A Large Scale Study of Text Messaging Use
Agathe Battestini, Vidya Setlur, Timothy Sohn
Nokia Research Center
955 Page Mill Road
Palo Alto, CA 94304 USA
Text messaging has become a popular form of communication with
mobile phones worldwide. We present ﬁndings from a large scale
text messaging study of 70 university students in the United States.
We collected almost 60,000 text messages over a period of 4months
using a custom logging tool on our participants’ phones. Our re-
sults suggest that students communicate with a large number of
contacts for extended periods of time, engage in simultaneous con-
versations with as many as 9 contacts, and often use text messaging
as a method to switch between a variety of communication medi-
ums. We also explore the content of text messages, and ways text
message habits have changed over the last decade as it has become
more popular. Finally, we offer design suggestions for future mo-
bile communication tools.
Categories and Subject Descriptors
H.5.m [Information Interfaces and Presentation (e.g., HCI)]:
Short Message Service (SMS), text messaging, mobile device, tex-
ting, large-scale study
Text messaging is a method for sending short 160 character mes-
sages between mobile phones that has become a popular, global
method of communication. In the past decade, text messaging
use has grown from 12 million to 135 billion messages sent ev-
ery month . The popularity of text messages can be attributed
to a variety of factors, such as cost per message compared to voice
minutes, social appropriateness, ease of use, and the lightweight
nature of sending messages [8, 4, 17]. Text messaging use has also
increased as a result of interaction with online services such as bank
statements, social networks, and chat clients.
There are differences in text messaging use around the world,
and numerous studies have been conducted in countries such as
Germany , Finland , Norway , United Kingdom and
Japan . The earliest text messaging studies were conducted in
Norway, examining the shift from voice to text in teenagers’ mobile
Copyright is held by the author/owner(s).
MobileHCI’10, September 7–10, 2010, Lisbon, Portugal.
phone usage . Since then, text messaging has been explored in
many countries, but few studies have been conducted in the United
States . These studies explored the motivation behind adopt-
ing text messaging as a dominant communication medium and pro-
vided details about how text messaging was being used.
There have been several notable pieces of work in the research
literature. Grinter and Eldridge offered some insights into the con-
tent and character of texting among a small group of British teena-
gers . They reported ﬁndings that teenagers tend to text for three
primary activities (chatting, coordinating and planning). Given the
immediacy and mobility of text messages, Kasesniemi and Rauti-
atainen found that text messages started to resemble online chat in
turn taking and discourse structure . In a questionnaire study of
British and American texters, Reid and Reid found a difference be-
tween ’texters’ and ’talkers’, with ’talkers’ being much more active
in conversations and frequently participating in simultaneous text
conversations . Our work explores these issues further through
an analysis of a large-scale text messaging data set.
In this paper, we present a comprehensive analysis of a 4-month
text messaging study conducted with 70 students at an American
university. Our work complements the aforementioned studies and
expands on the broad range of text messaging use. Unlike previous
studies that rely on self-reporting through diary studies and ques-
tionnaires, we instrumented each participant’s phone with a logging
client that captured all incoming/outgoing text messages along with
location information of where the texts were sent or received. We
collected a total of 58,203 text messages logged from participant’s
phones between December 2008-April 2009. The dataset we col-
lected gives us a unique opportunity to contribute additional anal-
ysis to previous text messaging research. Our ﬁndings reveal that
participants text with a large number of contacts, often have several
simultaneous conversations through texting, and use text messag-
ing to switch between multiple communication services. Although
previous research has identiﬁed that texters engage in some of these
behaviors, we present further analysis through a large dataset to un-
derstand the topics of text message conversations and how simulta-
neous conversations occur. Based on our ﬁndings, we make several
suggestions for designing future mobile communication technol-
There are many methods to capture data from participants in situ.
Previous SMS studies have used a variety of diary study methods
that involve the participants keeping a journal of their SMS activ-
ities . Diary studies help capture real data in the moment, but
suffer drawbacks of potentially missing data because participants
forget to record an entry. Another possible method is direct obser-
vation, but this can be quite time consuming and would not capture
data in certain parts of the day when the observer is not present.
We chose to collect data through a custom logging tool installed
on the participants’ phones. In contrast to a diary study where some
messages may not be recorded, we were able to capture all text
messages on the device. The logger recorded each incoming and
outgoing SMS message along with the location (via GSM cell posi-
tioning ) where the message was sent/received. Participants in-
teracted with the native text messaging application, but were aware
that a logger was running on the phone. We enabled several pri-
vacy features on the phone so that participants could control the
text message logger; turn off location monitoring or message gath-
ering altogether. Text messages were uploaded to our central server
each day and posted on a website for the participants to see their
messages. If the participants did not want certain messages to be
included in our analysis, they could delete the messages from our
system through the web interface. Our analysis and results do not
include any deleted messages.
