Who Do You Think You Are?

: The Performance of Identity in Microblogging

Candidate Number: 62371

Supervisor: Leslie Haddon

Dissertation (MC499) submitted to the Department of Media and Communication,

London School of Economics, September 2008, in partial fulfillment of the

requirements for the MSc in Global Media and Communication.

1
Abstract

This paper examines the performance of identity through short, deliberate messages on the
‘microblogging’ website Twitter. Informed by similar research conducted on Social
Network Sites (SNS) and Mobile Social Network Sites (MSNS), and drawing upon Erving
Goffman’s study of symbolic interaction and Pierre Bourdieu’s notion of cultural capital,
this study engages in a content analysis of microblog messages and attempts to establish
whether and how users employ microblogging as a means of identity performance and
impression management.

Introduction

In the performance of identity, an individual’s choice of references to and

associations with established and recognized cultural goods or ideas play an important role

in demonstrating that identity to an audience that is observing them (Goffman 1959). This

is true across any number of circumstances and situations, from the creation and

maintenance of an ‘imagined community’ (Anderson, 1983) that supports and sustains a

shared identity amongst globally dispersed diasporic groups, to the ‘taste statements’ (Liu,

2007) that individuals make in everyday life (eg. their choice of wardrobe or their selection

of a morning newspaper).

In the context of computer-mediated communication (CMC), a new landscape of

possibility for identity construction has been developing, particularly within the field of

social network sites (SNS) which provide users a virtual space to express their interests and

taste to a community of users (normally ‘real life’ friends, but also including online

acquaintances as well as, often, the general ‘internet public’). SNS use has been growing

over the past years according to marketing research (Nielsen, 2006), and this area of study

has provided rich fields of data for social scientists to examine, which they are doing with

increasing frequency (see boyd & Ellison, 2007, for a broad review of studies in this field).

Simultaneously, a parallel set of sites and services are developing that take

advantage of the increasing power and ubiquity of mobile telephony to provide hyper-local

2
and location-based services within the overall context of SNS (Johnson, 2007). So-called

mobile social network sites (MSNS) often combine internet-based profiles and account

controls with mobile phone-based communication inputs to allow users to employ their

services to various ends while away from a computer. These appear to be especially prized

for their immediacy, a function of the state of ‘perpetual contact’ described by Katz and

Aakhus (2002), as well as their emphasis on location-based services and ad hoc

collaborative functions, such as organizing a meeting between people who are in the same

area (Rheingold, 2002: 193)

While studies in this relatively nascent field of research are being published with

increasing rapidity, there remains ample unexplored territory to examine. Considerable

work has been done on the performance of taste on profile pages in SNS’s (see, for

instance, Liu 2007, boyd 2006 and Donath & boyd, 2004), but less research has been

published on the use of MSNS’s for the same purpose. This study is an effort to introduce

further evidence into the academic record, and it is focused on examining the performance

of identity through the microblogging MSNS Twitter1.

First, an explanation of the field of study will be presented, including definitions of

key terms and a brief history of microblogging. Next, a review of key literature in cultural

capital, impression management and taste performance will assess the ways in which these

theories have been applied to studies of SNS, and will provide important insight into

applying a similar framework to studying microblog messages. Finally, a content analysis

of microblog messages from the Twitter service will be described and analyzed in an effort

to answer the following research questions:

RQ1: Are users of microblogging services employing them for the purposes of
creating and maintaining their public identity?
RQ2: If users are employing microblogs for this purpose, how are they doing so?
1 Located online at: http://twitter.com

3
Describing the Field: Explaining Microblogging

Microblogging as a distinct phenomenon is of relatively recent vintage, however its

roots are deep within the annals of CMC. According to Wikipedia, a microblog (sometimes

micro-blog) is “a form of blogging that allows users to write brief text updates (usually 140

characters) and publish them, either to be viewed by anyone or by a restricted group which

can be chosen by the user.”2 In light of this definition, we can see that one of the essential

components of microblogs are their brevity, which is imposed as a constraint on the

amount of textual characters (including letters, symbols and spaces) allowed in each

message. In the case of Twitter, this constraint has an operative purpose: it ensures that the

messages are short enough to be compatible with distribution and syndication via text

messages (SMS) to mobile phones (Diaz, 2007). Additionally, we can see that microblogs

are defined by their communicative nature: they are intended to be viewed by an audience,

whether it is broad or restricted. Much like ordinary blogs, microblogs “often reflect the

personality of the author”3, and thereby provide the opportunity for their authors to perform

their identity through ‘taste performances’ (Liu, 2007).

This combination of brevity and taste performance online does not find its origin in

microblogs, but in fact can be found throughout the history of CMC. From the ‘signature

blocks’ of Usenet, E-mail and BBS systems which allowed users to append their messages

with additional information, to IRC auto-responders and instant messaging ‘away

messages’ that display a customized message to users who attempt to contact another user

who is not immediately available to respond, the history of brief, informative

2 Wikipedia is a user-created encyclopedia, and thus subject to bias. Although it is not
generally suitable for extensive citation, its definition of microblogging is useful and the
most authoritative to be found in the absence of more legitimate work in this new area.
3 “Glossary of Internet & Web Jargon”, ‘Blog’ entry. University of California, Berkeley.

See: http://www.lib.berkeley.edu/TeachingLib/Guides/Internet/Glossary.html#Blog

4
communications in CMC is long. Although originally ‘signature blocks’ were intended to

act as mechanisms for providing contact information4, users often appropriated this

capability as a means to express themselves and their personal taste by including favored

quotes or aphorisms, photos and graphics, or links to personal or other websites (Rains &

Young, 2006). These messages acted as markers for the audience to assist in identifying the

kind of person the user is, or at any rate the kind of person that they would like others to

think that they are.

In the case of instant messaging ‘away messages’5, an analysis by Baron et al

(2005) found that users employed away messages for overt self-presentation purposes – for

instance sharing a joke with friends or posting lyrics from a favorite song as a means of

telling other users something about themselves, their taste and their interests. A subsequent

study of speech acts in ‘away messages’ by Nastri et al (2006) found that the content of

messages were “primarily informational and expressive in nature”. They concluded that

expressive messages fulfilled multiple purposes, from indicating the taste of a user through

their choice of cultural references and associations, to sharing the emotional state of mind

of the user through the choice of language and the content of messages.

Why Study Microblogging?

While microblogging appears to be simply the latest iteration in a trend of terse

message-leaving technologies, it is important to study for several reasons. First of all, while

‘signature blocks’ generally remain static from one message to the next (that is, the

statement(s) contained in the signature block do not generally change frequently),

microblogs are characterized by their dynamic nature. Properly employed, a microblog is

4 RFC1855, sec. 2.1.1 – see http://www.faqs.org/rfcs/rfc1855.html

5 See http://www.aim.com/help_faq/starting_out/getstarted.adp#five

5
meant to emphasize the sense of immediacy that the medium enables, providing a glimpse

of the status of an individual that is only minutes or hours old. This is best expressed

through the ‘prompting question’ (Mischaud 2007) of Twitter - ‘What Are You Doing?’ –

which in its very construction as a sentence suggests a constantly changing account of

continuing actions into the future (as well as a record of previous actions in the past).

Secondly, microblogs are distinct from ‘away messages’ because they have the capacity to

be active. While an ‘away message’ can be as simple as ‘I am away from my computer’ or

as meaningful as a passage from a classic text, it is not normally the case that it is active in

its engagement with its audience. Away messages can be customized and changed,

however the norms of microblogs result in a more rapid flow of new messages and it is not

uncommon for users to change their microblog message more than a dozen times each

day6. Additionally, while ‘away messages’ tend to be limited to one particular purpose –

explaining the users absence from their keyboard – microblog messages have many more

potential meanings. Consequently, the corpus of microblogs potentially contain far more

breadth of message types, cataloged for any number of different purposes. These discrete,

deliberate and active messages are rich for study and analysis in ways that previous forms

of such messages (‘away messages’, ‘signature blocks’) are not.

