LINGUISTIC RUIN? LOL!

INSTANT MESSAGING
AND TEEN LANGUAGE

SALI A. TAGLIAMONTE
DEREK DENIS
University of Toronto

abstract: This article presents an analysis of Instant Messaging (IM), a one-to-one

synchronous medium of computer-mediated communication. Innumerable articles
in the popular press suggest that increasing use of IM by teens is leading to a break-
down in the English language. The analyses presented here are based on a unique
corpus involving 72 teenagers and over a million words of natural, unmonitored IM.
In addition, a corpus of speech from the same teenagers is examined for comparison.
Targeting well-known IM features and four areas of grammar, we show that IM is firmly
rooted in the model of the extant language. It reflects the same structured heteroge-
neity (variation) and the same dynamic, ongoing processes of linguistic change that
are currently under way in contemporary varieties of English. At the same time, IM
is a unique new hybrid register, exhibiting a fusion of the full range of variants from
the speech community—formal, informal, and highly vernacular.

Teenagers in the early twenty-first century are using home com-

puters for communication at unprecedented rates in ever-expanding virtual
communities. A particularly favorite medium, at least when we conducted
this research, was Instant Messaging (IM). IM is “a one-to-one synchronous
form of computer-mediated communication” (Baron 2004, 13). It is “direct,
immediate, casual online contact” (Schiano et al. 2002). In essence, IM is
real-time “interactive written discourse” (Ferrara, Brunner, and Whittemore
1991, 8).
Examples of IM in (1) show snippets of conversations in the IM environ-
ment. The numbers in square brackets indicate the individual (i.e., in 1a
[025] converses with [3], and in 1b [2] converses with [006]).1

1. a. [025] lol y dont i believe u

[3] its trueeee :P well fine then
[3] what have YOU been doing
[025] i dont kno nothing really
[TEEN/025, male, 17; TEEN/3, female, 17] 2

American Speech, Vol. 83, No. 1, Spring 2008 doi 10.1215/00031283-2008-001

Copyright 2008 by the American Dialect Society

3
4 american speech 83.1 (2008)

b. [2] when i drink tea

[006] ew u do!?
[006] hahahha
[2] yes b/c it tastes good
[2] with milk tea
[2] so it’s sugary
[2] actually i drink tea with chocolate powder
[2] it’s the best thing ever
[006] i’ll try it next time...
[2] YOU SHOULD TRY IT
[TEEN/006, female, 17; TEEN/2, female, 17]

Innumerable articles in the popular press have targeted IM. They suggest
that it is leading to a “breakdown in the English language,” “the bastardization
of language” (O’Connor 2005), even “the linguistic ruin of [the] genera-
tion” (Axtman 2002). In contrast, linguists argue that it is not the result of
students’ lax attitude toward spelling and grammar, but characteristic of a
general “linguistic whatever-ism” (Baron 2003a, 5). Indeed, some have sug-
gested that discourse on the Internet is a “new species of communication,”
complete with its own lexicon, graphology, grammar, and usage conditions
(Crystal 2001, 48).
In this study, we add to the building information on computer-mediated
communication (CMC) in general and the nature of IM language in par-
ticular by setting out to discover what IM language is like among the sector
of the population that uses it the most—teenagers. We approach this task
from two entirely different perspectives. First, we respond to the prevailing
lay perception that IM language is riddled with abbreviations, short forms,
and symbolic uses. The media presentation of these features suggests that
they infuse IM language, but how prevalent are they? Second, because one
of our goals is to contribute new information to the task of placing IM on the
written-to-spoken linguistic spectrum, we conduct a distributional analysis
of a feature identified as one of the “main differences between speech and
writing” (Yates 1996, 40)—the way individuals reference themselves and
others using personal pronouns (see, e.g., Chafe 1982). Finally, we offer
a new perspective on IM language by conducting an analysis of structured
heterogeneity in the linguistic system (Weinreich, Labov, and Herzog 1968,
99–100). Such systems may or may not operate the same way in a medium
such as IM as they do in spoken language.3 We draw on the findings of ear-
lier studies to inform our selection of variable features, focusing on those
that contrast different types of change in contemporary English because we
hypothesize this will give us unique insight into the nature of IM. Rapidly
innovating features among adolescents, such as intensifier so (Tagliamonte
forthcoming) and quotative be like (Tagliamonte and D’Arcy 2005, 2007a),
 Instant Messaging and Teen Language 5

will enable us to determine if IM language is on the forefront of linguistic

change. Longitudinal changes, such as the rise of the going to future (e.g.,
Poplack and Tagliamonte 1999) and deontic have to (Tagliamonte and
Smith 2006; Tagliamonte and D’Arcy 2007b), will enable us to assess the
extent to which IM language mirrors everyday language. We also contribute
a novel methodological approach to the study of CMC by utilizing compara-
tive sociolinguistic methods. These techniques enable the analyst to model
linguistic variation and change in related corpora (e.g., Tagliamonte 2oo2,
2006). Taken together, our analyses lead to an unexpected and intriguing
finding. IM language is characterized by a robust mix of features from both
informal spoken registers and more formal written registers—in essence it
is a hybrid register.
In subsequent sections, we provide a brief history of CMC and IM and
review the relevant literature, describe our data and methodology, present
the linguistic analyses, and offer our conclusions.

COMPUTER-MEDIATED COMMUNIC ATION:

HISTORY AND RESEARCH

Computer-mediated communication has developed at an astonishing rate

in its short history.4 With the launch of the World Wide Web in 1990 and
the mass popularization of the Internet, many forms of CMC became widely
used. These different types of CMC can be differentiated on two parameters:
(1) the number of recipients of a message and (2) the synchronicity of the
communicative event. Though it is possible to have multiple recipients in
IM, it is primarily used for one-to-one dialogue. IM is also synchronous, since
participants are “electronically present at the same time” (Paolillo 1999, 2).
Instant (or near-instant) reply is the norm, and spontaneous, real-time dia-
logue takes place. IM is unique in that it is the only one-to-one synchronous
type of CMC.
At the turn of the twenty-first century, 80% of Canadian teens and 74%
of American teens had used an IM program (Lenhart, Rainie, and Lewis
2001, 3; Randall 2002, 21). IM’s skyrocketing popularity among young people
has raised concern from a variety of people, including parents, teachers,
psychologists, and grammarians. This has led to an explosion of studies on
the topic, with more arising every day.
Nearly all types of CMC have been subject to academic study, many of
which appear in the online Journal of Computer-Mediated Communication (1995–,
ed. Susan Herring, http://jcmc.indiana.edu). Others appear independently
in journals from many fields, including linguistics, computer science, sociol-
ogy, and psychology. Unfortunately, as Cherny (1999, 2) states, “research
6 american speech 83.1 (2008)

on . . . CMC has occurred on the fringes of many disciplinary traditions, often

making it difficult to find previous work on the topic.” Since few studies have
singled out IM in particular, relevant research on all forms of CMC will be
discussed in order to lay the foundation of our study. The majority of research
focuses on providing a descriptive analysis of CMC’s “funky” (Palfreyman
and al Khalil 2003) features (Werry 1996; Hentschel 1998; Randall 2002),
although some studies conduct discourse and/or linguistic analysis (Ferrara,
Brunner, and Whittemore 1991; Yates 1996; Cherny 1999; Thurlow 2002;
Baron 2003a, 2004). Table 1 summarizes the empirical linguistic research on
CMC. It comprises a considerable range of different research foci, including

table 1
Summary of Linguistic Computer-Mediated Communication Research
Year Researcher(s) Form of CMC Corpus Size Feature
1991 Ferrara et al. e-messages (an early 18,769 words subject drop
form of IM) article deletion
copula deletion
tersed sentences
shortened words
1996 Yates computer conferencing Written: Lancaster- type/token ratio
(early newsgroups) Oslo/Bergen; analysis
Spoken: London pronoun use
Lund; CMC: modal auxiliaries
2,222,049 words
1996 Werry IRC no corpus conversation organi-
zation
addressivity
abbreviation
prosody
1998 Hentschel IRC no corpus prosody
lexicon
orthography
turn taking
1999 Paolillo IRC 6,317 lines use of r, u, z
use of code-switching,
Hindi-English
use of obscenity
2003 Palfreyman IM (MSN) 2,400 Arabic words use of Roman key-
& al Khalil & 2,000 English board set to write
words in Arabic
use of local dialect
2004 Baron IM (AIM) 11,718 words turn taking
conversation length
openings & closings
abbreviations
contractions
emoticons
 Instant Messaging and Teen Language 7

