philosophy, 9, 37-61.

Duncan, S. (1972).  Towards a grammar for dyadic
     conversation.  Semiotica, 9(1), 29-46.

Evard, R. (1993). Collaborative networked communication:
     MUDs as systems tools.  Paper given at LISA, Monterey,
     CA.

Ferguson, C. (1983). Sports announcer talk: Syntactic
     aspects of register variation.  Language in Society,
     12, 153-172.

Germain, E. (1993).  In the jungle of MUD.  Time,
     September 13, 49.

Goffman, E. (1963). Behavior in public places: notes on
     the social organization of gatherings. New York: the
     Free Press.

Goffman, E. (1974). Frame Analysis.  Boston: Northeastern
     University Press.

Grosz, B. and C. Sidner. (1986).  Attentions, intentions,
     and the structure of discourse.  Computational
     Linguistics, 12(3), 175-204.

Gumperz, J. (1982). Discourse strategies. Cambridge:
     Cambridge University Press.

Gumperz, J. (1972). Introduction. In J. Gumperz & D.
     Hymes (Eds.), Directions in sociolinguistics.
     Cowford: Blackwell.

Heritage, J. (1984)  A change-of-state token and aspects of
     its sequential placement.  In J. Heritage & J.M.
     Atkinson (Eds.), Structures of social action (pp.
     299-345).  Cambridge: Cambridge University Press.

Hiltz, S.R., & Turoff, M. (1982). The evolution of
     user-behavior in a computerized conferencing system.
     Communications of the ACM, 24, 739-762.

Holmes, M., & Dishman, J. (1994).  Social action in
     synchronous computer-mediated communication: a
     comparison of two genres.  Paper presented at Western
     States Communication Association, San Jose, CA.

Kendall, L. (1995).  Are you male or female? The
     performance of gender on MUDs.  Paper presented at the
     Pacific Sociological Association, San Francisco, CA.

Kiesler, S., Siegel J., & Mcguire, T. W. (1984).  Social
     psychological aspects of computer-mediated
     communication.  American Psychologist, 39(10),
     1123-1134.

McRae, S. (1995). Coming apart at the seams: the erotics of
     virtual embodiment.  Unpublished draft, University of
     Washington, Seattle.

Murray, D. (1989). When the medium determines turns:
     Turn-taking in computer conversation.  In H. Coleman
     (Ed.), Working with language (pp. 319-338). New York:
     Mouton de Gruyter.

Oviatt, S., & P. Cohen. (1988). Discourse structure and
     performance efficiency in interactive and
     noninteractive spoken modalities.  (Tech Report 454).
     Menlo Park, CA: SRI International.

Reid, E. (1991).  Electropolis: Communication and community
     on internet relay chat.  B.A. honors thesis,
     University of Melbourne, Australia.

Reid, E. (1994). Cultural formations in text-based virtual
     reality.  Unpublished master's thesis, University of
     Melbourne, Australia.

Sadock, J. (1988). Speech act distinctions in grammar. In
     F. Newmeyer (Ed.), Linguistics: the cambridge
     survey, vol. 2, (pp. 183-197).  Cambridge: Cambridge
     University Press.

Schegloff, E. (1968). Sequencing in conversational
     openings. American Anthropologist, 70(6).

Schegloff, E. and Sacks, H. (1974). Opening up closings.
     In R. Turner (Ed.), Ethnomethodology: selected
     readings.  Baltimore: Penguin.

Schiffrin, D. (1981). Tense variation in narrative.
     Language, 57(1), 45-62.

Searle, J. (1969). Speech Acts.  Cambridge: Cambridge
     University Press.

Searle, J. (1989). How performatives work.  Linguistics and
     Philosophy, 12, 535-558.

Short, J., Williams, E., & Christie, B. (1976). The social
     psychology of telecommunications.  London: Wiley.

Tang, J., E. Isaacs, & M. Rua. (1994).  Supporting
     distributed groups with a montage of lightweight
     interactions.  Unpublished manuscript, Sunsoft,
     Mountain View, CA.

Verschueren, J. (1994). The conceptual basis of
     performativity.  Unpublished manuscript.

Walther, J. (1992). Interpersonal effects in
     computer-mediated interaction.  Communication
     Research, 19(1), 52-90.

Whittaker, S., Frohlich, D. & Daly-Jones, O. (1994).
     Informal workplacecommunication: what is it like and
     how might we support it?  In Proceedings of CHI
     1994, 131-137.

Wolfson, N. (1982). CHP: the conversational historical
     present in american english narrative.  Dordrecht:
     Foris.