Our study took place at a local university in the United States
from December 2008 until April 2009. We recruited 70 partici-
pants (48 male, 22 female) through ﬂyers and online mailing lists.
Participants were all undergraduate or graduate students at the lo-
cal university studying a variety of majors such as economics, arts,
communications, computer/political/environmental science, biol-
ogy, linguistics, ﬁlm, international relations, product design, elec-
trical/mechanical/bio engineering, architecture, classics, and edu-
cation. We recruited over a wide range of college students to gain a
broader understanding of text messaging use in the university com-
munity. Although our participants come from a select sample of
university students, they help provide insights into novel uses of
text messaging. As text messaging continues to grow, we believe
many of these behavioral patterns will carry on with these students
from the university campus to the workplace.
All the participants owned a mobile phone and were experienced
with text messaging. Ages ranged between 17 to 26 years (mean =
20.36 years, σ= 1.92 years). Participants had an average of 6.26
years (σ= 2.51 years) of experience with a mobile phone. They
were also proﬁcient in text messaging, with an average of 4.54
years (σ= 2.3years) of experience.
We collected data from our participants through an initial online
survey and text message logging tool on the phone. The survey
asked each participant about whom they text with, why they text,
and general attitudes towards text messaging. After the survey, we
gave each participant a Nokia N95 to use with our pre-installed
logging software. The participants were all required to have SIM
cards that they could easily transfer into the N95 phone. As phone-
books can be saved to the SIM cards, the participants all had access
to their phonebook data from their previous phones. We asked the
participants to use the study phone as their primary phone for the
entire 4 months of the study. Most participants already had tex-
ting and data plans, but 10% of them did not have the unlimited
plan needed to cover the sending and uploading of SMS data to our
servers. At the time of this study unlimited text plans averaged $15
USD per month and unlimited data plans averaged $20 USD per
month. Instead of reimbursing the cost for an unlimited text and
data plan, we compensated the participants by letting them keep
the phone (≈$400 USD) at the conclusion of the study. The cost
of the phone covered the charges incurred for the text/data plans
and included additional compensation for their participation in the
All participants were asked to sign a consent form acknowledg-
ing that their text messages would be logged and later used for re-
search analysis. Consent was mandatory before participating in any
part of the study. We strongly encouraged the participants to also
notify any third parties that they communicate with about their par-
ticipation in our study. Third party consent is often a difﬁcult issue,
thus we created mechanisms for our study participants to remove
messages that any third party did not consent to. Further, the names
of the participants mentioned in our paper are pseudonyms.
3. SURVEY RESULTS
Figure 1: Percentile breakup of people with whom the partici-
pants typically text message, based on an online survey. Texting
is used for a wide variety of recipients. Friends have the highest
occurrence at 100%, classmates/projectmates are ranked sec-
ond at 74%, and siblings ranked third at 60%.
We conducted an initial online survey with the participants as a
means of self-reporting, asking questions about their normal tex-
ting usage and attitudes. Figure 1 shows the percentage of partici-
pants of whom participants reported that they send text messages.
All participants responded that they send text messages to their
friends, which is consistent with previous studies [11, 4]. Class-
mates and projectmates were ranked the second highest at 74%.
Siblings ranked third at 60%. A number of participants indicated
that they exchange text messages with their parents (37%). This
number is higher than previous reported numbers and suggests that
text messaging is becoming a popular method for interfamily com-
munication. Professors and teachers were also recipients of text
messages, using the method as a way of communication for class.
We asked participants to respond to 14 questions about their gen-
eral attitudes towards text messaging using standard communica-
tion research scales for interpersonal communication motives 
and afﬁnity . Figure 2 shows the results from these questions.
One of the classic mobile phone problems is text input. Participants
responded that they generally use abbreviations and predictive text
while text messaging. Although abbreviations can help lower the
Figure 2: Plotting general attitudes of text messaging usage.
character count, more than 40% of the participants indicated that
they are not careful about the length of their messages.
Participants indicated that they often send group messages to
multiple recipients. Although this is not naturally supported in
the SMS protocol, we found both in the survey and logging study
that group messages are an important part of daily communication.
Group messages were also some of the initial messages that started
simultaneous conversations with friends. We describe some exam-
ples of this in our logging study results.
Many of the respondents indicated that they would like to write
down or save text messages that are special. Taylor and Harper
have described similar sentiments through the embodied meaning
behind text messages . Over 60 % of the participants responded
that they would miss text messaging if it were not working. How-
ever, few participants considered text as one of the more important
things they do each day. Very few participants text in different lan-
guages or sent text messages from desktop applications. Finally,
most respondents seem to use more basic text messaging features
and less of multimedia messaging.