features of prosody, spelling variants, and dialect representations. In the next

section, we summarize the major findings of these studies, focusing on the
main research question in this literature—what type of register is IM?
Ferrara, Brunner, and Whittemore (1991, 22–23) were among the first
to address the question of whether CMC media are more like speech or more
like writing. To collect their data, an experiment was created in which 23
subjects used e-messages (an early form of IM) to exchange messages with
a controlled user who helped organize subjects’ travel plans. Though mes-
sages contained features often associated with speech, rather than writing,
well-known features of written registers occurred as well (24).
Yates’s (1996, 46) groundbreaking study compared computer confer-
encing data from corpora of written and spoken language constructed by
other researchers for other purposes. The results suggest a “neither simply
speech-like nor simply written-like” assessment of CMC. While CMC was found
to be similar to written discourse based on type/token ratio tests and lexical
density, he found that the use of pronouns and modals was similar to speech
(Yates 1996, 37, 39, 41, 45). Under the assumptiong that CMC has similar
effects cross-linguistically, additional research supports a more speech-like
profile. Palfreyman and al Khalil’s (2003, 20) work on Arabic showed that
individuals used local dialect features in IM rather than Modern Standard
Arabic. Because Arabic is typically written only in the standard dialect, the
use of vernacular lexicon and spellings showed that CMC language mirrors
speech rather than writing.
Werry (1996, 61) also argues that CMC “reproduces and simulates the
discursive style of face-to-face spoken language.” CMC displays informal and
speech-like features, including abbreviations, short turn-taking,5 and omission
of auxiliary verbs and pronouns, which he attributes to the “temporal, spatial,
and social strictures imposed by IRC [Internet Relay Chat], specifically that
messages be typed as quickly and efficiently as possible” (55). Werry’s research
highlights the importance of the type of medium studied. His study is based
on chat-room exchanges, a synchronous group medium. This could explain
the difference between his findings and Yates’s research, which was based on
asynchronous Newsgroup style CMC. Again, assuming similar cross-linguistic
effects in CMC, Hentschel’s (1998) research provides corroboration of this.
In his data, which come from a Serbian chat room, not only were the turns
short, but individuals broke sentences into several turns. Hentschel (1998,
10) argues that this strategy kept the attention of the interlocutor while at
the same time signaling that there was more coming.
More recent data from American college students suggests that IM is
not as quick paced as the chat room environment (Baron 2004, 409). While
this slower pace could be an indication of a more formal, written-like regis-
ter, two points argue against this: IM is a one-to-one form of CMC, and, as
8 american speech 83.1 (2008)

many researchers have discovered, teens multitask while using IM (Nardi,

Whittaker, and Bradner 2000; Lenhart, Rainie, and Lewis 2001; Grinter and
Palen 2002; Schiano et al. 2002; Baron 2004, 399). This can involve other
computer tasks, watching television, doing homework, or having a face-to-
face or telephone conversation. This means that the (apparent) slower pace
of IM interchanges is likely due to the fact that interlocutors direct less at-
tention to individual conversations, not that they are occupied with careful
composition and editing of messages.
Perhaps the most relevant study for our research is Baron (2004), which
is one of the few empirical studies of IM at the time of writing. Baron’s
study is based on a corpus collected from American college students. Her
remarkable discovery was that IM was more conservative than is suggested
by the press (419). A mere 0.3% of the words had typical IM abbreviations
(e.g., hrs, cuz), less than 0.8% were initialisms (e.g., lol, brb),6 and 0.4% rep-
resented emoticons (smiley faces and the like)7 (411, 413). Furthermore,
where lexical pairs could be contracted, as in I’m for ‘I am’, Baron (2004,
413) found that only 65% were shortened. The study concludes that IM
conversations in this corpus “represent a blend of both spoken and written
language conventions” (416).
The fact that IM diverges from written language at all raises an alarm
from teachers and parents. Indeed, the common perception is that IM, par-
ticularly among adolescents, is close to laughable, filled with grammatical
errors, incomprehensible words, and secret codes—by all counts, an inferior
mode of communication. Teachers often report that student assignments
are littered with typical IM abbreviations (Lee 2002). Although it is easy for
teachers to pigeonhole CMC generally as the root of this problem, Baron
(2003b, 88) offers an alternative explanation: “technology often enhances
and reflects rather than precipitating linguistic and social change.” Thus, one
possibility is that IM is simply mirroring the emerging tendency for written
genres to be more like speech, a process referred to as “colloquialization”
(Mair 1998; Hundt and Mair 1999; Leech and Smith, 2005). At the same
time, the graphic nature of some IM items (e.g., emoticons, elongated spell-
ings, etc.) makes it entirely unlike speech. Both developments reveal novel
tendencies, suggesting that IM may actually be a bellwether in the evolution
of the English language in general.

DATA AND METHODS

Examination of the literature on CMC reveals that it is still little studied;

there are few corpus based studies and only rare comparisons with other
forms of communication (e.g., speech, writing), so it has yet to be definitely
 Instant Messaging and Teen Language 9

characterized. IM, one of the most popular forms of CMC, is even less
known. Moreover, the question of precisely where to place it on a spectrum
between written and spoken language remains open. This provides us with
an intriguing starting point from which to conduct a study of IM. However,
embarking on such a study presents a fundamental problem: How does
the analyst—typically a middle-aged academic—gain access to the natural
informal discourse of the IM speech community, which essentially comprises
members of the teenage generation?
As it happens, the Faculty of Arts and Science at the University of To-
ronto has a unique program that encourages secondary-school students to
join university professors in their research projects—called the Mentorship
Program.8 Between 2004 and 2006, the first author (Tagliamonte) set up a
research project titled Teen Talk in Toronto that was specifically designed to
tap into IM communication. Over the course of the three-year project, two
teams of teenagers (Mentees) from high schools within the Toronto District
School Board worked on this project. The Mentees became coresearchers
and participant observers who engaged in fieldwork and data analysis of an
immense corpus of IM language. Table 2 shows the corpus composition by
the date of data collection and the birthdate and sex of the participants.
The second author (Denis) aided in the management of these projects
as well as the training and supervision of the teenage Mentees. He also
collected the data from the 17–20-year-old population from his own social
networks in the summer of 2005.
The IM portion of the corpus was constructed from conversational
histories—computerized records of the participants’ IM interactions with
friends. These massive text logs, comprising literally thousands of individual
conversations, were donated to the project. An important facet of this IM data

table 2
Corpus Constitution

2004–2005 Summer 2005 2005–2006 Total

Birthdate Male Female Male Female Male Female Male Female
1990 – – – – – 2 – 2
1989 – – – – 3 6 3 6
1988 4 4 – – 2 3 6 7
1987 9 17 3 2 3 2 15 21
1986 – 1 2 4 – – 2 5
1985 – – 4 – – – 4 –
total 35 (12) 15 (7) 21 (11) 71 (30)
note: Numbers in parentheses indicate how many spoken interviews were con-
ducted.
10 american speech 83.1 (2008)

is that the vast majority of the material came from conversational histories
that had been saved well before the study began,9 so the project largely avoids
the observer’s paradox (Labov 1972, 209)—a notable difference between
this and any previous study of IM (e.g., Baron 2004, 404). We believe this
may make it one of the most authentic pictures of teen language and inter-
active CMC discourse in existence and in particular a unique record of IM
communication among teenagers.
The Mentees also audio recorded themselves chatting with a subset of
the same friends with whom they engaged in IM communication frequently.10
These conversations were informal and unstructured, modeling the same
type of interaction they would have had had the audio-recording device not
been present. The data are typically gossip sessions and discussions of school
activities and projects.
In sum, our data are a singular documentation of teens interacting with
teens. All the conversations are between individuals who were well known to
each other, engaged in vibrant, interactive discourse. Moreover, and crucially,
the corpora comprise data from the same interlocutors recorded in different
media: one written (IM) and one spoken. This means we are able to contrast
an individual’s language use directly between one medium and another.
In total, there are 71 individuals in the corpus, 30 male and 41 female,
all born and raised in Canada, all between the ages of 15 and 20 in the years
2004–2006. We emphasize that these corpora have been made possible only
by the cooperation and generosity of our Mentees and their friends as well
as members of the second author’s social networks, all of whom willingly and
enthusiastically provided a wealth of data for study.
Following standard ethical guidelines, each individual signed standard
consent forms, and all names and screen names were systematically changed
in order to ensure anonymity.11 A formatting protocol was devised in order
to process the IM data using Concorder (Rand and Patera 1991). The speech
data were transcribed, digitized, and subjected to the same cataloging tech-
niques. All the data were then systematically processed by creating individual
concordances and word lists for each individual as well as for the data set as a
whole. All told there are over a million words of Instant Messaging and over a
quarter of a million words of spoken conversation. To our knowledge, these
materials are not directly comparable to any existing body of data.12

ANALY SES

In the sections that follow we systematically search and document a range

of linguistic phenomena found in the corpora.
 Instant Messaging and Teen Language 11

im forms. The media typically report IM language by describing a set of

characteristic forms with which IM is assumed to be riddled, among them
emotional language, such as laughter and other sounds, as in (2), and ab-
breviations, as in (3).13

2. a. [002] hahaah what were you on!