As a practical matter, microblogs are maintained by users who update them with

short messages as frequently as they choose. In the case of Twitter, this can be

accomplished through a number of means, including via SMS text messages and entries

composed through web-based applications (Twitter, which is privately owned by San

Francisco-based Obvious Corp., encourages third-party developers to build and distribute

applications for users to update their Twitter messages, known as ‘tweets’). The kind and

6 As will be seen in the results from the content analysis, different types of users can be
identified among whom the key differentiator is the frequency of message updates.

6
frequency of messages varies according to the user, but conventions within the user

community suggest that frequent updates (on the order of at least one a day) are the norm7.

The type of messages varies from the plainly informative (including messages that share

with the audience the physical location of a user) to the obscure (seemingly random

messages, inside jokes and utter nonsense), however in each case the user is undertaking a

deliberate performance to the audience about themselves.

Identity Construction and Performance

This study is primarily concerned with identifying whether and how users

appropriate the microblog services of Twitter as a means of expressing and performing

their identity. While Twitter provides a ‘prompting question’ that asks what users are

doing, it has been observed that users have adopted the technology for a number of other

uses beyond answering this question. In his study of Twitter messages, Mischaud (2007)

found that while users did indeed answer the prompting question, they were far more likely

to send messages to Twitter that had some other purpose. Only one third of the messages

analyzed by Mischaud were found to be answering the prompting question, with the

remainder consisting of numerous other types of messages from sharing links to engaging

in conversations with other users. In discussing his findings, Mischaud states: “The

majority of Twitter users observed in this study posted messages reflecting whatever kind

of communication they wanted to disseminate.” (ibid: 30)

Of particular interest to this study was Mischaud’s observation that users appeared

to be aware of an audience for their messages, and appeared to develop stylistic strategies

to reach out to that audience (ibid: 32). This awareness of an audience, and the subsequent

7See http://news.cnet.com/8301-17939_109-9697867-2.html for a primer on using Twitter

and the norms of the Twitter community.

7
effort to ‘play to’ that audience, is a crucial component of Canadian sociologist Erving

Goffman’s analysis of symbolic interaction in his seminal The Presentation of Self in

Everyday Life (1959). In this work, Goffman identified two kinds of expressive behavior:

the expressions that an individual gives (“communication in the traditional and narrow

sense”), and the expressions that an individual gives off (“a wide range of actions that

others can treat as symptomatic of the actor”)(both ibid: 2). The first kind of expressive

behavior is conscious and deliberate – these are the things that individuals intentionally do

and say to contribute to an impression of themselves in the eyes of others. Equally

important, Goffman argues, are the seemingly unintentional actions individuals take which

provide cues for others to determine what kind of person they are.

While Goffman’s work was focused on face-to-face interactions between

individuals, it provides a useful starting point for creating a framework of analysis for

expressive gestures in microblogging. Indeed, Goffman’s work has provided a backbone

for many studies of the performance of identity in CMC (see, for example, Liu 2007 in a

study of MySpace profile pages and Humphreys 2007’s analysis of ‘shouts’ on a MSNS

called Dodgeball) and no doubt will continue to be applied to studies of users’ efforts to

“type oneself into being” (Sundén 2003, quoted in boyd & Ellison 2007). As the nature of

the internet makes it difficult to be certain of the true identity of the user, cultural cues and

expressive actions become increasingly important both for users to tell others about

themselves as well as for others to infer characteristics about users based on their ‘taste

statements’. On interactive websites users are typically given multiple opportunities for

these displays of taste, and what now is the SNS profile page was once the selection of a

screen name in chat rooms or the choice of an avatar on bulletin boards or MUDs. In each

of these interactions the user is invited to explain themselves, and one of the ways users

8
accomplish this is by making associations and references to their taste.

The role of Cultural Capital

Discussions of taste owe much to the works of French sociologist Pierre Bourdieu,

particularly in his work Distinction: A Social Critique of the Judgement of Taste (1984).

For Bourdieu, the cultivation of taste (which he defines as manifested preference) in

cultural goods is a vital component of the construction of identity, and he suggests that an

individual’s taste is heavily influenced by their economic class. Different classes have

varying amounts of ‘cultural capital’, which is accrued by a number of methods, especially

inheritance and education. Bourdieu observed that individuals could gain social prestige

through their taste, as performed in front of others, in cultural institutions such as the

theater or the opera. The material boundaries imposed by the costs associated with

prestigious cultural venues or exclusive restaurants allow individuals to distinguish

themselves through their ability to afford such luxuries. The role of luxury, or the

perception of it, is central to the methods of distinction: it represents not only an

individual’s economic and social fitness through their ability to procure and retain goods

above and beyond the bare necessities of life, but also their superior taste in rare or

privileged cultural and material goods. Bourdieu, who situates this analysis within the

context of power relations and the domination of economic elites over the working classes,

writes: “This affirmation of power over a dominated necessity always implies a claim to a

legitimate superiority over those who, because they cannot assert the same contempt for the

contingencies in gratuitous luxury and conspicuous consumption, remain dominated by

ordinary interests and urgencies.” (Bourdieu, 1984: 56)

In other words, according to Bourdieu, when individuals signal their relative

9
cultural capital through ‘taste statements’, they are attempting to express their superior

fitness and dominant social position. What might at first glance appear to be the simple act

of purchasing shaving cream now becomes a highly symbolic assertion of status and taste.

Does the individual select the cheapest cream? Do they prefer to choose a branded can? Or

are they instead buying the boutique artisanal product that costs more and yields fewer

ounces of cream? In each case, the choice reflects a number of revelations about the

individual’s cultural capital and position within the class structure. Seen within this

context, then, few of the “seemingly innocent choices” (ibid: 269) of daily life can in fact

be seen as being mundane, or stripped of social and political meaning.

These notions of cultural capital have an important impact on the study of taste

statements in microblogs. Of course, the first taste statement that would need to be

examined through Bourdieu’s lens would be the use of a microblog in the first place, which

comes with its own attendant assumptions of cultural and economic capital. For an

individual to participate in microblogging, they must fulfill a number of criteria that

exclude large proportions of the potential user-base, and so the mere act of posting a

microblog could be considered to be a ‘taste statement’ that indicates the users’

technological savvy, economic resources (through their ability to ensure access to the

internet and mobile phone service) and the similar resourcefulness of their peer group,

among other possible signals. Investigations of class issues on SNS are relatively few,

however boyd (2007) has published one of the first comparing class status on the SNS’s

Facebook and MySpace. Although this observation is not terribly useful in itself for the

purposes of this study of taste performance on Twitter, it does nevertheless provide a

valuable (and necessary) ‘reality check’ on efforts to attempt to extend the findings of this

study to a broader audience whose difficulty in providing daily necessities prevents them

1
from having the luxury of maintaining a microblog.

When analyzing ‘taste statements’ on microblogs, notwithstanding the various

limitations mentioned above, Bourdieu’s explanation of the means of appropriating

distinction through association with cultural goods is very relevant for this study.

Describing the motivation of intellectuals in attending cultural performances and art shows,

Bourdieu writes: “They expect the symbolic profit of their practice from the work itself,

from its rarity and from their discourse about it (after the show, over a drink, or in their

lectures, their articles or their books), through which they will endeavor to appropriate part

of its distinctive value.” (Bourdieu, 1984) Brought to the present-day, the components of

the discourse about a work could include their blogs, microblogs and SNS profile pages.

Additionally, while he was describing the means used to appropriate cultural capital from

works of art by intellectuals, the same procedure can be applied to different classes in their

discourses regarding sport, music, television or other aspects of popular culture. As a

practical matter for this study, efforts of appropriation can be observed in the conscious

references to cultural products such as films or television programs, as well as more

obscure allusions to the same through recitations of lyrics from a song or memorable lines

in a film (the latter may be seen as an effort to establish still further distinction in relation

to those cultural goods by demonstrating mastery over their contents to others who are

similarly well-informed on the same products).

Bourdieu’s work was centered on the premise that socio-economic conditions

determined taste according to class lines. This is instructive and was a revelatory theory at

the time it was introduced. However, in the context of this study it is difficult to implement.