In the last part of our survey, we explored some of the moti-
vations behind texting (Figure 3). We asked the participants why
they send text messages, and provided twenty possible responses
for them to rate. Most respondents send text messages more for its
utilitarian beneﬁts (e.g. to reply to a question, to announce some-
thing, to request something from someone, to ask where somebody
is, to plan a meeting), and less for its emotional beneﬁts (e.g. be-
cause it’s exciting, because it makes me feel less lonely, to gos-
sip, let others know I care about their feelings). Using Principle
Components Analysis, we found that all but one of the motivations
items (escape, pleasure, inclusion, relaxation) and afﬁnity measures
form a single and highly reliable factor (Cronbach’s α= .96). By
averaging the scores on those items, we created a single positive
orientation toward text messaging factor. The affection and con-
trol motivations did not ﬁt into this factor. By regressing number
of years of texting experience upon this positive orientation toward
text messaging factor, we found that numbers of years of texting
experience predicted a negative orientation toward texting among
respondents to our survey (model R2=.64, b=-.80, p<.05). The
number of years of experience of texting did not signiﬁcantly pre-
dict an orientation toward controlling others or seeking afﬁnity with
others. This suggests that the people who had more experience with
text messaging tend to use it more often, but they did not not seek
escape, pleasure, or relaxation from it.
The survey results provide some initial insights into the text mes-
saging usage patterns of the study participants and their perceived
opinions about text messaging. This information along with the
quantitative analysis described further in the paper, can help moti-
vate informed design choices for text messaging interfaces regard-
ing what users presume and their actual behavioral patterns.
Figure 3: Plotting responses to why people send text messages. Respondents tend to send text messages more for its utilitarian
beneﬁts and less for its emotional beneﬁts.
4. LOGGING STUDY ANALYSIS
The participants sent a total of 58,203 messages during the study
(24,806 outgoing, 33,397 incoming). There were an average of
848.6messages exchanged during the study by a single participant
(σ= 1000.5,min = 7,max = 5304). The average time to reply
to a text message received or receive a reply to a text message sent
was 6 : 04 minutes.
We did not ﬁnd signiﬁcant gender differences. Female partici-
pants communicated slightly more than male participants with an
average of 862 messages per female user (σ= 974,min = 26,
max = 3522) versus 817 messages per male user (σ= 1000,
min = 7,max = 5304). Text messages sent by female partic-
ipants were also slightly longer with an average of 58 characters
(σ= 48,min = 1,max = 513) versus 47 characters (σ= 44,
min = 1,max = 517) for male participants. Hoﬂich and Geb-
hardt also noticed that women send more and longer messages than
men, but their results showed a more signiﬁcant gap between gen-
4.1 Text Message Conversations
Our logging tool captured both incoming and outgoing messages
sent by each participant. The lightweight nature of text messages
means that all messages are treated as single units, even though they
may form part of a larger conversation between two individuals. To
address this, we applied a conversation grouping algorithm to all of
the messages. Messages that cannot be grouped into a conversation
are treated as single messages (e.g., messages without responses).
We deﬁne a conversation as two or more text messages exchanged
between contacts with at least one incoming and one outgoing mes-
sage. We applied a conservative time metric of a 20 minute re-
sponse time in order to group messages together. If a message be-
tween two individuals was followed with another message within
the 20 minute window, the messages would be grouped together
as a conversation. Our algorithm identiﬁed a total of 8,590 con-
versations with an average of 122.7conversations per user. The
average conversation contained 4.93 text messages exchanged be-
tween contacts. We found that there is a strong correlation (correla-
tion factor = 0.95) between the total number of text messages sent
and the number of conversations a user was engaged with. In to-
tal, 72.8% (42,390 messages) of the total messages were part of a
(a) Messaging network of 1255 users (pilot participants and their
contacts) from the study dataset
(b) Zoomed in view of a portion of the messaging network
Figure 4: Visualization network of the text messaging data. Here, each node denotes a user. An edge between two nodes indicates
that a text message was sent between two users. A single text message is denoted by one line as the edge. A blue edge indicates
an outgoing text message, while an orange edge indicates as incoming message. If two nodes text message to one another, then the
orange color would appear roughly in the middle of the edge. A conversation is denoted by the thickness of the edge. Thicker the
edge, the more back-and-forth text messages there are in a given conversation.
larger conversation and 27.2% (15,813 messages) of the messages
were single messages.
4.2 Categorizing Conversations
A large dataset of text messages provides a unique opportunity to
explore the topics that are often discussed in text message conver-
sations. Three researchers independently categorized 35% (3,000
out of 8,590) of the conversations, tagging them with words that
identify the content and purpose of the message. We chose only
to do a subset of the conversations given the overwhelming number
of conversations. We believe that this subset still maintains insights
into the broader set of data.