[002] those are great
[999] I was on 1 hours of sleep :P
[999] and Blintz
[DER/002, female, 17; DER/999, male, 20]
b. [006] thanks darling
[999] noooooooo problemo :)
[006] hehe
[DER/006, female, 18; DER/999, male, 20]
c. but now after this horrible editing experience.. i’m back to square one.
argh. [TEEN/008, female, 16]
3. a. [3] hehe anywhoo i want to go watch tv and eat more advent calender lol
ill ttyl
[025] lol ok later
[TEEN/3, female, 17; TEEN/025, male, 17]
b. [024] im doing work i leave on wednesday for kenora so i have to get a
ton of work done
[3] kenora for what? basketball? yeah im being buried in hwk
[TEEN/024, female, 17; TEEN/3, female, 17]
c. do you have susan peterson’s email btw?
nvm, clair is getting it [TEEN/027, male, 17]
d. [999] I’l be down for 7 then and meet you dudes whenever
[012] k wicked.
[DER/999, male, 20; DER/012, female, 18]

The first question we pose is how frequent are these forms in IM? Table
3 presents the 16 most numerous of these and their proportion of total
word count.
Notice that there are thousands of uses of haha and its variants.14 Indeed,
this is the most productive feature by far. The highest frequency form stereo-
typically associated with IM is lol, which, although initially standing for ‘laugh
out loud’, is used by our participants in the flow of conversation as a signal
of interlocutor involvement, just as one might say mm-hm in the course of
a conversation.15 The frequency of lol is well below haha, at just over 4,500
times. Next is hehe, another laughter variant but with connotations of gig-
gling.16 Next comes omg, the IM abbreviation for ‘oh my god’, and then hmm,
a fairly mundane form expressing contemplation. Thereafter the numbers
decrease markedly, and this is where we find many of the famed IM forms,
12 american speech 83.1 (2008)

table 3
Characteristic IM Forms: Frequency and Proportion of Total Word Count

haha (laughing) 16,183 1.47%

lol ‘laugh out loud’ 4,506 0.41%
hehe (laughing) 2,050 0.19%
omg ‘oh my God’ 1,261 0.11%
hmm (thinking) 1,038 0.09%
brb ‘be right back’ 390 0.04%
ttyl ‘talk to you later’ 298 0.03%
btw ‘by the way’ 249 0.02%
wtf ‘what the fuck’ 218 0.02%
arg (frustration) 197 0.02%
hwk ‘homework’ 99 0.01%
nvm ‘nevermind’ 78 0.01%
gtg ‘got to go’ 68 0.01%
np ‘no problem’ 65 0.01%
lmao ‘laugh my ass off’ 63 0.01%
nm ‘not much’ 32 0.00%
total 26,795 2.44%

note: The data contain many more IM forms, but these are the most frequent.

such as brb ‘be right back’, ttyl ‘talk to you later’, and wtf ‘what the fuck’.17
Frequencies of the remaining forms recede into obscurity.
The sheer infrequency of the so-called “characteristic IM forms” is dra-
matic when we look at the results through accountable proportional analysis.
They represent a minuscule proportion of the total number of words in the
discourse. Indeed, all the forms combined represent only 2.4% of the cor-
pus. This is remarkably consistent with Baron’s (2004) findings for young
adults in their early 20s in the early 2000s and a more recent study of college
students’ use of text messaging on cell phones (Baron and Ling 2003; Ling
and Baron 2007).18 These results confirm—and with a substantially larger
and unprompted corpus—that the same is true of the teenage generation
in Canada between 2004 and 2006. The use of abbreviations, short forms,
and symbolic uses in IM is without a doubt a new vogue, but much rarer than
the media have led us to believe.19

lol. Perhaps the most eminent IM abbreviation is lol ‘laugh out loud’. The
frequency of this feature, 4,506 instances in the IM corpus, provides an op-
portunity to examine the social distribution of laughter variants within this
teenage speech community. In particular, since our corpus is comprised of
a range of ages, it is possible to test for change over time. Figure 1 shows
 Instant Messaging and Teen Language 13

figure 1
Distribution of Laughter Variants in IM Data

1.6

1.4 LOL
haha
1.2 hehe
1.0
Percentage

0.8

0.6

0.4

0.2

0
15–16 17–18 19–20
Age

the distribution of the IM laughter variants lol, haha, and hehe according to
the age of the individuals.
The use of lol declines systematically according to age, with the younger
individuals using it the most. In contrast, the more conservative form haha
is the clear preference of the oldest individuals. Since it is also the case that
the older individuals have had the longest exposure to IM, we speculate that
this trajectory is the result of incremental loss of the stylized form, lol, in favor
of haha.20 Perhaps as a result of habituation to the IM environment, it seems
that adolescents quickly outgrow at least some of the IM forms. Indeed, a
quip from the first author’s 16-year-old daughter, “I used to use lol when I
was a kid,” provides anecdotal confirmation of lol’s puerile association.

i and u. Another often cited feature of IM is the use of lowercase letters,

particularly i and u for I and you as in (4).

4. a. [002] i’m a nerd too!

[999] yay!
[002] but whats worse is i’m a hippie nerd
[DER/002, female, 17; DER/999, male, 20]
b. i tried getting a hold of u sooo many times [TEEN/024, female, 17]

This variation presents us with a quick and easy way of examining IM language
since alternate forms can be tabulated straightforwardly from the text files.
Table 4 shows their distribution in the data.
14 american speech 83.1 (2008)

table 4
Distribution of (you) and (I)

you 21,491 91.41%

u 2,020 8.60%
I 10,581 25.93%
i 30,222 74.07%

Orthographic you represents the vast majority of variable (you) forms—

91%. Contrast this with variable (I), where the lowercase variant reigns at
74%. Thus, use of i stands out as a frequent feature of IM overall. Of course
the choice of an abbreviated form (which applies only to u) and use of up-
per- versus lowercase (which is fundamental to the use of lowercase i) are
different processes. This may explain the contrasting findings. However,
the next step is to consider whether or not these features are used similarly
across individuals.
Figures 2 and 3 show the distribution of variable (you) and (I) by indi-
vidual. Figure 2 shows that by far the majority of individuals use the standard
form, orthographic you categorically. There are only a few individuals who use
lowercase u, and of these most use it the majority of the time. Interestingly,
very few individuals are variable. Figure 3 shows that by far the majority of
individuals use the nonstandard form lowercase i. Only some use the stan-
dard form I. The norm for sociolinguistic variables is that individuals will use
competing forms (Sankoff 1980, 55). Here, however, only a few individuals
do. Thus, the use of neither lowercase u nor lowercase i exhibits the profile
one would expect of a typical linguistic variable. Instead, it appears that the
selection of one variant or the other is an individual’s stylistic choice, with

figure 2
Distribution of Second-Person Singular Pronouns you and u

100 you
u
80
Percentage

60

40

20

0 b c d i l mn o p s u v f g h j k l m o p q r @ 1 s i n d k c 3 z 2 h r 4 q y b t f j e w a a e g x
Speaker Code
 Instant Messaging and Teen Language 15

figure 3
Distribution of First-Person Singular Pronouns I and i

100 i
I
80
Percentage

60

40

20

0 m p d s u v w x y b e g j h q i pm3 z h j a c k 1 i g 4 a n t k c b f e 2 o l r q @ n r o s l d
Speaker Code

most individuals opting for one form or another.21 Indeed, previous studies
suggest that these IM forms can have strong in-group affiliations (Cherny
1999, 92). This is another result pointing to the stylized nature of the so-
called IM features; however, we still do not know much about the nature of
the IM linguistic system. We begin this inquiry by targeting a feature thought
to be a good measure of the spoken-to-written continuum.

personal pronouns. According to earlier research, first- and second-person

pronouns are more frequent in speech than in writing. In order to situate
CMC on this spectrum, Yates (1996) examined the distributions of pronouns
of personal reference. He found that the language used in computer con-
ferencing was much more like speech than writing.
Taking Yates’s lead, we extracted all pronouns of personal reference in
both the IM and speech data. Figure 4 presents these results proportionally
and compares them to Yates’s written data. In the speech data, personal refer-
ence is dominated by first-person pronouns at 53%. The IM data exhibits a
very similar pattern—first-person pronouns appear 62% of the time. These
findings are parallel to Yates’s findings. Yet, comparison of the IM data with
Yates’s written data reveals a stark contrast. In writing, third-person pronouns
dominate. This supports the hypothesis that IM is similar to speech and
different from writing. Unlike the choice between i and u, the distribution
of personal pronouns reflects the content of the discourse, such as topic
of discussion, suggesting that IM’s register/genre attributes are speechlike
in nature.