Users on Twitter are not prompted to provide any information that might suggest their

economic class, and so it is not possible to objectively apply Bourdieu’s theories to the

1
sample data. However, notions of cultural capital appear to be an essential component of

‘taste performance’ in microblogging. Indeed, as a result of a lack of true data on the

economic class of users, the analysis of cultural cues and references can indicate both the

possible realities of a users class, as well as their aspirations towards class mobility. Future

study could benefit by combining an analysis of messages with interviews or

questionnaires that establish a users class status for the purposes of comparison across class

and other boundaries. Still, analysis of ‘taste statements’ benefit immensely from the work

of Bourdieu on cultural capital and distinctions of taste.

1
Methodology

Selecting the Method

In seeking to discover whether and how users of microblogging services such as

Twitter employ them for the purpose of performing their identity, it was necessary to create

a method of measuring and classifying microblog messages. Several methods were

considered, including surveys and interviews with microblog users as well as a discourse

analysis of microblog messages, but a content analysis was deemed to be the most useful

method to accomplish the research goals. Surveys and interviews would be interesting for

determining some of the conscious choices that microblog users make when they compose

and send status updates. However, the performance of identity is as often accomplished in

the unconscious choices of individuals as in the considered and deliberate ones (Goffman,

1959), and there was a risk of interviewees internal perceptions of their ‘taste statements’

differing from what can be observed externally. Critical discourse analysis would provide

rich depth of analysis, and should be considered a useful method for future study of the

various discourses at work in microblog networks, however I was interested in developing

greater breadth of research in the field, and found that the quantitative nature of content

analysis would provide more valuable results and help establish a body of data for future

study.

Still, content analysis was not without its own drawbacks. Twitter has grown over

the years (the service launched publicly in 20068) and is reported to have millions of

registered users (Arrington, 2008), although the privately-funded company refuses to

publicly state how many users have signed up for the service. As a result of the increasing

awareness of microblogging, the volume of messages that it receives has multiplied and
8 See http://twitter.com/help/aboutus

1
averages in the hundreds per second (Payne, 2007). As a practical consequence on the

sampling for this study, the flow of messages on Twitter’s Public Timeline9 are generated

on this researcher’s computer according to the settings of Twitter’s server and are beyond

my control. Additionally, although the default setting for Twitter profiles is to have all

messages publicly displayed, an option exists for a user to make their messages private and

only viewable by other users to whom they grant access. The percentage of private Twitter

accounts is not publicly available, but it can be assumed to be statistically significant.

Indeed, as an anecdotal observation, one of the one hundred users in the sample activated

their privacy settings after sampling began, perhaps because they had been notified that

users such as this researcher were ‘following’ them.

Although these factors mean that the sample of Twitter users collected for this

research project cannot be called truly statistically random, they nevertheless were

collected through a deliberate and structured method and can be considered useful as a

convenience sample. According to Riffe et al (2005: 99), convenience samples, although

not favored, can be justified under certain circumstances and in this case proved necessary

given the lack of access to pure data held by Twitter itself (Obvious Corp. declined to

respond to emails in July requesting an opportunity to access their database). The criteria

for justifying the use of a convenience sample are identified by Riffe et al as being: (1) the

difficulty in obtaining a random sample, (2) the prohibitive cost of generating a random

sample or (3) the necessity for research in an underresearched area (ibid 99). In the case of

this study, the first criteria applies because of the technical challenges presented by the role

of Twitter’s server in delivering data and by the presence of private user accounts that were

not a part of the available data. Additionally, as noted by Mischaud (2007: 7) the field of

study is not yet well developed and there is little published material on microblogging, so
9 See http://twitter.com/public_timeline

1
the third criteria can be applied as well, with the justification of providing a basis for

further study in the future (Riffe et al: 100).

Developing the Coding Framework

Because the goal of the study was to investigate the ways that users employed

microblogs to perform their identity for others, a coding framework had to be developed

that would allow the sample messages to be categorized by their content. Informed by

Mischaud’s methods, which included the classification of each message’s content into

categories like ‘personal’ or ‘work’ (Mischaud 2007: 19), as well as a pilot study that had

been conducted on a sample of users on the competing microblogging platform of

Facebook10, a revised coding framework was constructed to allow the classification of

messages into the following categories: Sharing Links, News/Politics, Sharing Location,

Answering the Prompting Question, Asking a Question, Work, Taste Performance,

Personal, Future Plans, Conversations and Other – a catch-all category for messages that fit

no other classification (see Appendix A).

Although an effort was made to rely on previous researchers practices in fashioning

the structure so as to allow better comparison between the results, some changes were

deemed necessary, as does some explanation of the interpretations of meaning of the

categories and how they were applied feel required. Twitter status updates are short,

typically discrete messages with a maximum length of 140 characters. Even so, the brevity

of the message does not necessarily have a bearing on the complexity of its meaning, and

individual messages are capable of satisfying the terms of classification for more than one

category at a time. In response to one of the challenges discovered during the pilot study on

Facebook, as well as other large social networking sites like MySpace, began supporting
10

microblogging on their sites shortly after Twitter’s popularity rose in 2007 (Slee: 2007)

1
Facebook microblog messages, the coding schedule was revised to include guidelines for

the prioritization of categories under certain conditions in an effort to establish rules to

govern the classification of messages.

For instance, a convention exists on Twitter to allow users to signal that a given

message is intended for a specific user. The notation - @username – publicly directs the

message to the named user. In coding the sample, the guidelines instructed that any

message that began with the @username notation was to be coded in the ‘Conversation’

category, without regard for the content of the conversation. Similarly, any message that

contained a URL in the http:// format was to be coded as ‘Sharing Links’, even if it could

have potentially been interpreted as ‘News/Politics’ because the link led to an article about

the American presidential election. This system was adopted (see Appendix D for the order

of priority), although it was ultimately found to be dissatisfying – future study would be

better served by attempting to code each of multiple potential classifications in a message

instead of attempting to determine which of the meanings was most significant, a

dangerously subjective choice. Indeed, a specific study of the content of @username

messages could be warranted as they represent an interesting adoption of technology

beyond its originally intended use (see Mischaud 2007).

In order to generate the full sample of 100 users, a systematic sampling of every

fifth user with a suitable message (messages in languages other than English were

discarded) was done on July 14th, 2008 in which ten users were chosen from the Public

Timeline, a feature provided by Twitter that lists the most recent public messages in

chronological order (see Appendix C). Using a timer, further samples were taken, ten at a

time each hour for ten hours, resulting in a total of 100 users whose messages would be

recorded for seven days. These efforts were made in an attempt to broaden the possibility

1
that a given user would be able to be a part of the sample, instead of limiting it to a

particular period of time that could bias the sample towards particular time zones or

regions.

Once the sample was completed, the first round of coding was done to encode

various descriptive attributes of each user, including their gender and location if provided

by the user, the number of total messages that the user has posted since creating their

account, and whether or not they included a personal link on their profile. This information

was taken from the user’s profile page, where users can elect to share their location (88%

of users did), as well as a short biography (see Appendix B). In the case of encoding

gender, a policy of not making assumptions but allowing for common sense was employed,

as users do not in fact have a response field to positively identify their gender. However,

presuming that users were not being intentionally deceptive for the purposes of this study, a

common-sense appraisal of the user’s photograph and/or username was employed to

determine the gender for coding. In the event of any ambiguity, the user was coded as

‘Unknown Gender’ (12% of the sample had an unknown gender). While there are of course

valid concerns about whether or not the true gender of the users was ascertained by this

methodology, in practical terms there was little ambiguity.

After this round of coding had been done, the encoding of the content of individual

messages began. After seven days of sampling, the 100 users had posted a total of more

than 7,800 messages, which was determined to be a larger sample than necessary. Using a

random number generator11 the sample was reduced to 33 users with a total of 1,505

microblog messages, and the encoding of message content could begin. After encoding all

of the messages, a second coder was trained to employ the coding framework and

conducted a second coding on 619 messages from 11 users selected by systematic

11 Random numbers were selected by a sequence generator at http://www.random.org

1
sampling. These messages were compared with each other and achieved an overall inter-

coder reliability (ICR) of .75, which is comfortably above the lower bound of .70

established by Lombard et al (2005) as an appropriate level of ICR in content analysis (see

Appendix E for selected ICR findings).