After all three researchers categorized the subset of data, we
grouped the tags into higher-order categories. Each conversations
was then given votes based on which category each researcher had
placed them in. If a conversation had two or more votes, then it
was classiﬁed into the appropriate category. Some conversations
spanned multiple topics and could be placed in multiple categories
if there were enough votes to do so. These categories only reﬂect
what we observed through this study and are not meant to be an
exhaustive list. Moreover, our study only captures a population of
college students, but still offers insights into behaviors and patterns
for text messaging use. Table 4.2 shows each category, a represen-
tative example taken from the conversation subset, and the number
of conversations in that category.
Planning (31.7%) accounted for almost a third of the conversa-
tion topics. Many of these conversations were related to planning
future events/get togethers, coordinating around meal times, and or-
ganizing rides. Given the lightweight nature of text messaging, co-
ordination is a popular use to send quick messages back and forth.
Previous research has also observed that planning and coordination
are dominant topics among teenagers . Our results indicate that
this topic is similar among college students.
Relationships (15.3%) were the second most frequent topic of
conversation. These conversations involved gossip about others,
‘thinking of you’ type messages, and discussions about deep rela-
tionship issues. We observed several ﬁghts and breakups between
signiﬁcant others that were communicated over text messaging.
The conversation with the most messages lasted 3hours 39 min-
utes between two signiﬁcant others ﬂirting from the morning until
afternoon. The entire conversation contained 108 text messages
Chatting was a popular use for text messaging. We categorized
messages that did not have a clear topic of conversation into this
category. Examples included conversations asking "what’s up",
"how’s it going", and "good morning". The longest conversation
our algorithm discovered was 7hours 58 minutes and was catego-
rized in this category. The conversation consisted of 38 messages
(every 12 minutes) sent between friends discussing their life and
making weekend plans (also listed in the planning category).
All our participants were part of the university community and
would sometimes discuss school and job related information thro-
ugh text messaging (10.9%). Text messaging was popular for an-
nouncement, such as when practice would be or what time a frater-
nity gathering was happening. Students also used text messaging
to discuss homework answers with each other. A secondary topic
Figure 5: Trend lines of average times to respond to simultaneous and non-simultaneous conversations.
that is popular among university students is food, which accounted
for 9.5% of the conversations.
Places (10.2%) and Information Seeking (10.2%) had an equal
amount of frequency and are both closely related to mobile infor-
mation needs. Conversations about places typically involved letting
others know one’s location or querying others to know their loca-
tion. In a study of mobile information needs, Sohn et. al. observed
that information about friends and places were a frequent type of
information sought after by mobile users. Moreover, they also ob-
served that people would call a friend to address their information
needs, which is represented in our study through text messaging.
Information seeking through text messaging others has a unique
way of being asynchronous and allowing recipients to respond at
With the popularity of social media that encourages broadcasting
status messages to the world, we found text messaging an interest-
ing place to selectively broadcast these messaging. Current status
messages (9.0%) through text messaging naturally afford picking
the recipients that know what you are doing. Other related topics
to current events included Sports/TV/News (6.8%) that were about
scores for ongoing sports games and newsworthy topics.
Communication (5.3%) was an interesting use of text messag-
ing that we believe provides insights into the future design of mo-
bile technology. Text messages conversations in this category in-
volved continuing conversations from other communication medi-
ums, such as responding to emails, Facebook messages, or phone
calls. Some participants would move to text messaging if a conver-
sation needed to switch from online chat to mobile, or they wanted
to share email/web addresses with each other. This category of con-
versations reveals that users communicate with each other across a
variety of services, and text messaging is often used as a media-
tor between them. Creating an easier way to switch between these
communication services could be a great beneﬁt for mobile users.
Illicit Activities (2.9%), Health (1.7%), and Money (1.1%) round
out the list of categories and were discussed with low frequency.
4.3 Simultaneous Conversations
Given the large percentage of messages that were part of a con-
versation, we assumed that there would be a number of simulta-
neous conversations that occurred. Participants engaged in a sig-
niﬁcant number of simultaneous conversations, sometimes with as
many as 9different contacts. Figure 4 shows a network snapshot
of messages sent between people during the study. In the ﬁgure,
each node represents a user. An edge between two nodes indicates
that a text message was sent between two users. A single text mes-
sage is denoted by one line as the edge. A blue edge indicates an
outgoing text message, while an orange edge indicates an incom-
ing message. If two nodes text message to one another, then the
orange color would appear roughly in the middle of the edge. A
conversation is denoted by the thickness of the edge. A thicker
edge represents more back-and-forth text messages there are in a
given conversation. From Figure 4b, even though it represents a
subset of the data, we can see that signiﬁcant number of conversa-
tions occur between users based on the thickness and colorations of
the edges. These conversations also occur with numerous people as
evidenced in the subset of participant 35 (Figure 4b).