Thus far, the emerging profile of IM language reveals a stylized medium

containing a relatively small set of identifying features but which is more
 16 american speech 83.1 (2008)

figure 4
Distribution of Personal Pronouns by Medium

100
Percentage of Total Personal Pronouns

1st Person
80 75
2nd Person
62 3rd Person
60
53

40
29
21 21
20 18 17

4
0
Spoken Instant Messaging Written (Yates 1996)
Medium

conservative than the popular press has suggested. Yet patterns of linguistic
usage, as demonstrated by pronominal choice, suggest a speechlike regis-
ter. It is at this point that we launch into a series of quantitative variationist
analyses of different areas of English grammar. Use of variable grammatical
features is thought to reveal deeper, systemic aspects of the linguistic system
(Poplack and Tagliamonte 2001, 93). As such, they permit the analyst to tap
into the underlying grammar. We have selected these particular features for
two reasons. First, they are features that are undergoing change in the English
language more generally, thus enabling us to assess where IM is positioned
vis-à-vis ongoing linguistic change. Second, each of the variables has been
studied extensively. In addition to having access to the results of earlier stud-
ies, we also mimic these previously established methods and analysis, making
the results obtained here comparable across studies. As far as we are aware,
such an analysis of IM has not yet been conducted. An added benefit of this
corpus is that it allows for cross-genre comparison of the same teenagers in
near identical contexts, but in different media.22 This provides a window
into the nature of IM. If it is more like spoken language, then we can expect
it to pattern similarly to the speech data.23 If, however, IM is on the cutting
edge of innovation, we can expect it to have greater use of incoming forms
than speech. Furthermore, because the grammatical features we target for
investigation involve a combination of formal, standard, and colloquial vari-
ants, the comparative distribution of variant forms in the IM and speech
corpora will provide a unique opportunity to document the sociolinguistic
nature of IM language.
 Instant Messaging and Teen Language 17

intensifiers. The intensifier system in Toronto English is undergoing rapid

change, particularly among adolescents (Tagliamonte forthcoming). Intensi-
fiers are adverbs that maximize or boost meaning, as in (5).

5. a. my clean room is so weird [TEEN/004, male, 18, IM]

b. haha it was kinda creepy, but very cool. [TEEN/002, female, 17, IM]
c. I don ’t have any stories though my life is really boring. [DER/003, male,
19, speech]

This area of grammar is well known to change continuously (Stoffel 1901,

2), partly because of a “speaker’s desire to be ‘original,’ to demonstrate his
or her verbal skills, and to capture the attention of their audience” (Peters
1994, 271). This presents us with an ideal feature to investigate in the IM
medium and to test the common hypothesis that IM is on the leading edge
of linguistic innovation (Schiano et al. 2002, 1). In earlier research on spo-
ken corpora of British English, Ito and Tagliamonte (2003, 257) discovered
that very, the most frequent intensifier in contemporary English usage (Fries
1940, 201; Bäcklund 1973, 290), is frequent only among older individuals. In
contrast, the newer variant really is increasing among the youngest generation.
In Toronto English, the same variety spoken by the teenagers in this study,
younger individuals are shifting dramatically toward use of really. Moreover,
a new form is on the horizon, intensifier so (Tagliamonte forthcoming).
In order to study intensifiers in the speech and IM corpora, we adapted
the methodology in Ito and Tagliamonte (2003), including in the analysis
the 25 most commonly used adjectives from Tagliamonte’s (2003–2006,
forthcoming) corpus of Toronto English.24 Following standard variationist
practice and the principle of accountability (Labov 1972, 72), we extracted
all of these adjectives in contexts in which they were capable of being in-
tensified, whether they were or not. This amounted to over 16,320 tokens
across the two data sets.25
The questions are: What intensifiers are used in IM? Where will IM be
positioned vis-à-vis speech? If IM is a speechlike medium, then we can ex-
pect parallelism with the speech data. Furthermore, assuming IM is on the
leading edge of change, we might expect newer forms such as really and, in
particular, so to be more frequent as well.
Figure 5 shows the distribution of the main intensifiers represented as
the proportion each intensifier represents of all the adjectives considered.
The most frequent forms in these data, both speech and IM, parallel those
found in spoken data in earlier research—they are really, so, and very.26 Con-
sistent with earlier research, very is rare among adolescents. However, it is the
relative proportion of really and so that are informative. In our speech data,
really and so are nearly equally used for intensifying adverbs. In contrast, in
IM the teenagers are much less likely to use intensifiers at all. Since the use
18 american speech 83.1 (2008)

figure 5
Distribution of Major Intensifiers by Medium

15

Percentage Intensification
12 really
10 very
10
so
7

5 4
2
1
0
Spoken Instant Messaging
Medium
note: Percentages were calculated based on the total number of intensifiable adjec-
tives. As is typical, most were not intensified, which explains the low percentages.

of intensifiers is typically associated with colloquial usage and nonstandard

varieties (e.g., Stoffel 1901; Fries 1940), this suggests that IM is a more for-
mal register than speech. However, when the teenagers do intensify their
adjectives, their preferred choice is so. Thus, while the two media mirror
each other in exhibiting the same intensifying adverbs overall, there are
some notable differences. This is the first indication that IM cannot simply
be labeled a speechlike register. It mixes innovative trends alongside an
overarching conservative nature.

the quotative system. We now turn to a feature of English that is another

excellent test for language change in action—the quotative system. In this
area of grammar, the vernacular variant be like is strongly associated with young
people and has increased exponentially in the last decade (e.g., Tagliamonte
and D’Arcy, 2004a, 2004b, 2007a). Quotative be like, similar to intensifier so,
provides a choice opportunity to find out how a rapidly innovating feature
operates in IM.
The examples in (6) show typical usage of quotative verbs in the cor-
pora.

6. a. when I asked him, “So do you still like Anastasia?”

. . . Ø “Well, of course she ‘s good-looking,”
. . . and I was like, “Woah! Not what I asked!” [TEEN/8, female, 16,
speech]
 Instant Messaging and Teen Language 19

b. He’s like “come on come on”

Ø “pow”
Ø “oh god.”
He’s like– he’s like, “if we fought anyone, it would be The-Antics
[DER/001, male, 19, IM]
c. I e-mailed him, I said, “I despise skirts, can I please wear dress pants?”
[TEEN/039, female, 16, speech]
d. some guy’s running around the street going “World War Three, oh my
god, it ‘s world war three!” Ah, do you remember that? [DER/001,
male, 19, speech]
e. [003] we were taking to PL, and she was like . . . you guys will need to
explain why your error was so huge
[003] and he was like “uh . . . yeah . . . .”
[003] I don’t know. People at uts can be so fucki
[1] yeah
[TEEN/003, female, 16; TEEN/1, female, 16; IM]

In earlier research on spoken corpora, the incoming form be like was

found to be encroaching on all the other forms in the contemporary Eng-
lish quotative system, particularly in the younger generation and especially
among young women (e.g., Tagliamonte and Hudson 1999; Buchstaller
2001a, 2001b, 2004; Singler 2001; Cukor-Avila 2002; Tagliamonte and
D’Arcy 2004a, 2007a). In contrast, the standard variants, say, think, and the
like are the favored forms among older individuals.
For this analysis, we extracted the first 50 quotative verbs from individu-
als for whom we had both IM and speech data. This amounted to nearly 500
tokens. Then we followed the methodology for coding and analysis outlined
in Tagliamonte and D’Arcy (2004a).
The questions are: What quotatives are used in IM? Where will IM be
positioned vis-à-vis speech? If IM is speechlike, then we can expect a greater
use of nonstandard forms, especially be like. Furthermore, assuming IM is
on the leading edge of change, we might expect be like to be used more
frequently than in speech. Figure 6 presents the distributions of the three
most frequent quotatives in the data—be like, said, and zero.
There is a dramatic difference between IM and speech media. The spoken
materials reflect the profile of teen language we would expect. Consistent
with many other studies, be like is by far the dominant form, at 68%. However,
the unexpected result is how conservative IM is in comparison. Although
it is true that there is not nearly the same frequency of quotative verbs in
the IM data as in the speech data,27 the fact that we have used proportional
analysis provides an opportunity to compare the two media accountably. Zero
forms are dramatically more frequent in IM (41%) than in speech (12%).
20 american speech 83.1 (2008)

figure 6
Distribution of Quotative Verbs by Medium

80
68 be like
60 said
Percentage Ø
41
40
27
24
20
11 12

0
Spoken Instant Messaging
Medium

This is undoubtedly due to the fact that alternating dialogue can be marked
clearly by separate transmissions and/or punctuation, as in (6b), suggesting
that the heightened use of zero in IM is a genre-specific effect. However, this
cannot explain the curtailed use of be like in IM as compared to speech. In
the speech data, the use of be like dominates the system, whereas in IM the
formal variant said and be like are robust competitors. This difference is all the
more remarkable since quotative said is rarely used by the teenagers in their
speech. Yet in IM the very same teenagers use it just as much as be like.28
Furthermore, the IM data comprises an entirely different balance in the
inventory of quotatives: be like, said, and zero appear in proportions of 24%,
27%, and 41%, respectively. In contrast, the speech data is characterized by
overwhelming use of a single form—be like, which supports the conclusion
that IM has a wider range of variants and greater use of formal and standard
variants than spoken language.