A closer examination of the ICR data found that for certain categories a perfect ICR

was recorded. This is likely as a result of the prioritization schedule which prescribed strict

protocol for interpretation of messages and left little room for interpretation. As stated

previously, @username messages were expected to always be coded as Conversations,

regardless of their content as a message. Not surprisingly, then, an examination of ICR

values for Conversation categories was without any variance. The restrictive protocol of

the prioritization schedule is noticeable in several categories, however there were also

observed instances of disagreement even within categories that were governed by its strict

rules. This revealed errors in the coding schedule that allowed for an unfortunate level of

ambiguity in interpreting the type of message. According to the prioritization schedule,

messages that contain hyperlinks were meant to be coded as Sharing Links regardless of

their content. However, ICR revealed a major disconnect between coders with a poor ICR

of .48, which is well below the desired agreement. Closer examination of the data found

that the categories failed to be mutually exclusive, as messages that should have been

coded as Sharing Links became instead News (often sharing a link to a news story),

Prompting Question (normally sharing a link as a way of answering the question) and

Work (posting a link that is directly related to the users work). Having examined other

categories, notably the Taste Performance category, one can at least take some solace in the

strong ICR figures for the demonstration of cultural capital, the measurement of which was

after all the purpose of the study. In messages that were coded as being taste statements,

1
ICR showed a comfortable .77 agreement.

Nevertheless, the shortcomings of the coding schedule, and the failure of the

remedy of the prioritization schedule, loom over the findings of the study. Obviously when

constructing a coding framework the goal is to create categories that are not ambiguous.

Although there are a number of valuable results to be gleaned from this undertaking, the

weaknesses of the coding schedule have become readily apparent under close examination,

and function to limit the extent to which the study can be said to have accomplished its

goals. Future studies can, at the very least, take notice of the errors made here in an effort

to avoid similar pitfalls.

1
Results and Interpretation

The principle goal of this study was to determine whether and how users employed

microblogging for the purposes of expressing their identity. As a result, the coding variable

of most interest for the study is the seventh, the classification of message type. If

microblogging services like Twitter are useful for performing and expressing a personal

identity, then we would expect to find that a large proportion of the messages in the sample

would be categorized as Taste Performance.

2
 Chart 1 - Message Type

Other Links
4% 5%
News
16%
Conversation
26% Location
1%
Prompting Q
6%
Future Plans Asking Q
1% 1%
Personal Work
13% 6%
Taste
Performance
21%

Representing 21% of the total messages in the sample, taste performances were indeed a

high proportion of all of the messages in the same, however they were less frequent than

the actual top category of microblog message: conversation. 26% of all messages were

coded as being conversations – that is messages that began with the notation @username –

and were intended to be directed at an individual user, although they were public for all to

see.

The Role of @Conversations

It was not surprising to find conversations to be the most common message type, as

Mischaud (2007: 30) found in his study a slightly higher percentage (32%) of messages

“directed at or about a person or people within the user’s network.” Additionally the role of

prioritization that was developed to deal with messages that might have multiple meanings

was biased towards interpretation as Conversations over all others. In reality the

2
@username notation is less of a message type than a message attribute. Messages that were

encoded as conversations had a specific content type but, under the coding framework

adopted, that content type was not interpreted. While the content of a conversation could

have been of a particular type (a conversational message by user ‘tzaddi’, for instance,

reads “@reese, not sure, but maybe the DFN tag is what you’re looking for.

http://www.htmlquick.com/reference/tags/dfn.html” which would be categorized as

Sharing Links), due to limitations in the coding framework that message was simply

encoded as a conversation. Improvements in the coding framework could allow messages

to be encoded for multiple categories that can be discerned in it, or, due to the apparent

prevalence of @username messages, the nature of a message as either ‘conversation’ or

‘not conversation’ could be encoded as a separate variable, which would allow for closer

examination of the nature of conversation messages, as well as the segregation of their

conversational component from other components of the message type.

The emergence of the conversation notation on Twitter has been the subject of

some discussion, as interpersonal conversations seem at odds with the broadcast medium of

microblogging – akin to reflexively saying ‘hi, Mom!’ when a person unexpectedly finds

themselves standing in front of a television camera. Indeed, the role of conversation

messages on Twitter is not without controversy among some users in regards to etiquette

(see Popkin, 2007 or Gartenberg, 2008), especially considering the fact that Twitter

features a private, interpersonal Direct Message feature. Although Direct Messages are

both private and beyond the scope of this study, it is interesting to note that many users

appear to be happy for the public at large to be capable of seeing both who they are

interacting with (@username posts include hyperlinks that allow readers to click through to

the profile of the person named in the message) and what they are saying to them. Happily,

2
this has relevant consequences for the performance of identity, as any user is able to engage

with another, including the more high profile user accounts12. Indeed, the role of naming

the intended recipient with the @username notation enables microbloggers to engage in

taste statements in the in the mere choice of their conversation partners.

In the following message user ‘FarFromSubtle’ writes in response to a blog post by

Internet entrepreneur and Silicon Valley figure Jason Calacanis13: “@JasonCalacanis Little

slow to the draw there Calacanis”. Thus, while this message is clearly classified as a

conversation due to the use of the @username notation, it also serves to identify

‘FarFromSubtle’ as someone who is familiar with Jason Calacanis and the subject of Mr.

Calacanis’ blog post that he is responding to. He also demonstrates himself to be the kind

of person who is willing to publicly criticize him, perhaps suggesting an effort to establish

an air of intellectual competence as someone who is well-informed or even snarky. The

presence of a message directed at a celebrity is as meaningful an expression of cultural

competency and taste as is the content of the message itself, particularly in the way that it is

presented publicly to an audience of peers. Although the data to confirm it is not a part of

this study, the chances are that Mr. Calacanis did not receive this particular message among

the flood of others (Calacanis has more than 30,000 followers, any number of whom might

direct conversation messages at him daily), and that wasn’t even necessarily the point.

‘FarFromSubtle’ may well have hoped that Mr. Calacanis read his comment, but more

important to him may have been that his friends and colleagues saw it and added that

association to the constellation of cultural cues and connections that inform their regard for

‘FarFromSubtle’s character.

12 Until very recently, the user with the most followers was Kevin Rose, a well-known
figure in the so-called ‘web 2.0’ community. He was recently passed by Barack Obama.
13 Mr. Calacanis was a successful early blogger who sold the blog network Weblogs, Inc. to

America Online and continues to be an influential technology business blogger.

2
 As noted, an improvement to the coding framework would have broken out

messages that had the attributes of a conversation from other messages to be more closely

investigated, but some additional analysis of the category was possible under the

framework employed. When controlling for gender, women used the conversation notation

more frequently than men did, with 42% of all messages sent by women being coded as

conversations, compared to 25% for men. The finding that a plurality of messages sent by

women was conversational is notable, and indicates that interaction is an important

attraction to microblogging for those users. This could suggest a direction for future study,

one that is concerned with looking closely at what the content of these conversations is.

Marketers and advertisers have begun with experiments in monitoring microblog services

like Twitter for rapid-response public relations activity (Lashinsky, 2008), and even within

the confines of the sample there were interesting patterns that could be observed among the

subject of discussion throughout the body of Twitter as a whole, as well as within distinct

groupings within it.

There appeared to be a regional difference in terms of the prevalence of

conversation messages, with conversational message accounting for 53% of all messages

from the South, as compared to six, sixteen and seven percent for the West, Midwest and

Alaska/Hawaii respectively, and forty-two percent of messages from the Northeast. The

number of total updates, a rough approximation of seniority14, also showed a positive

relationship with conversational messages. The more total microblog messages a user had

sent, the more likely he was to send conversational messages, with the exception of the

most productive microbloggers who, for reasons that will be explained shortly, are less apt

to engage in conversations. Nevertheless, users with fewer than 100 total updates posted
14No data is available to ascertain the date a user account was created, but a running tally
of a user’s total messages is provided and can, when cross-referenced with the frequency of
their updates, distinguish veteran or early-adopter user accounts from new ones.