Simultaneous conversations were a signiﬁcant part of our par-
ticipants text messaging habits. Despite a text messaging interface
that does not explicitly support multithreading messages, we found
numerous simultaneous conversations in our dataset. We deﬁne a
simultaneous conversation as two or more conversations overlap-
ping in time. Previous studies have rarely observed simultaneous
conversations in text messaging , but have seen this type of be-
havior in instant messaging use [5, 16]. IM studies have reported
that simultaneous conversations only occur with a handful of con-
tacts. Our results suggest that simultaneous conversations are be-
coming more frequent in text message conversations.
Given many of the traditional problems with mobile interfaces
(i.e., screen size, input) we did not expect users to be able to carry
on so many conversations at once. This behavior is interesting be-
cause text messaging clients on mobile phones are not designed
like IM clients. Most desktop messaging clients can take advan-
tage of larger screen space so multiple conversations can be viewed
at once. In contrast, the text messaging interface of most mobile
phones (particularly the phone used in our study) only allow the
user to view one message at a time. Despite these constraints, some
of the conversations in our study were with as many as 9unique
contacts at a time. Using the same metric for detecting conversa-
Category Number of Conversations Percentage Example
out: “What time are you leaving for class”
951 31.7 in:“Im at the dining hall. I’m leaving straight from here at like 1:05?”
out: “Hey. I thought about u today’"
460 15.3 in:“Aww thanks that so sweet , did you really think of me so cute”
out: “Yes i did cuz u werent answering :( awwww i wanna cuddle”
Chatting 412 13.7 out: “How’s LA? did you really wake up at 8?’"
in:“Hehe got to my house at 9:30 so i got up at around 9. LA is good”
School/Jobs 326 10.9 out: “When are our presentations due?’"
in:“Like in 2 weeks? Or maybe next week”’
in: “Hey i’m in the french house where you at?’"
305 10.2 out:“Dance ﬂoor away from the doors”’
out: “We are leaving. You ok?"
in: “Yep i’m with people walking"
Information Seeking 305 10.2 out: “Who was that artist you put on in the cluster last night?’"
Food 285 9.5 out: “Im hungry i want orange chicken lol’"
in:“i’m coming over right now”’
out: “Where are you?"
270 9.0 out:“Denzel washington is in the crowd!!!”’
in:“I am on the tv platform. Hopefully I don’t get kicked out.”’
Sports/TV/News 205 6.8 out: “Whats the score?"
in:“3 and a half minutes left uconn up by 5”’
Communication 159 5.3 out: “Yo can u email me those ids when u can"
in:“Whats your email?”’
Illicit Activities 86 2.9 out: “We headed home now? Got some to smoke"
out: “I have obstructive sleep apnea... :("
52 1.7 in:“?”’
out:“From the sleep clinic...”’
in: “I need to pay my mom back tomorrow :)"
33 1.1 out:“Haha whoops i forgot. How much do i owe?”’
in:“I’m not sure, ask carla (she kept the check)”’
Table 1: Categories of a subset of conversations. Here, incoming text messages are prepended with ‘in:’ and outgoing text messages
are prepended with ‘out:’
tions described above, we found that each participant had an aver-
age of 7.9simultaneous conversations (σ= 12.1, min=0, max=57)
during the study. The top 5users all engaged in 30 or more simul-
There were 554 simultaneous conversations during the study.
These conversations can overlap in a natural way as text messages
are sent and received with different contacts, but they can also be
initiated by a multi-recipient text message. 56% of the simulta-
neous conversations were initiated with a multiple-recipient text
message. The text messaging application on the N95 phone lets
the user compose a message and select multiple recipients. Unlike
email that shows a full recipient list, text message recipients will
not be able to see who the text message has been sent to. The fol-
lowing conversation shows a user trying to ﬁnd out who is coming
to his party by sending out a single message to 8 contacts:
4.3.1 Variety of Simultaneous Conversations
The conversation highlights a message with a quick response that
others can respond to. There were a number of conversations that
began with a more open-ended text message, which led to several
longer conversations. Table 2 shows several topics of simultaneous
conversations that were started with a multi-recipient text message.
Some messages prompted short conversations such as sports sco-
Alice Are yall coming tonite?
Response 1 I just woke up from a nap and I need
to work on my ihum essay. Its a
night-in for me :(
Response 2 Yea. Are you already there?
Response 3 Yup
Response 4 Hells yeah! Is it poppin?
Response 5 Coming now.
Response 6 Isabelle anton and i are coming soon
Response 7 I’m sick have fun tonight!