future temporal reference. What if we go deeper into the grammar and

examine a feature that has been variable in English for a very long time? The
future temporal reference system of English is in the process of longitudinal
change in which the incoming form going to has gradually been increasing
since the 1400s (see, e.g., Poplack and Tagliamonte 1999). In this change,
none of the variants are particularly stigmatized; however, the go future has
connotations of informality generally, and its reduced variants, including
gonna, can be considered vernacular.
The examples in (7) show that the IM data has the gamut of forms at-
tested in contemporary English—going to, gonna, will, ’ll, shall, and the simple
or periphrastic present as in (7d).
 Instant Messaging and Teen Language 21

7. a. star wars episode III is going to suck just as much as the previous two and
everyone is going to be really sad [DER/004, male, 18, IM]
b. by the way there will be a pre drink/jam session at my house new years
eve [DER/010, male, 16, IM]
c. hmm. . . .11:30. . .i think i shall hit the sack tonight . . . [TEEN/050. male,
19, IM]
d. and, as far as i know, i’m going shopping with her *alone* on sunday
[TEEN/034, male, 16, IM]
e. ill get my shift switched! nick too? how’re u gonna fit us all in, lol.
[TEEN/003, female, 16, IM]

We adapted the methodology deployed in earlier research (Poplack and

Tagliamonte 1999) but included only future-in-the-present contexts.29 This
amounted to over 4,500 tokens across the two data sets.
The questions are: What future variants are used in IM? Where will IM
be positioned vis-à-vis speech? If IM is more speechlike, then we can expect
parallelism with the speech data. Furthermore, assuming IM is leading
change, we might expect newer, incoming forms such as going to to be more
frequent. Figure 7 shows the distribution of future variants in the data.
Overall, the two media pattern in tandem. Will is the most frequent form.
The two variants of the go future, going to and gonna, make up the bulk of
the remainder, while the use of the simple and periphrastic present is low
and shall is virtually nonexistent. As with intensifiers and quotative verbs,
these relative frequencies parallel the findings of earlier research. As far as
differences between the two media are concerned, the standard variant will
is used more in IM (56%) than in speech (46%), while going to/gonna is used

figure 7
Distribution of Future Temporal Reference Variants by Medium

80
shall will going to gonna present

60 56
Percentage

46
40
30
25
20
13
10 10 8
0.4 1
0
Spoken Instant Messaging
Medium
22 american speech 83.1 (2008)

less in IM (35%) than in speech (43%).30 Focusing on the go future alone,

this trend is repeated. The more vernacular variant gonna is more frequent
in speech (30%) than in IM (25%). In sum, when these teenagers use IM,
they tend more toward the standard variants than when they are talking to
each other. IM is, once again, more conservative than speech.

modals of necessity. The last feature we will consider is also a feature of

English undergoing change. This area of grammar is the deontic modality
system, which involves the expression of obligation and necessity, as in (8).

8. a. As soon as this is over I gotta go up to Forrest-Hill, grab Audrey, then we

have to go up to Yorkdale. [DER/999, male, 20, speech]
b. [999] I wonder what my maxim magazine says about a girl inviting a guy
to watch a movie. . .hmmm
[999] I must consult my Man-Bible
[002] report back
[DER/999, male, 19; DER/002, female, 17; speech]
d. You’ve gotta send me the pics. [TEEN/8, female, 16, IM]
e. [2] bought the dvd, came out Y ESTERDAY
[015] OH OKAY, I NEED TO GET THAT
[TEEN/2, female, 17; TEEN/015, female, 17; IM]

In earlier research on Toronto English, Tagliamonte and D’Arcy (2007b)

demonstrated that teenagers use have to most of the time. The other variants
of deontic modality appear infrequently and, with the exception of need to,
which is a minority variant across people of all ages, are mostly restricted to
the older individuals. As with the increasing use of the go future, the change
is below the level of awareness and, unlike intensifying adverbs, the incoming
form have to is not overtly associated with teenagers.
Once again our methodology replicates that of earlier research, in this
case Tagliamonte and D’Arcy (2007b), who limit the analysis to tokens of
deontic modality that are variable between must, have to, have got to, got to,
and need to (including all morphological forms, such as have to, has to, and
needs to). This provided over 2,200 contexts.
The questions relevant for this analysis are: How does IM reflect ongoing
change in deontic modality in Toronto English? Where will IM be positioned
vis-à-vis speech? If IM is more speechlike, we can expect parallelism with the
speech data. Furthermore, assuming IM is on the leading edge of change, we
might expect the incoming form have to to be more frequent than in speech.
Figure 8 shows the distribution of variants of deontic modality.
As might be expected, have to dominates. Notice, however, that its fre-
quency is highest in the speech data, at 83%. Furthermore, in the speech data,
 Instant Messaging and Teen Language 23

figure 8
Distribution of Deontic Modals by Medium

100
must
83
80 got to/gotta
have got to/gotta
need to
Percentage

60 have to 54

40

22
20 18
14

3 4 3
0 0
0
Spoken Instant Messaging
Medium

neither must nor have got to/have gotta nor got to/gotta represent more than a
few tokens apiece. Only the variant need to—the feature that is stable across
both time and space in Toronto—is found with any degree of frequency, at
14%. Moreover, this rate approximates the rates reported in Tagliamonte
and D’Arcy (2007a).
This time the comparison between IM and speech uncovers two im-
portant results that build on what we have already found. First, observe the
comparatively infrequent use in IM of the variant that dominates in spoken
language: compare the proportion of have to in speech (83%) and that in IM
(54%). Second, IM shows the full complement of all the variants present in
this system in the language—formal, (i.e., must), standard (i.e., have got to),
and colloquial (i.e., got to). The common theme is the unique combination
of variants in the IM medium.

summary. The overall findings are summarized in table 5, which reveals a

number of striking convergences in the nature of IM language. Contrary to
expectation, speech has a more innovative profile in comparison to IM. The
incoming forms so, be like, and have to are all more frequent in the speech
data. At the same time, speech tends to be more vernacular, containing
higher rates of going to than will. In contrast, IM language is consistently more
varied, exhibiting a wider range of variants than speech and, in particular,
containing a higher proportion of standard forms than speech.
24 american speech 83.1 (2008)

table 5
Summary of Findings

Analysis Results
CMC Features
acronyms, short forms, use of these forms is rarer than common percep-
emotional language tion
lol systematic decline of lol according to age suggests
such forms are outgrown
you/u and I/i use of one form or the other is an individual
stylistic choice
Inherent Variability
intensifiers less overall intensification in IM; in SP really and
so predominate, but in IM innovative so is the
leading form
quotatives innovative form be like dominates in speech; more
robust use of varied forms in IM including a
heightened use of the stardard variant said
future temporal reference vernacular gonna more frequent in speech; a
wider range of variants in IM, more frequent
use of standard will
modals of necessity incoming form have to more frequent and domin-
ant in speech; wider range of variants in IM
than in speech (including formal must and
vernacular gotta)

CONCLUSIONS

Taken together these findings permit us to make a number of observations.

In a million and a half words of IM discourse among 71 teenagers, the use of
short forms, abbreviations, and emotional language is infinitesimally small,
less than 3% of the data. Indeed, the proportion of use by these adolescents
is nearly identical to the rates found for the same features among 20-year-olds
(Baron 2004). These results challenge the adverse perceptions of IM pro-
mulgated in the media and suggest that they have been overblown. Indeed,
consistent with Baron’s (2004) suggestion that the college students in her
study had outgrown the stylized IM forms (quoted in Ghenu 2005, 8) the
teenagers in our study are already conservative. Indeed, if our results are
indicative, adolescents appear to abandon the “funky” IM features at a very
young age. Nevertheless, we have clearly tapped into a vibrant new medium
of communication with its own unique style (see also Ferrara, Brunner, and
Whittemore 1991; Crystal 2001; Herring 2003, 2004). Moreover, our results
corroborate earlier CMC research in demonstrating that language use in IM
 Instant Messaging and Teen Language 25

is part of a much broader contemporary trend toward more informal lan-

guage generally (e.g., Biber 1988; Biber et al. 1999; Hundt and Mair 1999).
Yet, the variety of English used in the IM corpora we have studied here is
neither a caricature of real language nor some kind of basilectal lowlife. But
the question is—what is it?
Our foray into the IM environment through quantitative sociolinguistic
analysis, encompassing four areas of grammar and over 20,000 individual
examples, reveals that IM is firmly rooted in the model of the extant lan-
guage, reflecting the same structured heterogeneity (variation) and the same
dynamic, ongoing processes of linguistic change that are currently under
way in the speech community in which the teenagers live. For each of the
linguistic variables we studied, the IM materials mirror the larger Toronto
community in terms of the major variants in the system. Yet the register we
have tapped into is “a language variety that never existed before” (Ferrara,
Brunner, and Whittemore 1991, 26). The consensus in the literature is that
IM is a hybrid. The findings we have presented here permit us to document
just what kind of species it is. For every linguistic variable, IM demonstrates
a unique fusion of variants. Simultaneously, it makes use of formal variants
such as shall and must; informal variants such as will and have to; and highly
colloquial variants like gonna and gotta. It also contains heightened use of
new innovations such as intensifier so, providing evidence to confirm that
IM is a medium on the forefront of change.
This blend of features is easily visible in many of the examples throughout
the article, but the following excerpt from a single conversation illustrates it
particularly well. In this interchange, the two male interlocutors are engaged
in a discussion about their new band. They are both excited about the pos-
sibility of playing together and writing new songs. The conversation revolves
around arrangements having to do with renting recording equipment and
getting together.31