2
conversational messages 20% of the time, followed by moderate users (100-499 total

updates) at 26%, heavy users (500-999 total updates) at 30% and heavier users (1000-4999

total updates) at 32%.

The Performance of Taste: The Case of Dr. Horrible

While some of the conversational messages might have contained assertions of

cultural capital, as described in the case of ‘FarFromSubtle’ and Jason Calacanis, other

messages were more directly and unambiguously assigned to the category of ‘taste

statements’. These messages typically came in the form of an expression of affinity or

association with some form of cultural production, such as a film, place, product or idea

(“At Caesars Palace. Had a great steak at The Palm”, “Finally testing out the new Firefox.

Like it so far”). These messages, meant consciously or unconsciously as a statement of

taste, are important to study and have both theoretical significance and practical

applications. Firstly, they provide a useful mechanism for examining how individuals

manage their identities in a social space, making efforts to impress their peers as well as

express their interests and desires to others. Secondly, it provides an (imperfect) means of

observing the impact of individual cultural products on a public.

For the marketer or producer of a cultural product, there is little they can take more

professional pride in than seeing a campaign theme, slogan or website being repeated by

users on a microblog service – effectively extending the marketing message to their peers

for them. This method of seeding marketing messages within user communications is often

described as ‘viral marketing’, whereby marketing messages are passed among consumers

through their social networks. Even in the one week of sampling for this study, several

cultural products could be observed ‘going viral’ - making their way through the

2
Twittersphere by passing from user to user.

One, a web video series entitled “Dr. Horrible’s Sing Along Blog” could be

observed in the messages of several of the users in the sample. The messages carrying the

film’s name ranged from simple notifications (“watching dr horrible”) to supportive

endorsements (“just finished watching the last act of ‘Dr. Horrible’s Sing Along Blog’.

Everybody needs to go to www.drhorrible.com and watch all 3 acts”) and through to the

critical (“Am I right in assuming Dr. Horrible will only be watchable via American IP

addresses?”). Twitter users deftly employed references to Dr. Horrible in their microblogs,

associating themselves with the cultural product and expressing their taste in a performance

before the community. In a complimentary relationship, the user who included a link to the

film’s website might have gained in cultural capital among his peers by sharing his find

with them, while the producers of the film gained in free exposure and viral promotion of

their product by a valuably authentic.

As noted before, ‘taste statements’ were the second most common kind of message

in the study, consisting of just over 20% of all messages. Compared across gender, men

were found to be much more likely to express their taste in a microblog message, with 29%

of all of their messages including references to cultural products or demonstrations of

cultural knowledge, as compared to just 10% of women’s messages. More senior users,

those with between 500 and 999 total updates, had the highest percentage of messages

being ‘taste statements’, 31%. At other levels of seniority, both higher and lower, the

percentage of messages in that category fluctuated between 18 and 23 per cent.

Geographically, users from the Western region of the United States were much more likely

to post ‘taste statements’, with 38% of their messages falling into the category as compared

with percentages as low as seven and eight in the South and Alaska/Hawaii, respectively.

2
Outside the United States, 25% of posts were concerned with references to cultural

products or ideas and 18% in the Northeast of the US.

A Bestiary of Microblogger Types

In order to better understand these figures, it is necessary to take a closer look at the

users themselves and establish a few types that can be observed in the body of Twitter. One

of the first differentiators to consider is the number of messages from a user in the sample,

and in this sample there were 9 users (representing 27% of the 33 total users) who posted

more than 100 messages in the sample period. Among these 9 users, they averaged 22

messages a day, compared to the overall average of 9.7 messages a day for all users (the

overall average, when controlled for the nine biggest users, further reduced to 3.4 messages

a day). This disproportionately busy group of microbloggers can be further broken up into

three categories: heavy users, promotional users and bots.

Heavy users are maintaining personal microblogs, just like the light users, but

happen to post messages much more frequently. Sometimes the content of heavy users

messages have a strong work component, where microblogging is used as a log of tasks or

events. User ‘majornelson’ is a heavy user (with an average frequency of 28 messages a

day) with a professional streak to his microblog (29% of his messages were coded as Work

– he appears to be a journalist covering the video game industry). An example message:

“Just did a Todd Howard interview about Fallout 3 for Inside Xbox.”

Promotional users are accounts operated by individuals working in the name of an

organization, for instance in the sample there is user ‘TCU’, which is operated by Texas

Christian University and disseminates information of interest to students there (“Athletics:

Four Frogs Named to Preseason All-Conference Team”). Promotional users are interested

2
in using microblogs to gain traffic to their websites, and so their postings often include

hyperlinks (88% of ‘TCU’s postings were coded as Sharing Links). Organizational

microblogs sometimes develop a ‘character’ to their postings (such as the British Prime

Minister’s Twitter account15), while others are impersonal, handing out information with

no-nonsense style.

Bots are automated applications that use microblogs as a means of distribution for

content – often a blog or other website. One of the highest volume users in the sample,

‘celebritypaycut’, is a bot operating for the website celebritypaycut.com and provides links

to new blog posts on the site (a close inspection of the user’s 203 sample messages finds

that 86% of them are classified as Sharing Links). Although bots and promotional users

appear to have similar attributes, it is important to note the difference. Promotional users

are operated by human actors, often to be used for purposes of gaining web traffic or

providing information. Bots are automated scripts, software that operates simply to

distribute links to external websites when they are updated. The methodological framework

of this study was not sophisticated enough to distinguish between human operators posting

links and automated bots posting links, and would benefit from more flexible application in

the future.

Having identified these high volume users, it appears that future study might

benefit from constructing a methodological framework for identifying and isolating them

from other kinds of users, perhaps using a threshold percentage of diversity of message

type within each user to screen out bots that simply post links devoid of meaning or human

agency. The interest of this study is primarily in personal user accounts, however a study of

promotional accounts could also be valuable, with a possible goal of examining what types

of messages organizations relied on, whether they preferred to mostly post links to external
15 http://twitter.com/DowningStreet

2
content or attempted to adopt a ‘personality’ and drive traffic to external sites more

indirectly. Informal studies have shown that there is a danger of over-promotion through

frequent or irrelevant postings (Driver, 2008), and a more carefully controlled study could

attempt to corroborate those tentative findings.

Unfulfilled Promise: Accounting for Limits in the Coding Framework

Two message category types that had been presumed to be promising turned out to

return poor results, signifying either their importance was less than expected or that their

nature was more liable to fall afoul of limitations of the coding framework and

prioritization schedule. Messages informing of a users location, as well as ones referring to

future plans, consisted of only a combined 2.1% of all messages. This contrasted with

expectations and anecdotal observation, where messages consisting of a declarative

statement of the user’s present location are commonly seen. This type of message is seen to

be especially common when users are traveling to announce their departures and arrivals.

One probable reason for this disparity is the limitation in the coding framework imposed by

the prioritization of certain categories over others might have disproportionate effect on

location and future-plan type messages. Often when users announce their location, there is

an implicit cultural cue imbued in the naming. For instance, a user announcing their arrival

to Las Vegas may be priming their audience to expect a string up microblog messages

consistent with the ‘Sin City’ image of the place. In such a situation the announcement of

their present location is subordinated to the ‘taste statement’ of the cultural reference, and

subsequently fewer messages will be coded as sharing a location than actually do in fact

share the present location of the user.

Nevertheless, there are some kernels of information to be extracted from the data

2
related to these figures. Firstly, the Sharing Location category demonstrated a noticeable

bump when it was cross-referenced with the method of input: it hopped up to 5% of all

messages from 1% when looking at messages inputted via text message. Twitter has a

versatile platform, and can receive messages from users in various forms: inputs via the

website, inputs via SMS messages (text messages), and inputs from various third-party

software applications. Each message is encoded with metadata that reveals the method of

input used, and the results do show that the messages change depending on the method of

input employed. For users who are on the move, it appears that they are more liable to send

an update via their mobile phone to record their location (“In Palo Alto”, “At the science

center”) for the benefit of any of their friends who might need or want to know where they

are.