Response 8 We are heading over as we speak
res. A number of categories spawned several simultaneous conver-
sations because of their more open-ended nature. One of the big
incidents that occurred during the study was a participant’s dorm
room being robbed. She sent text messages to 4contacts that in turn
spawned conversations with each one. The conversations lasted for
the next four hours. Interestingly, the conversations never trans-
ferred over to face-to-face or any other communication medium
during that time frame. We believe that this may suggest text mes-
saging is becoming both a common and preferred method of com-
munication for some individuals.
Figure 6: Time chart of a user across the duration of the study. Notice that the frequency of text messages is higher towards the
evening and night, probably typical of a student.
Future Plans Where are y’all? You are making 012
look bad. C’mon. 680 is where its at!
Sports Scores WE WON 84 66 AWESOME AWESOME
General Greeting Hey u good morning. Whats up?
Thanks Hey girls thanks so so much for coming
tonight! It meant so much to me to have
u both there! :)
Big Incidents My room got robbed
Looking For Items Hey do you know anyone with a truck or
large van that would maybe let me bor-
row it 2m to move into my apartment in
Announcements Practice moved to tomorrow, Meetings
are today at 2:30 still
Future Communication hey call me at thls number my phone is ab
to die and i need to talk to u ab something
Chain Letter If I LeT u GRAB 1 pArT oF My BoDy
WhaT pArT WouLd u GRAB?? Keep this
going to both sexes! And see what u get
Table 2: Categories of simultaneous messages
There were several chain letter messages in our dataset that we
were surprised to ﬁnd (Table 2). Chain letters are typically associ-
ated with email, but it seems they have started to make an appear-
ance in text messaging. The content of these chain letter messages
included topics that were sexual in nature, and about bringing good
luck if passed along to several friends. It is unclear how far the
chain letter was passed on, but it was interesting to see this type of
4.3.2 Simultaneous Conversation Response Times
We were surprised at how users were able to handle these si-
multaneous conversations on their mobile phone. We analyzed the
dataset to assess how well people multitask when they have si-
multaneous conversations with different contacts. 50 users in our
dataset had both multiple simultaneous conversations, and multi-
ple non-simultaneous conversations. Across all users, the average
time to respond during simultaneous conversations was 431.28 sec-
onds ( min = 42 seconds, max = 1217 seconds, σ= 274.453
seconds). The average time to respond for non-simultaneous con-
versations was 391.88 seconds (min = 72 seconds, max = 937
seconds, σ= 156.141 seconds). There are a variety of factors that
affect response time; participants could be driving, in a meeting, or
missed a text message notiﬁcation. In general though, we noticed
that users tend to take longer to respond to an incoming text mes-
sage when they are engaged in conversations with multiple contacts
4.4 The When and Where of Conversations
4.4.1 When Conversations Occur
Figure 6 shows a time chart of when participants sent/received
text messages. Many of the participants being students sent text
messages towards the end of December, which coincides with ﬁ-
nals and winter break. This pattern also occurred near the end of
March when the next school term ends and spring break begins.
The topics of conversation during these time periods often related
to ﬁnal exams, school work, and procrastination.
We generally found that text messages were sent/received to-
wards the evening and night time. This is probably representative of
most participants in our study who were students. Some messages
were sent in the very early hours of the morning when participants
stayed up late. There were several occurrences of a higher number
of messages sent in the morning between 7 a.m and 10 a.m. Many
of these messages were general greeting messages (Table 2) that
started several conversations.
4.4.2 Where Conversations Occur
The ubiquitousness of text messaging enables people to commu-
nicate from anywhere. Our study spanned several vacation peri-
ods in the university quarter, so students traveled to many different
places across the country. Text messages were sent from 20 dif-
ferent states, but most of them were sent in the area around the
university. We were particularly interested in where conversations
occur - how much movement there is during a conversation. We
analyzed each of our 8590 conversations for any cell id changes
from our participants during a conversation. We considered cell
id changes to indicate that the participant was moving while main-
taining a conversation. In 87% of the conversations we found that
users did not move, in other words, they stayed at the same cell id.
This may have implications for designing around mobility for text
messaging, particularly for conversation threads.
4.5 Text Message Lengths
With the changes in text message usage, we wanted to see how
the 160 character limit would affect messages. Modern phones
have recently begun to abstract the character limits from the user
by automatically deconstructing and recombining messages that
exceed the character limit. This can help provide a seamless ex-
perience for the user since she does not need to think about how
to partition each text messages. Not all phones automatically parti-
tion messages though, so users still need to be conscious about how
messages are going to be seen by the recipient.