9. [001] ye guy were gonna do lotta different shit for sure

[001] thesongs are gonna totally range
[999] for sure
[001] so you come out tonight
[999] i have to see, but i will try, i wanna
[001] ok just us 3
[001] were gonna party and write shit
[999] nice
[001] ye guy we went to long and mcquade the all in one
[001] is soooooooo cheap
[001] we can rent it for 2 monthes for 240 bux
[001] all in one mixer
[999] really? that’s wicked
26 american speech 83.1 (2008)

[001] yep so like rest assured all our band money will be able to cover record-
ings and me and dick gotta talk to you about recordings, as you already
know but were gonna see how it goes at anti warped and see if it should
just be us 3 doing the recordings and what not, i dunno its something
dick was goin on about but yes meet tonight i must run and shower be
back later
[DER/001, male, 19; DER/999, male, 19; IM]

There are numerous typos, misspellings, swear words, and colloquialisms

(i.e., slang), along with numeric forms and a propensity toward lowercase;
however, notice that there are no IM forms at all. Moreover, a closer exami-
nation of the data reveals the type of linguistic mixing that we have seen re-
peated feature after feature in the quantitative analyses. We have highlighted
the forms that have been investigated here, although there are many others
that are clearly worthy of study. It is important to point out that the type of
combination of features that the quantitative analyses have revealed is not
due to amalgamating different conversations from different registers. Even a
single turn may contain variants of contrastive formality, as with have to, will,
and wanna in line five. Note especially in the lengthy last turn by individual
[001] that the standard variants will and formal must appear alongside gotta
and gonna. Moreover, among innumerable other nonstandard and colloquial
features, the phrase structure that underlies the discourse is actually quite
complex, including subordinators typical of written registers, such as but and
if and even a highly formal use of as.
We have selected the following vibrant examples to illustrate contras-
tive formality in the same turn. The use of shall occurs alongside serious as
an intensifier and the slang term jam in (10a). The use of quotative says in
(10b) occurs where we would certainly expect be like in speech. The use of
deontic must in (10c) appears with the formal variant very in (10d). In addi-
tion, these particular examples also include three IM abbreviations, omgod,
lol, and ttyl.

10. a. aaaaaaaaagh the show tonight shall rock some serious jam [DER/001,
male, 19, IM]
b. Jeff says “lyk omgod omgod omgodzzzzzZZZzzzzz!!!11one” [TEEN/8,
female, 15, IM]
c. heheh okieee! must finish it now ill ttyl [TEEN/003, female, 17, IM]
d. lol. . as u can tell im very bitter right now. [TEEN/008, female, 18,
IM]

This consistent juxtaposition of “forms of a different feather” is the quintes-

sential characteristic of IM discourse.
 Instant Messaging and Teen Language 27

What would lead to such a development? In formal written language,

colloquial variants are illegitimate, prohibited by “language police” of all
persuasions (teachers, editors, etc.). Indeed, anecdotal reports suggest that
teachers are increasingly penalizing students for the use of abbreviations in
written assignments. In teenage conversations, however, formal variants are
equally undesirable. IM appears to be a venue in which teenagers are free to
use all these features together. This linguistic fusion is endemic to the register
itself. Individuals pick and choose from all the available variants that their
linguistic system has to offer and draw from the entire stylistic repertoire
of the language that exists at a given point in time. If the teenagers did not
already possess skilled command of their linguistic system, this would be
impossible. The character and nature of IM we have uncovered here reveals
fluid mastery of the sociolinguistic resources in their speech community. We
conclude that IM, and perhaps computer-mediated communication more
generally, is not the ruin of this generation at all, but an expansive new
linguistic renaissance.
Indeed, this study of IM language is likely already behind the times and
taps only a very small part of what is even now developing. The insurgence of
new media into the contemporary world of communication is always expand-
ing. Since we conducted this study, newer and trendier ways to interact online
have developed, including mushrooming social-networking Web sites (e.g.,
MySpace, Facebook) (see Baron et al. 2005), multiplayer online role-playing
games (e.g. World of Warcraft), and undoubtedly untold other newfangled
ways to communicate online. Simultaneously, the use of text messaging on
mobile phones has gained in popularity in North America, providing yet
another medium that will shape and reshape the nature of communication
(see Baron and Ling 2003; Ling and Baron 2007). All these provide yet
to be discovered venues in which the foremost commodity is language. To
seek out and study the “intriguingly new and still evolving linguistic variet-
ies” (Crystal 2006, 271) that will emerge from these developments will be a
gateway to understanding the future of human communication and perhaps
even greater insights into the language faculty itself.

NOTES

Sali Tagliamonte gratefully acknowledges the support of the Social Sciences and Hu-
manities Research Council of Canada for research grant #410-2003-0005, Linguistic
Changes in Canada Entering the 21st Century; Derek Denis wishes to acknowledge
the support of the University of Toronto Excellence Award in the Social Sciences and
Humanities, Summer 2006. We would like to express our gratitude to our remarkable
28 american speech 83.1 (2008)

Mentees: Milen Foto, Catherine Kierans, Vivian Li, Macy Siu, Helen Tsang, Anita Li,
and Tamar Friedman. Special thanks are due to Jenny Seppänen, Sonja Molfenter,
and Cori Hanson for data transcription, extraction, and helping to manage the cor-
pora and train assistants. We also thank Maria Abdoullaeva, Patrick Dennis, David
Hodgkiss, Chris Latendresse, Amy Levy, Holly Lloyd, Doug Marks, Stuart Marks, Katie
Mayerson, Jesse McLean, Matt Scarlino, and Jess Spindler. We could not have done
this research without you! The initial inspiration for this research was Naomi Baron’s
(2004) article “See You Online.” Subsequent discussions with Naomi have contributed
importantly to this work, although we have only “seen” her online ourselves (albeit
via email not IM). Several anonymous reviewers as well as the editor Michael Adams
offered especially invaluable feedback and suggestions. None of them, of course, is
responsible for any errors that remain.

1. We preserve the IM material precisely as it was typed, complete with typos, spell-
ing errors, spacing anomalies, and the like.
2. The referencing system for our examples records the corpus and individual.
“DER” indicates that the example was collected by Derek Denis, the second
author. “TEEN” indicates that the example was collected by the Mentees. The
numbers following the corpus designation indentify each individual: three digits
indicate the interviewee; single or double digits indicate the interviewer.
3. “Structured heterogeneity” is the idea that variation is inherent to language and
further that it is neither random nor free, but systematic and rule governed.
The question then becomes how these systemic patterns are reflected across
different media.
4. See Baron (2003a) for an in-depth discussion on the history and usage of dif-
ferent types of CMC.
5. “Turn” in synchronous CMC is the term for an individual message of any length
sent by a user.
6. Baron (2004, 411) excluded abbreviations and initialisms that are common in
offline usage (e.g., hrs ‘hours’) or are direct representations of spoken usage
(e.g., cuz ‘because’).
7. Emoticons are among the most novel features of IM. They are “abbreviations of
expressions of mood, tone of voice, or instruction to the reader” (Randall 2002,
27) made using typographical symbols (e.g., :-) ‘smile’, ;-) ‘wink’, :-P ‘stick out
one’s tongue’).
8. For information about the University of Toronto’s Mentorship Program, see
http://www.artsci.utoronto.ca/prospective/highschoolprograms/enrichment#1.
9. Note that the MSN software allows users the option of saving a text record of
their conversations. It was our luck that the Mentees in our study had previously
saved documents of this type on their computers when our project began. Many
of them simply opted to donate these files to the project.
10. Due to the amount of IM data that the Mentees provided, the overall corpus is
more heavily weighted toward IM than spoken communication.
11. Individual conversations were not tabulated separately. All parties involved were
given the option to remove any sensitive interchanges; however, such excisement
 Instant Messaging and Teen Language 29