Users who send their messages through the website were much more likely to

discuss news and current events than other forms of input (35% of web inputs were news

related, as compared to 13% in third party applications and 0% of SMS inputs – although

one should note that the web input figure includes messages from news bots). More than

half of all SMS messages were conversations, which suggests that some SMS users employ

Twitter as a means to coordinate with friends (user KatyLinda posted the following

message by SMS: “@MichMaybeNot it should be fun! let me know if you need

directions.”). This follows an observed trend in the use of ICT’s, with social gatherings

being coordinate and planned on an ad hoc basis via mobile phones (Ling & Yttri, 2002),

and now perhaps microblogging.

Users who posted messages through third part software applications16 had a much

more varied distribution of message types. ‘Taste statements’ were the most common kind
16Programs like Twitterific and Twhirl are popular third party software applications for
Twitter – they allow users to update their microblogs, as well as collect, organize and track
the microblogs they subscribe to. See: http://twitter.com/downloads

3
of message, with a total of 23% of message falling in this category. Following that,

however, Personal Messages and Conversations each consisted of 16% of the messages,

while Sharing Links and Discussing News were each 13%, showing a quite diverse blend

of message types.

Examining Twitter Profile Pages

As described before, Twitter ‘profiles’ are very spare by the standards of social

networking sites (see Appendix B for an example). There is room for one photograph, a

username, a biography entry of 140 characters or less and one hyperlink. The coding

schedule included encoding whether the user had posted a link to a personal website or

blog or not, or whether they posted a link to non-personal site. 66% of users posted a link

to a personal site or blog, and 27% posted a link to a non-personal site (often a mainstream

news or business site). Only 6% posted no link at all. These links have potential roles to

play in identity performance, and the presence of a link to a personal website or blog

suggests some level of invitation from the user to the audience to learn more about them.

User ‘kathysart’, for instance, provides a link to her own website (kathysart.com) which

includes a blog, examples of her paintings (including opportunities to purchase her work)

and entreaties to users to sign up for an email newsletter about upcoming art shows.

The role of Twitter and other microblogging platforms is in fact connected to other

arenas of identity performance such as blogs or social networking sites, and a cursory

examination of sample users personal blogs found that they often used blogging software

plugins that allowed their Twitter status to be displayed on their blog’s main page. User

‘inko9nito’ includes a link to her personal blog (inko9nito.wordpress.com) which contains

posts about technical issues related to her work as a web designer, as well as posts related

3
to her personal life and taste expressions. The page also features a Real Simple Syndication

(RSS) display of her latest Twitter posts for users to see. The practice of embedding a

Twitter display on a personal blog suggests that users feel like it expresses their identity in

a way that is complimentary to the purposes of a personal blog. Investigations of the

relationship between the content of microblog messages and lengthier blog posts could

potentially offer interesting results of the thought process of the user, in both short and long

interval messages.

Suggestions for improving the coding framework

Having conducted a rudimentary examination of the Twitter ecosystem, it becomes

clear that the accurate study of this field requires knowledge of and accounting for these

various constituent elements – from bots to promotional users to high volume personal

users and more. As the field is still fledgling, it is hoped that some of the findings of this

survey, as well as some of its errors, will be able to inform future studies which can benefit

from these efforts. With both the depth and breadth of the field waiting to be further

explored, ample room exists for studies of both ambitious and narrow focus to come.

As demonstrated by some of the shortcomings of this study, a methodological

framework that is interested in categorizing different types of messages needs to be flexible

enough to account for complex messages. Unfortunately, in the case of this study there was

a level of ambiguity that thwarted the design of the coding schedule. Efforts to create rules

for precedence of different types of messages did result in somewhat less ambiguity in the

classification of messages, but at the expense of their depth and complexity. In the case of

identity performance, the broadness of the research question also provides a handicap for

this study. Ultimately it became difficult to concretely determine which messages were not

3
able to be categorized as ‘taste statements’, as each one was deliberately constructed and

contained both conscious and unconscious cues to the audience, as described by Goffman

(1959). A better method would be to very narrowly and specifically define what

constitutes that type of message. For instance, one could restrict the categorization of ‘taste

statements’ to messages that contain a brand name or trademark, as a means of measuring

user appropriation of cultural goods (as suggested by Bourdieu) instead of a measure of

identity performance.

Theoretical Relevance of Findings

In regards to the research goals, the data, even if flawed, does show that users

express aspects of their identity through microblogs. One in five messages coded included

a positive message by a user making a statement about their taste, and as a consequence a

performance of their identity for the benefit of the audience. Although it was determined

that @username messages intended for conversations were more commonly recorded in the

analysis than messages that were recorded as ‘taste statements’, limitations of the coding

schedule prevented complete analysis of the content of those and other types of messages

from being recorded. As described, in many cases it could be observed that conversational

messages contained other types of messages as well.

Indeed, as alluded to earlier, depending on how broadly one interprets a ‘taste

statement’ to be, all microblog messages could be construed as being deliberate

announcements of relative cultural capital as a means of performing identity to an audience

of readers. Using such a broad definition introduces difficulties in subsequently drawing

conclusions about the meaning of such a state of affairs, but does contribute to a general

3
agreement that users do in fact make use of microblogging for the purpose of taste

performance and identity management.

Regarding the second question of how users employ microblogs to this end, the data

does allow us to make some conclusions, however more importantly it indicates a direction

to take improved research methods in the future. Users were observed to engage in taste

performances through direct appropriation of cultural goods in the content of their

messages. Often this was achieved by naming a specific good (“Dr. Horrible’s Sing Along

Blog”, for instance) as a way of making the audience aware of the users familiarity with the

product. These messages could be more complex, with the user providing an opinion or

criticism of a work as a means of demonstrating how the product conformed with or

deviated from the users tastes. This more complex performance allows the user to both

‘name check’ a cultural product as well as establish themselves as a tastemaker, one whose

opinions or suggestions can influence the tastes of others.

Additionally, users employed the conversational notation of @username to allow

themselves to create associations between themselves and other users (both famous and

obscure) which serves the purpose of creating their identity as one in relation with others,

introducing the opportunity for greater social capital promotion. In conversations, users

both identified themselves as actors within the social space of the MSNS as well as

demonstrated their identity through the choice of which users they elected to direct their

conversations towards. While one might think that at times this choice is innocent and

simply in reaction to events, and this can certainly be true, it also lends itself to yet another

opportunity for taste statements, as described in the analysis of a user’s conversation

message directed at public figures of the ‘web 2.0’ community who are heavily invested in

microblogging.

3
 Further research could create a framework for identifying these tastemakers and

examine the extent to which their pronouncements on various cultural goods had a

noticeable affect on their followers. Or, a focus on @username conversational messages

could be fruitful in examining more closely the interactions between users in the social

space of microblogging. While some users might engage in conversations only with

established friends, others may have a predilection towards efforts to engage public or

notorious characters in conversation. These, among other options, suggest that microblogs

as a field of study have a rich future ahead of them.

Conclusion

This study endeavored to examine the short messages of an emerging

communication medium known as microblogging for evidence of the active performance

of identity and taste. Although the content analysis of messages from the Twitter microblog

network utilized a flawed coding schedule that resulted in severe limitations in the scope of

the application of its findings, the overall goal of establishing that users do in fact employ

microblogs for these purposes was accomplished. Combined with the suggestions for

improvement in future research, the study provides a useful starting point for more rigorous

and careful examination of this interesting and rich new field of study. While

microblogging remains a niche activity (and one that excludes large volumes of potential

users through material, cultural and other barriers), it is nevertheless growing in its

prevalence. Beyond microblogging sites such as Twitter (and its several competitors such

as Jaiku, Plurk and Pownce), mainstream SNS’s such as Facebook, MySpace and even the

more staid and business-directed LinkedIn have added microblogging functions that have

3
found popular success. As a pilot study of Facebook ‘status updates’ suggested, there could

be rich opportunities to examine the way that users in different virtual environments

employ microblogs for identity performance as well as other purposes as a result of

differences in the networks’ approach to perennial CMC issues of privacy and trust.