The majority of messages sent in our study were within the char-
acter limit. Participants sent or received an average of 50.9char-
acters per message (σ= 46.2characters). 2.4% of the messages
(1373 messages) exceeded the character limit. In our initial survey
results (Figure 2), almost 50% of our participants indicated that
they are not careful about their text message lengths. Given some
of the novel uses of text messaging, we witnessed many messages
that exceeded the 160 character limit. One individual sent an 11-
part message with each text averaging 505 characters. The longest
message had 520 characters, 3.25 times the length of a single text
Yeah but will you like me if I call it off??? and will
you still like me if I do what my parents say....???....if
I decide to like you and be your friend...be your cam-
panion ratheer than your girlfriend and in a relatikon-
ship??? ...if I call it off to be "lovers" and title it com-
panionship I feel like you won’t like me :( ...I tried to
stand up to them but to be honest I’m scared. This is
stressing my mom. Daniel has a gf it turns out...she
just found out I was lying to her and she was on verge
We are not sure what compelled the participants to send such
long messages over SMS, but believe a lot of the motivation may
be due to the immediate nature of text messaging. Email may be
better suited to handle these messages, but previous studies have re-
ported that a contact is more likely to check a mobile phone rather
than email . This trend may be changing though with the perva-
siveness of mobile internet access.
5. SMS CONTACTS
Previous mobile phone studies have found that address books
are often ﬁlled with many contacts, but only few are actually in-
volved in any communication and an even smaller number in SMS
communication [3, 7] . We found similar results on a larger scale.
Whereas Grinter and Eldridge reported that participants text with
10 −15 individuals , our study participants had text message
conversations with an average of 47.1different contacts (σ= 35.3,
min = 3,max = 148).
Not all contacts were involved in regular communication. We
analyzed the dataset to ﬁnd regular contacts based on frequency of
communication. We considered a contact to be a regular contact if
messages were exchanged on at least 10% of the days during the
study ( 12 days out of 121 total days). As text messaging is only
one method of communication in the grander setting of other con-
versation tools, we considered 10% a good metric for determining
if one was a regular contact. 51 participants had regular contacts
with an average of 5.2contacts (σ= 4.1,min = 1,max = 18).
Of all regular contacts, participants communicated with them an
average of 25.3days (σ= 14.5) for the duration of the study.
One participant communicated with his most frequent contact for
99 days of the study.
These results suggest that text messaging has grown to be a com-
mon practice even with acquaintances, based on the number of con-
tacts that our participants corresponded with. We believe that this
may be because our study was conducted among university stu-
dents who are in an environment that is more conducive towards
communication (e.g., classes and social events). This would natu-
rally increase the amount of text messages exchanged with others
based on ongoing activities. Although text messaging is used to
communicate with many more contacts, there is still a select hand-
ful of contacts that people communicate with frequently.
6. TEXT MESSAGES AND SERVICES
One of the upcoming uses of text messaging is interaction with
web services. Popular social networking services lend themselves
to having a natural mobile component. Services such as Twitter and
Facebook that rely on users sending status messages beneﬁt greatly
from a mobile interface. With the plethora of heterogeneous mobile
phones, ﬁnding a common denominator to provide access to these
services is difﬁcult. Text messaging is the natural candidate for
these web services because it is ubiquitously available on all mobile
phones. The cost of text messaging is also decreasing, which makes
it a more viable candidate for wide adoption.
We had 1,059 messages (1.8%) that were associated with a web
service (e.g., Twitter, Facebook). Although the number of mes-
sages seems small compared to the total number of messages col-
lected, we had a signiﬁcant number of participants (89%) interact
with web services through text messaging. Some examples of these
web services included mobile banking (balance information), sta-
tus updates, and instant messages. The interaction with services
demonstrates a different way that text messaging is being used. In
the case of status updates to a public social networking site (e.g.,
Twitter) a text message is a broadcast as opposed to a point-to-point
message like traditional text messaging. We have already seen that
text messaging is used to send messages to multiple recipients, even
though the technology was not designed for such uses. The avail-
ability of web services that broadcast helps complement the differ-
ent types of communication available using text messages. In the
future we would like to explore some of the motivation behind us-
ing these broadcast type services from a mobile device and compare
them to other uses of text messaging.
7. DESIGN IMPLICATIONS
Our study results indicate that text messaging has evolved in
many ways over the last decade. Text messaging has become an es-
sential and sometimes preferred method of communication. Based
on the observations in our study, we offer two design suggestions
for future designers of mobile communication tools.
First, mobile communication tools should consider supporting
simultaneous conversations. Our results show that users often com-
municate with multiple contacts on their mobile device. Multi-
threading habits over instant messaging are common on desktop
computers with chat interfaces, but become more difﬁcult to sup-
port on a constrained mobile device. There are several chat clients
available on phones today, such as Nokia Chat. These chat clients
are similar to their desktop counterparts, but are still designed with
one conversation thread in mind. Users are unable to start a multi-
recipient message and then maintain these messages as responses
arrive. Users need a way to manage simultaneous conversation
threads when initiating them. 56% of the simultaneous conversa-
tions started with a multiple recipient message and had to be man-
aged as single messages.