must have been rare since the data often contain extremely personal interac-
tions.
12. The closest comparison is the corpus collected by Yates (1996), which comprises
written materials from the Oslo-Lancaster Corpus, the London-Lund Corpus
of Spoken British English. The CMC data come from computer conferencing
data at the Open University in the United Kingdom—mostly communications
between students and professors about course work (Yates 1996, 30–33).
13. We did not consider emoticons in this study. The versions of the program at the
time of our study automatically converted the keystrokes that make up emoticons
into a graphical representation. These graphics, however, do not translate across
platforms straightforwardly, making it difficult to study them systematically using
the methods we have employed here.
14. This tabulation comprises all variants of haha, including hahaha, hah, ahahaha,
and the like.
15. The use of lol as a signal for interlocutor involvement is consistent with Baron’s
(2004, 416) comment that lol functions as a “phatic filler comparable to okay,
really, or yeah in spoken discourse.”
16. This tabulation comprised all variants of hehe, including hehehe, heh, eheheh, and
the like.
17. Note that earlier CMC abbreviations reported in Baron (2003a, 21) and Crystal
(2001, 85–86) seemed to our participants to be old-fashioned. This may well be
why they are not used in the current corpus. Moreover, according to anecdotal
reports from our participants, many of the frequently attested IM forms in the
literature were unknown to them at the time of writing. For example, one of
our participants had been using IM for 4–5 years but had never seen lmao. As
one of our participant-observers reported: “it’s not that she’s uncool but rather
lmao is.” Such observations, however anecdotal, point to the transitory nature
of IM abbreviations and acronyms generally.
18. Baron (2004) analyzed nine conversations between females; nine between males;
and five between male and female. The total number of words in her corpus was
11,718.
19. Although people use abbreviations all the time in speech (e.g., TV, DVD, US,
UK ), to our knowledge no one has yet conducted an accountable study of ab-
breviations or short forms in regular use in speech or writing or how such usages
have changed over time.
20. Since lol is among the most well established IM forms (e.g., Cherny 1999, 92),
it seems unlikely that the heightened use of lol among the younger individuals
is the result of increasing frequency due to innovative change.
21. One might think it informative to examine the individuals who are variable to
determine whether they exhibit systematic patterning of these alternate forms
or not. Preliminary analysis revealed no obvious factors that could distinguish
these individuals. They were both male and female, older and younger.
22. This is a major strength of the present study. It studies data from Instant Mes-
saging and spoken conversations from (mostly) the same individuals, whereas
30 american speech 83.1 (2008)

earlier research comparing CMC and speech rely on different data sources and
different individuals.
23. A further test would be to accumulate written materials from the same speakers.
Future research should include such data.
24. These 25 most commonly used adjectives represent over 75% of all adjectives
from the Toronto English Corpus, thus we extrapolated from this that they would
comprise the majority of the adjectives in these IM and speech data as well.
25. Extraction was greatly facilitated by running the huge data files through Con-
corder (Rand and Patera 1991), selecting the relevant forms, and then importing
the text files comprising them directly into Goldvarb X (Sankoff, Tagliamonte,
and Smith 2005).
26. In this analysis, as well as the subsequent ones, we will not go into the details of
variant distributions by age and sex. The few contrasts between male and female
speakers did not detract from the major trends we will report and in all cases
followed expected sociolinguistic trends (e.g., females had higher frequencies
of standard variants). A more in-depth consideration of social factors is left for
future research.
27. There are 6.3 quotatives per thousand words in the speech data and only 1.7
per thousand words in the IM data.
28. While both intensifier so and quotative be like are incoming forms, why is it that
only so is innovative in IM? The answer undoubtedly lies in the contrasting
trajectories of these two changes. While be like has been developing for several
generations and is used by individuals right up to the late 30s (Tagliamonte
and D’Arcy 2007a), the increasing frequency of intensifier so is restricted to
adolescents. Its innovative status in IM is consistent with this profile.
29. This circumvents methodological problems arising from defining the variable
context for future-in-the-past contexts (see Poplack and Tagliamonte 1999).
30. One might alternatively hypothesize that, because will is a shorter form than
going to or gonna, it is predisposed to occur in the IM environment. The same
argument could be made in the case of intensifier so, which is the shortest of
the intensifiers. However, we are reticent to propose this explanation. Having
heard this suggestion numerous times, we analyzed the two corpora for average
word length (not shown here) and found that the IM data actually have a longer
average word length score than the speech data, 3.87 vs. 3.78 letters per word.
31. Several lexical items require explanation: “long and mcquade” is a Canadian
music store; “all in one” refers to a type of recording device; “anti warped” is
the name of a concert where the band was to play; and “bux” is a slang term for
the dollar.

REFERENCES

Axtman, Kris. 2002. “‘r u online?’: The Evolving Lexicon of Wired Teens.” Christian
Science Monitor, Dec. 12, 1. Available at http://www.csmonitor.com/2002/1212/
p01s01-ussc.html.
 Instant Messaging and Teen Language 31

Bäcklund, Ulf. 1973. The Collocation of Adverbs of Degree in English. Studia Anglistica
Upsaliensia 13. Uppsala, Sweden: Acta Universitatis Upsaliensis.
Baron, Naomi S. 2003a. “Language of the Internet.” In The Stanford Handbook for
Language Engineers, ed. Ali Farghali, 1–63. Stanford, Calif.: CSLI Publications,.
———. 2003b. “Why Email Looks Like Speech: Proofreading Pedagogy and Public
Face.” In New Media Language, ed. Jean Aitchison and Diana M. Lewis, 85–94.
London: Routledge.
———. 2004. “See You Online: Gender Issues in College Student Use of Instant
Messaging.” Journal of Language and Social Psychology 23: 397–423.
Baron, Naomi S., and Rich Ling. 2003. “IM and SMS: A Linguistic Comparison.”
Paper presented at the Fourth International Conference of the Association
of Internet Researchers, Toronto, Oct. 16–19. Available at http://jls.sagepub
.com/cgi/reprint/23/4/397.pdf.
———. 2007. “Emerging Patterns of American Mobile Phone Use: Electronically-
Mediated Communication in Transition.” In Mobile Media 2007: Proceedings of an
International Conference on Social and Cultural Aspects of Mobile Phones, Convergent
Media and Wireless Technologies, ed. Gerard Goggin and Larissa Hjorth, 218–30.
Sydney: University of Sydney.
Baron, Naomi S., Lauren Squires, Sara Tench, and Marshall Thompson. 2005.
“Tethered or Mobile? Use of Away Messages in Instant Messaging by American
College Students.” In Mobile Communications: Re-Negotiation of the Social Sphere, ed.
Rich Ling and Per E. Pederson, 293–311. London: Springer.
Biber, Douglas. 1988. Variation across Speech and Writing. Cambridge: Cambridge
Univ. Press.
Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad, and Edward Fin-
egan. 1999. Longman Grammar of Spoken and Written English. Harlow, England:
Longman.
Buchstaller, Isabelle. 2001a. “An Alternative View of like: Its Grammaticalisation in
Conversational American English and Beyond.” Edinburgh Working Papers in Ap-
plied Linguistics 11: 21–41.
———. 2001b. “He Goes and I’m Like: The New Quotatives Re-visited.” Paper pre-
sented at the 30th annual meeting on New Ways of Analyzing Variation (NWAV
30), Raleigh, N.C., Oct. 11–14. Available at http://www.ling.ed.ac.uk/~pgc/
archive/2002/proc02/buchstaller02.pdf.
———. 2004. “The Sociolinguistic Constraints on the Quotative System: British
English and US English Compared.” Ph.D. diss., Univ. of Edinburgh.
Chafe, Wallace. 1982. “Integration and Involvement in Speaking, Writing, and Oral
Literature.” In Spoken and Written Language: Exploring Orality and Literacy, ed.
Deborah Tannen, 35–53. Norwood, N.J.: Ablex.
Cherny, Lynn. 1999. Conversation and Community: Chat in a Virtual World. Stanford,
Calif.: CSLI Publications.
Crystal, David. 2001. Language and the Internet. Cambridge: Cambridge Univ. Press.
———. 2006. Language and the Internet. 2nd ed. Cambridge: Cambridge Univ.
Press.
32 american speech 83.1 (2008)