In any event, this study has validated, at least in the eyes of this researcher,

microblogs as a useful and important field of study. Although there are prominent critics

who suggest that microblogging sites such as Twitter encourage users to record useless and

mundane details about their daily lives, it is in the recording of these passing events,

actions and ideas that a body of microblogs like Twitter gains value as a repository of

meaningful, if sometimes ephemeral, assertions of identity and cultural capital, complete

with an automated timestamp. Microblogs have a value both on a granular level - looking

at the way an individual user expresses themselves through the choice of messages to

broadcast – and across the whole body of the population – measuring the level of attention

paid to a topic, the number of times a key word is repeated by all of the millions of users.

Whether the content of the conversation is informative (numerous news agencies,

government agencies, corporations and other organizations have created profiles and share

information and news through Twitter (Lashinsky, 2008)), entertaining (not surprisingly

one format that has taken to the hyper-local and instantaneous nature of microblogging has

been celebrity news updates) or even life-saving (numerous articles have been written on

the use of Twitter in case of disasters like fires and earthquakes (Ingram, 2008), as well as

the widely-reported ‘Twitter jailbreak’ (Simon, 2008)), it contributes to a growing database

of readings of the public psyche.

Of course one must take a critical eye to such techno-utopian notions of collective

intelligence (Levy 1999) or a ‘hive mind’. Several challenges crop up against the notion

3
that a microblog service like Twitter allows social scientists a reliable stand-in for society

broadly. The nature of Twitter and other microblog services are inherently exclusive,

requiring users to negotiate access via internet or mobile phone, creating a financial barrier.

In addition, language and skill barriers can bar access to a large population of users even if

they are equipped with the necessary information and communication technologies. These

intersecting inequalities often result in disadvantaged groups becoming further isolated

through lack of access, and act to limit the representativeness of microblogs and society

generally.

This is a legitimate and accurate criticism of microblogs, as well as the Internet and

ICT’s generally. Access is limited by barriers, typically material ones, which

disproportionately bar the poor and poorly educated from receiving the benefits of the

technology. As a consequence of these barriers, it is important to recognize the limits of

technology’s reach, and account for this problem when making projections and assessments

of the impact of digital media and ICT’s. However, having acknowledged this inherent

limiting factor, it is not then fair to say that there is no value in the data revealed through

the study of microblogs. Due attention must be made in observing that the conclusions

drawn from study do not extend beyond their bounds, but operating within the limitations

imposed by the medium we can still draw rich results.

3
Works Cited:

Anderson, Benedict. “Imagined Communities: Reflections on the Origins and Spread of

Nationalism.” New York: Verso, 1983.

AOL Instant Messenger. “FAQ About ‘I’m Away’ Message” ©2008 AOL LLC. Accessed
15 Aug 2008. http://www.aim.com/help_faq/starting_out/getstarted.adp#five

Arrington, Michael. “End of Speculation: The Real Twitter Usage Numbers.” TechCrunch
[blog] 29 April 2008.
http://www.techcrunch.com/2008/04/29/end-of-speculation-the-real-twitter-usage-
numbers/

Baron, N. S., Squires, L., Tench S., & Thompson, M. “Tethered or mobile? Use of away
messages in instant messaging by American college students.” In R. Ling & P. Pedersen
(Eds.), Mobile Communications: Re-Negotiation of the Social Sphere (pps. 138-163).
Thousand Oaks, CA: Sage.

Bourdieu, Pierre. “Distinction: A Social Critique of the Judgement of Taste.” Cambridge:

Harvard University Press, 1984.

Boyd, danah. “Friends, Friendsters, and MySpace Top 8: Writing community into being on
social network sites.” First Monday. 11.12 (2006).

Boyd, danah & Ellison, Nicole B. “Social Network Sites: Definition, History, and
Scholarship.” Journal of Computer-Mediated Communication 13.1 (2007): 210-230.

Boyd, danah. “Viewing American Class Divisions Through Facebook and MySpace.”
Apophenia Blog Essay. 24 June (2007).
<http://www.danah.org/papers/essays/ClassDivisions.html>

Diaz, Sam. “Life, in Little Chirps.” Washington Post. 9 June 2007.

Donath, J. & boyd, d. “Public Displays of Connection.” BT Technology Journal 22.4

(2004): 71-82.

Driver, Thayer. “Twitter Survey – the Results!” Weblog Entry. Tall. Eats a lot. Talks too
Much. 26 February 2008. <http://thayer18.livejournal.com/10903.html >

Gartenberg, Michael. “Twitter is not a chat room… or does it just feel that way some
times?” Jupiter Research Analyst Weblogs. 12 March 2008.
<http://weblogs.jupiterresearch.com/analysts/gartenberg/archives/2008/03/twitter_is_not.ht
ml>

Goffman, Erving. “The Presentation of Self in Everday Life.” New York: Anchor, 1959

Humphreys, Lee. “Mobile Social Networks and Social Practice: A Case Study of
Dodgeball.” Journal of Computer-Mediated Communication 13.1 (2007): 341-360

3
Ingram, Mathew. “Quaking and Twittering.” The Globe and Mail [Toronto]. 31 July 2008.

Johnson, Carolyn Y. “Social networking Sites Breaking Free From the PC.” The Boston
Globe. 17 September 2007.

Katz, J. E., & Aakhus, M. (Eds.). “Perpetual Contact: Mobile Communication, Private
Talk, Public Performance.” New York: Cambridge University Press. (2002)

Lashinsky, Adam. “The True Meaning of Twitter” Fortune. 7 August 2008.

Levy, Pierre. “Collective Intelligence: Man’s Emerging World in Cyberspace.” New York:
Basic Books. (1999)

Ling, R. & Yttri, B. “Hyper-coordination via mobile phones in Norway.” In Katz, J. E. &
Aakhus, M. (Eds.). Perpetual Contact: Mobile Communications, Private Talk, Public
Performance. New York: Cambridge University Press. (2002)

Liu, Hugo. “Social Network Profiles as Taste Performances.” Journal of Computer-

Mediated Communication. 13.1 (2007): 252-275.

Lombard, M., Snyder-Duch, J., & Campanella Bracken, C. “Practical Resources For
Assessing and Reporting Intercoder Reliability in Content Analysis Research Projects.”
[web site] retrieved August 10, 2008 from <http://www.temple.edu/sct/mmc/reliability/>

Mischaud. Edward. “Twitter: Expressions of the Whole Self.” Media@LSE Electronic

Dissertation Series. Dr. Bart Cammaerts (supervisor). London: Media@LSE Sept., 2007

Nastri, Jacqueline, Peña, Jorge and Hancock, Jeffrey T. “The Construction of Away
Messages: A Speech Act Analysis.” Journal of Computer-Mediated Communication. 11.4
(2006)

Nielsen//NetRatings. “Social Networking Sites Grow 47 Percent, Year over Year,

Reaching 45 Percent of Web users, According to Nielsen//NetRatings”. PR Newswire. 11
May, 2006.

Payne, Alex. “Scaling Twitter.” Railsconf 2007, May 17-20 2007, Portland, OR, United
States. Presentation accessed online on 15 August, 2008.
<http://www.slideshare.net/al3x/scaling-twitter-railsconf-2007>

Popkin, Helen A. S. “Twitter Nation: Nobody Cares What You’re Doing.” MSNBC 8 May
2007. <http://www.msnbc.msn.com/id/18445274/>

Rains, Stephen A. & Young, Anna M. “A Sign of the Times: An Analysis of

Organizational Members’ Email Signatures.” Journal of Computer-Mediated
Communication. 11.4 (2006).