Second, we noticed messages involve other communication chan-
nels such as responding to Facebook messages, sending web ad-
dresses, or continuing conversations from SMS through online chat.
We believe that this provides support for exploring interoperability
between different communication mediums. We imagine that as
mobile phones become more capable and have access to a wide
range of messaging technologies, users may want an easy way to
manage these multiple communication mediums. We imagine fu-
ture interfaces integrating these technologies together so that con-
versations can seamlessly continue from online chat to SMS and
then to another medium such as email. There are concerns about
maintaining these seams since different technologies may have dif-
ferent social implications. Also as evidenced through our analy-
sis of text message lengths, users are not always careful about the
160 character limit inherent to SMS. Providing a way to switch to
appropriate message delivery methods would be useful to address
some of these limitations.
Text messaging has become a popular method of communica-
tion that has encouraged a variety of novel uses. The current tra-
jectory of text messaging seems to indicate that this communica-
tion medium will continue to grow as text messaging habits trans-
fer from teenagers, college students, and into the workplace. We
have presented an in-depth analysis of text messaging use in the
daily lives of 70 university students over a period of 4 months.
Using a custom logging tool on the participants’ mobile phones,
we captured 58,203 text messages grouped into 8,590 conver-
sations. Categorizing a subset of these conversations our study
revealed that text messaging has several novel uses such as con-
tinuing and responding to conversations from other services (e.g.,
email, instant messaging). Our study also showed that participants
engaged in simultaneous conversations with as many as 9 differ-
ent contacts on their mobile phone. Based on these observations
we made several design suggestions for future mobile communica-
tion tools. Designing for simultaneous conversations and interop-
erability between multiple communication channels holds promise
to help mobile users manage their many communication threads.
 N. Doering. "have you ﬁnished work yet? :)" communicative
functions of text messages (sms).
 R. Grinter and M. Eldridge. Wan2tlk?: everyday text
messaging. In CHI ’03: Proceedings of the SIGCHI
conference on Human factors in computing systems, pages
 R. E. Grinter and M. A. Eldridge. y do tngrs luv 2 txt msg?
In ECSCW ’01: Proceedings of European Conference on
Computer-Supported Cooperative Work, pages 219–238,
 R. E. Grinter and L. Palen. Instant messaging in teen life. In
CSCW ’02: Proceedings of the 2002 ACM conference on
Computer supported cooperative work, pages 21–30, 2002.
 R. Harper, L. Palen, and A. Taylor. The Inside Text : Social,
Cultural and Design Perspectives on SMS (The Computer
Supported Cooperative Work Series). Springer, October
 M. Ito and D. Okabe. Intimate Connections: Contextualizing
Japanese Youth and Mobile Messaging. Springer-Verlag,
 M. Ito, D. Okabe, and M. Matsuda. Personal, Portable,
Pedestrian: Mobile Phones in Japanese Life. The MIT Press,
 E.-L. Kasesniemi and P. Rautiainen. Mobile culture of
children and teenagers in Finland, pages 170–192.
Cambridge University Press, New York, NY, USA, 2002.
 A. Lamarca, Y. Chawathe, S. Consolvo, J. Hightower,
I. Smith, J. Scott, T. Sohn, J. Howard, J. Hughes, F. Potter,
J. Tabert, P. Powledge, G. Borriello, and B. Schilit. Place lab:
Device positioning using radio beacons in the wild. In
Proceedings of the Third International Conference on
Pervasive Computing, May 2005.
 R. Ling. Norwegian teens, mobile telephony and text
messages. Technical report, Telenor Research and
 D. Reid and F. Reid. Insights into the social and
psychological effects of sms text messaging. February 2004.
 P. Riedman. U.s. patiently awaits wireless texting that’s
soaring overseas. Advertising Age, 2002.
 A. M. Rubin. An examination of television viewing
motivations. Communication Research, 8:141–165, 1981.
 R. B. Rubin, E. M. Perso, and C. A. Barbato.
Conceptualization and measurement of interpersonal
communication motives. Humon Communication Research,
 D. J. Schiano, C. P. Chen, E. Isaacs, J. Ginsberg,
U. Gretarsdottir, and M. Huddleston. Teen use of messaging
media. In CHI ’02: CHI ’02 extended abstracts on Human
factors in computing systems, pages 594–595, New York,
NY, USA, 2002. ACM.
 Z. Smoreda and F. Thomas. Use of sms in europe.
 A. S. Taylor and R. Harper. Age-old practices in the ’new
world’: a study of gift-giving between teenage mobile phone
users. In CHI ’02: Proceedings of the SIGCHI conference on
Human factors in computing systems, pages 439–446, New
York, NY, USA, 2002. ACM.