Cukor-Avila, Patricia. 2002. “She say, She go, She be like: Verbs of Quotation over Time
in African American Vernacular English.” American Speech 77: 3–31.
Ferrara, Kathleen, Hans Brunner, and Greg Whittemore. 1991. “Interactive Written
Discourse as an Emergent Register.” Written Communication 8: 8–34.
Fries, Charles Carpenter. 1940. American English Grammar: The Grammatical Structure
of Present-Day American English with Especial Reference to Social Differences or Class
Dialects. New York: Appleton-Century-Crofts.
Ghenu, Mmike. 2005. “Ill Comunication: How Email, Text and Instant Messaging
Affect Language.” Varsity (Univ. of Toronto), Feb. 28, 8.
Grinter, Rebecca E., and Leysia Palen. 2002. “Instant Messaging in Teen Life.” In
CSCW 2002: Proceedings of the ACM 2002 Conference on Computer Supported Coop-
erative Work, 21–30. New York: Assoc. for Computing Machinery. Available at
http://www.cs.colorado.edu/~palen/Papers/grinter-palen-IM.pdf.
Hentschel, Elke. 1998. “Communication on IRC.” Linguistik Online 1. http://www
.linguistik-online.de/irc.htm.
Herring, Susan C. 2003. “Gender and Power in On-line Communication.” In The
Handbook of Language and Gender, ed. Janet Holmes and Miriam Meyerhoff,
202–28. Malden, Mass.: Blackwell.
———. 2004. “Slouching toward the Ordinary: Current Trends in Computer-Medi-
ated Communication.” New Media and Society 6: 26–36.
Hundt, Marianne, and Christian Mair. 1999. “‘Agile’ and ‘Uptight’ Genres: The
Corpus-based Approach to Language Change in Progress.” International Journal
of Corpus Linguistics 4: 221–42.
Ito, Rika, and Sali Tagliamonte. 2003. “Well Weird, Right Dodgy, Very Strange, Really
Cool: Layering and Recycling in English Intensifiers.” Language in Society 32:
257–79.
Labov, William. 1972. Sociolinguistic Patterns. Philadelphia: Univ. of Pennsylvania
Press.
Lee, Jennifer 8. 2002. “I Think, Therefore IM.” New York Times, Sept. 19, G1. Available
at http://www.nytimes.com/2002/09/19/technology/circuits/19MESS.html.
Alternatively titled “Nu Shortcuts in School R 2 Much 4 Teachers.”
Leech, Geoffrey, and Nicholas Smith. 2005. “Extending the Possibilities of Corpus-
based Research on English in the Twentient Century: A Prequel to LOB and
FLOB.” ICAME Journal 29: 83–98.
Lenhart, Amanda, Lee Rainie, and Oliver Lewis. 2001. “Teenage Life Online: The
Rise of the Instant-Message Generation and the Internet’s Impact on Friendships
and Family Relationships.” Pew Internet and American Life Project. http://www
.pewinternet.org/pdfs/PIP_Teens_Report.pdf (accessed Nov. 25, 2004).
Ling, Rich, and Naomi S. Baron. 2007. “Text Messaging and IM: Linguistic Com-
parison of American College Data.” Journal of Language and Social Psychology 26:
291–98.
Mair, Christian. 1998. “Corpora and the Study of the Major Varieties of English:
Issues and Results.” In The Major Varieties of English: Papers from MAVEN 97, ed.
Hans Lindquist, Staffan Klintborg, Magnus Levin, and Maria Estling, 139–57.
Växjö, Sweden: Växjö Acta Wexionensia.
 Instant Messaging and Teen Language 33

Nardi, Bonnie A., Steve Whittaker, and Erin Bradner. 2000. “Interaction and Outerac-
tion: Instant Messaging in Action.” In Proceedings of the 2000 ACM Conference on
Computer Supported Cooperative Work, 79–88. New York: Association for Computing
Machinery. http://www.acm.org/pubs/contents/proceedings/cscw/358916/.
O’Connor, Amanda. 2005. “Instant Messaging: Friend or Foe of Student Writing.”
New Horizons for Learning Online Journal 11.2. http://www.newhorizons.org/
strategies/literacy/oconnor.htm.
Palfreyman, David, and Muhamed al Khalil. 2003. “‘A Funky Language for Teenzz to
Use’: Representing Gulf Arabic in Instant Messaging.” Journal of Computer-Mediated
Communication 9.1. http://jcmc.indiana.edu/vol9/issue1/palfreyman.html.
Paolillo, John. 1999. “The Virtual Speech Community: Social Networks and Language
Variation on IRC.” Journal of Computer-Mediated Communication 4.4. http://jcmc
.indiana.edu/vol4/issue4/paolillo.html.
Peters, Hans. 1994. “Degree Adverbs in Early Modern English.” In Studies in Early
Modern English, ed. Dieter Kastovsky, 269–88. Berlin: Mouton de Gruyter.
Poplack, Shana, and Sali Tagliamonte. 1999. “The Grammaticalization of going to in
(African American) English.” Language Variation and Change 11: 315–42.
———. 2001. African American English in the Diaspora. Malden, Mass.: Blackwell.
Rand, David W., and Tatiana Patera. 1991. Concorder: Concordance Software for the
Macintosh. Montreal: Les Publications CRM.
Randall, Neil. 2002. “Lingo Online: A Report of the Language of the Keyboard
Generation.” Unpublished MS. Available at http://arts.uwaterloo.ca/~nrandall/
LingoOnline-finalreport.pdf.
Sankoff, D., S. A. Tagliamonte, and E. Smith. 2005. “Goldvarb X: A Multivariate
Analysis Application.” Dept. of Linguistics, Univ. of Toronto. http://individual
.utoronto.ca/tagliamonte/Goldvarb/GV_index.htm.
Sankoff, Gillian. 1980. The Social Life of Language. Philadelphia: Univ. of Pennsylvania
Press.
Schiano, Diane J., Coreena P. Chen, Jeremy Ginsberg, Unnur Gretarsdottir, Megan
Huddleston, and Ellen Isaacs. 2002. “Teen Use of Messaging Media.” CHI ’02:
Proceedings of ACM Conference on Human Factors in Computing Systems, 594–95.
New York: Assoc. for Computing Machinery. Available at http://portal.acm
.org/ft_gateway.cfm?id=506500&type=pdf&coll=GUIDE&dl=GUIDE&CFID=4
9265163&CFTOKEN=26387788.
Singler, John Victor. 2001. “Why You Can’t Do a VARBRUL Study of Quotatives and
What Such a Study Can Show Us.” University of Pennsylvania Working Papers in
Linguistics 7.3: 257–78.
Stoffel, Cornelis. 1901. Intensives and Down-Toners: A Study in English Adverbs. Heidel-
berg: Carl Winter’s Universitätsbuchhandlung.
Tagliamonte, Sali. 2002. “Comparative Sociolinguistics.” In Handbook of Language
Variation and Change, ed. J. K. Chambers, Peter Trudgill and Natalie Schilling-
Estes, 729–63. Malden, Mass.: Blackwell.
———. 2003–2006. “Linguistic Changes in Canada Entering the 21st Century.”
Research Grant. Social Sciences and Humanities Research Council of Canada
(SSHRC). #410-2003-0005. http://individual.utoronto.ca/tagliamonte/.
34 american speech 83.1 (2008)

———. 2006. Analysing Sociolinguistic Variation. Cambridge: Cambridge Univ. Press.

———. Forthcoming. “So Different and Pretty Cool: Recycling Intensifiers in Toronto,
Canada.” In Intensifiers, ed. Belén Mendez-Naya. Special issue of English Language
and Linguistics.
Tagliamonte, Sali, and Alexandra D’Arcy. 2004a. “He’s like, She’s like: The Quotative
System in Canadian Youth.” Journal of Sociolinguistics 8: 493–514.
———. 2004b. “Mom Said, and My Daughter’s Like: Tracking the Quotative System
through the Generations.” Paper presented at the 33rd annual conference on
New Ways of Analyzing Variation (NWAV 33), Ann Arbor, Mich., Sept. 30–Oct.
3.
———. 2005. “When People Say, ‘I Was Like . . .’: The Quotative System in Canadian
Youth.” In Pennsylvania Working Papers in Linguistics 10.2: 257–72.
———. 2007a. “Frequency and Variation in the Community Grammar: Tracking
a New Change through the Generations.” Language Variation and Change 19:
199–217.
———. 2007b. “The Modals of Obligation/Necessity in Canadian Perspective.”
English World-Wide 28: 47–87.
Tagliamonte, Sali, and Rachel Hudson. 1999. “Be like et al. beyond America: The
Quotative System in British and Canadian Youth.” Journal of Sociolinguistics 3:
147–72.
Tagliamonte, Sali, and Jennifer Smith. 2006. “Layering, Change and a Twist of Fate:
Deontic Modality in Dialects of English.” Diachronica 23: 341–80.
Thurlow, Crispin. 2002. “Generation Txt? The Sociolinguistics of Young People’s
Text-Messaging.” Discourse Analysis Online 1.1. http://extra.shu.ac.uk/daol/ar-
ticles/v1/n1/a3/thurlow2002003.html.
Weinreich, Uriel, William Labov, and Marvin Herzog. 1968. “Empirical Founda-
tions for a Theory of Language Change.” In Directions for Historical Linguistics:
A Symposium, ed. W. P. Lehmann and Yakov Malkiel, 95–188. Austin: Univ. of
Texas Press.
Werry, Christopher C. 1996. “Linguistic and Interactional Features of Internet Relay
Chat.” In Computer-Mediated Communication: Linguistic, Social and Cross-Cultural
Perspectives, ed. Susan C. Herring, 47–63. Amsterdam: Benjamins.
Yates, Simeon J. 1996. “Oral and Written Linguistic Aspects of Computer Confer-
encing: A Corpus Based Study.” In Computer-Mediated Communication: Linguistic,
Social and Cross-Cultural Perspectives, ed. Susan C. Herring, 29–46. Amsterdam:
Benjamins.