RFC1855 sec 2.1.1. FAQs.org. Oct., 1995. FAQs.org. Accessed 15 Aug 2008

3
http://www.faqs.org/rfcs/rfc1855.html

Rheingold, Howard. “Smart Mobs: The Next Social Revolution.” Cambridge: Basic Books
(2002)

Riffe, D., Lacy, S. & Fico, F. “Analyzing Media Messages: Using Quantitative Analysis in
Research.” New York: Routledge (2005)

Simon, Mallory. “Student ‘Twitters’ His Way out of Egyptian Jail.” CNN.com. 25 April,
2008

Slee, Mark. “What’s Everyone Up To?” The Facebook Blog. 21 April 2007.
<http://blog.facebook.com/blog.php?post=2334332130>

Sunden, J. “Material Virtualities.” New York: Peter Lang (2003)

Micro-blogging. (2008, August 18). In Wikipedia, the Free Encyclopedia. Retrieved 18

August, 2008. <http://en.wikipedia.org/w/index.php?title=Micro-
blogging&oldid=232789905>

“Glossary of Internet and Web Jargon”. [web site] Retrieved on 10 August 2008 from
<http://www.lib.berkeley.edu/TeachingLib/Guides/Internet/Glossary.html>

4
Appendix A – Coding Schedule:

1. Tweet Number (1-1,505)

-a number that identifies the individual message within the sample.

2. Twitter User ID# (1-100)

-a number that specifies which account in the sample published the tweet

3. Average frequency of individual user (in tweets per day)

-total up all the tweets in 7 days by a user, divide by 7. That is the average frequency of tweets.

3. Gender of user
1 = male : 2 = female : 99 = cannot be determined
-Look for indications in username and/or photo. If there is any ambiguity, code 99.

4. Location of user
1 = West : 2 = Midwest : 3 = South : 4 = Northeast : 5 = Alaska/Hawaii : 6 = Outside USA
99 = cannot be determined
-Regional groupings are taken from US Census

5. Number of total updates

-the total number of updates posted by this user since creating their account with Twitter

6. Do they list a web links?

1 = Yes, Personal Page : 2 = Yes, but not personal page : 3 = Yes, but unclear if personal or not
4 = No link provided

7. Classify the kind of message

1 = Sharing Links : 2 = News : 3 = Sharing Location : 4 = Answering the Prompting Q
5 = Asking a Q : 6 = Work : 7 = Taste Statement : 8 = Personal : 9 = Future Plans
10 = Other : 11 = Conversation
-Note: refer to the prioritization schedule to determine which kind of message you have

8. Method of Tweeting
1 = via website : 2 = via SMS : 3 = via external application
-Note: the external app category is very broad, everything but twitter.com and twitter SMS

4
Appendix B – Sample of User Profile

In this example of a user page, the small ‘About’ section can be seen to display the various
fields that allow users to enter information about themselves for other users. Additionally,
the ‘Stats’ field displays descriptive data about the user, including his total updates (in this
case, 3,611).
Note in the list of microblog messages that each one is marked with a time stamp as well as
an indication of the method used to input the message (in this user’s case, all of his recent
microblog messages, excepting the most recent one were inputted via the web interface at
http://twitter.com. The most recent one was inputted via the third-party Twitter application
Ping.fm).

4
Appendix C – Sample of Twitter Public Timeline

This example of the Public Timeline (http://twitter.com/public_timeline) displays the most

recent public messages received by Twitter at that time. Several different types of messages
can be observed in this example, including a @username conversation by user ‘Marti_L’,
sharing links by user ‘FFFFIND’ and a taste statement by user ‘austinscorpio23’ who refers
to the Nintendo video game console system Wii.

Additionally, the three different input mechanisms are on display in this example, with a
prevalence of third-party applications (such as Twhirl, Twitterfeed and TwitterBerry), as
well as web inputs by user ‘yep_that_url’ and SMS inputs by user ‘moixoxo’

4
Appendix D – Prioritization Schedule
When coding for message type, refer to this schedule to determine the priority message type. The following
lists, in descending order from the highest priority, how to assess which category a given microblog
message will be encoded into.

11: Conversation. When a microblog message is encountered that employs the @username construction, it
is always encoded as a Conversation.

1: Sharing Links. When a microblog message is encountered that includes a hyperlink in the standard
notation (eg. http://www.example.com), it is always encoded as Sharing Links provided that it is not
contained within a Conversation message.

7: Taste Statement. When a microblog message is encountered that includes a reference to cultural or social
products, ideas, places or things (as defined within the theoretical framework of the study), it will always
be encoded as a Taste Statement provided that it is not a Conversational or Link Sharing message.

8: Personal. When a microblog message is encountered that includes self-referential information, thoughts,
concerns, anxieties or other kind of expression that are principally concerned with sharing some insight into
the user’s personal mood, status or state of mind, these will be encoded as Personal provided they do not
satisfy an above category.

4: Answering the Prompting Q. When a microblog message is encountered that includes references to
current actions (especially when constructed using the active voice – verbs ending in –ing), it will be
encoded as Answering the Prompting Q so long as they do not satisfy an above category.

5: Asking a Q. When a microblog message is encountered that concludes with a question mark (‘?’), it will
be encoded as Asking a Q so long as it does not satisfy an above category.

3: Sharing Location. When a microblog message is encountered that includes a reference to a physical
location that the user indicates they are presently at, it will be encoded as Sharing Location provided it does
not satisfy an above category.

9: Future Plans. When a microblog message is encountered that includes references to future plans
(especially if referring to specific event or date in the future), it will be encoded as Future Plans so long as
it does not satisfy an above category.

6: Work. When a microblog message is encountered that includes references to tasks or responsibilities
related to a user’s professional life, they will be encoded as Work provided that they do not satisfy any
above category.

2: News. When a microblog is encountered that makes specific reference to current events or news, it will
be encoded as News provided it does not satisfy an above category.

10: Other. When a microblog is encountered that does not satisfy any above category, it will be encoded as
Other.

4
Appendix E – Selected Inter-Coder Reliability Figures for overall agreement, as well as
broken out figures for message categories: Sharing Links, Taste Statements, Conversation

1st 2nd
UserID Coder Coder Agreement Sharing Taste Conversation
1 7 3 n n
1 7 7 y y
1 6 6 y
2 3 4 n
2 8 8 y
2 8 8 y
4 2 1 n n
4 2 1 n n
4 2 1 n n
5 6 6 y
5 11 11 y y
5 5 5 y
8 7 7 y y
8 7 7 y y
8 7 7 y y
10 1 1 y y
10 11 11 y y
10 11 11 y y
13 7 7 y y
13 11 11 y y
13 11 11 y y
14 2 4 n
14 4 1 n n
14 1 1 y y
16 8 8 y
16 11 11 y y
16 7 7 y y
17 4 8 n
17 11 11 y y
17 7 7 y y
18 7 7 y y
18 8 8 y
18 11 11 y y
19 5 8 n
19 11 11 y y
19 11 11 y y
22 1 1 y y
22 1 1 y y
22 2 1 n n
27 4 1 n n
27 5 5 y
27 7 7 y y
30 11 11 y y
30 11 11 y y
30 11 11 y y

4
32 8 8 y
32 5 5 y
32 11 11 y y
36 11 11 y y
36 11 11 y y
36 7 7 y y
37 7 7 y y
37 4 4 y
37 7 7 y y
46 4 4 y
46 4 4 y
46 1 1 y y
48 7 7 y y
48 4 4 y
48 7 7 y y
52 8 8 y
52 4 4 y
52 4 4 y
56 7 4 n n
56 4 7 n n
56 7 7 y y
64 6 1 n n
64 11 11 y y
64 11 11 y y
65 4 4 y
65 4 4 y
65 3 4 n
66 7 8 n n
66 11 11 y y
66 8 4 n
69 7 7 y y
69 7 7 y y
69 1 1 y y
72 7 7 y y
72 7 7 y y
72 7 7 y y
77 8 3 n
77 7 4 n n
77 3 4 n
80 6 4 n
80 4 4 y
80 7 4 n n
88 7 7 y y
88 11 11 y y
88 8 7 n n
90 7 7 y y
90 7 7 y y
90 7 7 y y
92 4 4 y
92 4 2 n

4
92 4 4 y
92 4 4 y
93 11 11 y y
93 11 11 y y
93 11 11 y y
100 total 13 total 31 total 23 total
75 yes 6 yes 24 yes 23 yes
25 no 7 no 7 no 0 no
0.75 0.46153846 0.77419355 1