What Drives Content Tagging:
The Case of Photos on Flickr
Oded Nov
Polytechnic University
New York, NY, USA
onov@poly.edu
Mor Naaman
Yahoo! Inc.
Berkeley, CA, USA
mor@yahoo-inc.com
Chen Ye
University of Illinois at Chicago
Chicago, IL, USA
cye1@uic.edu

ABSTRACT
We examine tagging behavior on Flickr, a public photo-
sharing website. We build on previous qualitative research
that exposed a taxonomy of tagging motivations, as well as
on social presence research. The taxonomy suggests that
motivations for tagging are tied to the intended target
audience of the tags – the users themselves, family and
friends, or the general public. Using multiple data sources,
including a survey and independent system data, we examine
which motivations are associated with tagging level, and
estimate the magnitude of their contribution. We find that the
levels of the Self and Public motivations, together with social
presence indicators, are positively correlated with tagging
level; Family & Friends motivations are not significantly
correlated with tagging. The findings and the use of survey
method carry implications for designers of tagging and other
social systems on the web.
Author Keywords
Tags, Flickr, photo sharing, motivations, social presence.
ACM Classification Keywords
H5.m. Information interfaces and presentation (e.g., HCI):
Miscellaneous.
INTRODUCTION
Tagging, or using keywords in order to add metadata to
content [5], is gaining much popularity in recent years
[3,5,11]. Tagging is used to annotate various types of content,
including images, bookmarks, blogs, and videos, through
web-based services such as Flickr, del.icio.us, Technorati,
and YouTube, respectively. The importance and popularity
of tagging are attributed, at least in part, to the benefits users
gain from effective sharing and organization of very large
amounts of information [2,3,11].
Users’ motivations for tagging on Flickr, a public photo-
sharing website, were explored qualitatively by Ames and
Naaman [2]. In their study, the researchers drew the
distinction between motivations stemming from three
categories of target audience for the tags added by the user.
These categories include: Self, Family & Friends, and the
general Public of Flickr users. Within each category, the
researchers identified two functional dimensions for tagging,
representing the tag's intended use: Organization and
Communication. Organization is tied to categorization and
future retrieval of images, while Communication involves
providing additional context to viewers of the image. For
example, in the Self category, the Organization function is
intended to facilitate future search and retrieval by the user,
and the Communication function involves adding context to
the image for the user’s own future recall or understanding
(e.g., "where did I take this photo?"). In the Public and
Family & Friends categories, the Organization function is
intended to facilitate future search and retrieval by others; the
Communication function reflects the motivation to add
information that explains the image and its context to
viewers. For convenience, refer to the taxonomy below in
Figure 2 (where we use the taxonomy to develop our survey
items).
The findings of [2] suggest that social presence plays a role
in tagging behavior. According to social psychology
research, behavior is affected by presence – actual, imagined,
or implied – of others [1]. The effect of perceived social
presence was found to exist even when such presence was
computer mediated (e.g., [10,12]). Perceived social presence
was also found to have a positive effect on tagging in
del.icio.us, a bookmark managing system in which tagging is
used extensively [8]. Indeed, creating social presence is seen
as key to developing successful virtual (computer-mediated)
communities [7]. In the taxonomy of [2], the Public and
Family & Friends motivations would not exist without the
user’s awareness of other people in the system who might be
viewing the user’s images.
Flickr has various avenues through which a user seeks and
perceives social presence. In this study, we look at two
indicators of social presence: groups and contacts. A Flickr
user can belong and post photos to multiple user groups,
which are normally formed around a common subject of
interest (e.g., trains, Chicago). In addition, a Flickr user can
designate other users as “contacts,” people whose photos the
user follows (contacts are often reciprocal). By marking
certain contacts as “friends” or “family,” a user may provide


Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that copies
bear this notice and the full citation on the first page. To copy otherwise,
or republish, to post on servers or to redistribute to lists, requires prior
specific permission and/or a fee.
CHI 2008, April 5–10, 2008, Florence, Italy.
Copyright 2008 ACM 978-1-60558-011-1/08/04…$5.00
CHI 2008 Proceedings · Search April 5-10, 2008 · Florence, Italy
1097

these contacts with a special access for to his semi-private
photos. When a user joins a group or adds people to his
contact list, the user implicitly accepts that his images will be
exposed to members of the group or the user’s contacts,
thereby leading to a perception of social presence, and
possibly affecting the user’s tagging behavior.
In this work we present a quantitative study which builds on
the findings of [2], and which allows us to explore the effects
of the various motivations and social presence on actual
tagging behavior on the Flickr site.
The contributions of this work include: a) a research model
capturing the effect of user actions, motivations, and social
environment on tagging; and b) a first quantitative study of
tagging motivations on the photo-sharing website Flickr.
We begin by laying out the hypotheses, and then describe the
study method and results.
RESEARCH MODEL
Since we are studying tagging behavior, the dependent
variable we measure is the total number of unique tags
applied by users to images in their Flickr photo collection.
We expect to see evidence that this dependent variable is
influenced by both stated user motivations, and by social
presence indicators, representing the user's perceived social
presence on Flickr.
Based on Ames and Naaman's findings [2], we would expect
to find the following:
H1: The level of users’ Self motivation will be positively
correlated with their number of tags.
H2: The level of users’ Public motivation will be positively
correlated with their number of tags.
H3: The level of users’ Family & Friends motivation will be
positively correlated with their number of tags.
We further hypothesize that indicators of social presence on
Flickr will be correlated with the number of tags, as
previously shown for other systems [8], and as suggested by
the Public and Family & Friends tagging motivations [2].
This view is supported by evidence that users’ viewing
activity is affected by the number of their contacts [9]. Since
on Flickr, groups and contacts can be taken to imply
perceived social presence, we would expect the following:
H4: The number of contacts a user has will be positively
correlated with the user’s number of tags.
H5: The number of groups in which a user is a member will
be positively correlated with the user’s number of tags.
Another potential driver of tagging, which serves as a control
variable, is the level of participation, as evident by the
number of images a user has in their Flickr account.
Regardless of other motivations and social factors, a larger
number of photos introduces both an opportunity and a more
pressing need for a user to tag their photos. We expect, then,
that the more photos in the user’s account, the more tags he
or she will have. Controlling for the number of images is
therefore critical for understanding the other factors that
influence tagging behavior. Notice that although the duration
of the user’s activity on Flickr may also serve as a control
variable, we feel that the number of photos in the user’s
account is a more direct measure. To confirm this, both
control variables can be tested.
To summarize, our research model attempts to explain
tagging activity using three elements: stated motivations
(Self, Family & Friends, Public), social presence indicators,
and participation level. The model is described in Figure 1.

Figure 1. Summary of the research model
METHOD
In recent years there has been an increasing use of surveys in
the HCI field (e.g., [4,6]). However, much of the research on
users’ tagging motivations so far has been qualitative, and
therefore provides a useful conceptual background but no
statistically significant quantitative assessment of the
motivations.
In this study, we execute a larger-scale study of tagging on
Flickr, using independent sources: user-reported data (via a
survey) and Flickr system data about actual usage. Using
system data is a suggested approach for avoiding common
method bias, which often poses a methodological problem in
interpreting results from survey studies [13], and can be
avoided by measuring the dependent variable using objective
data. In this study, users’ tagging data was retrieved from the
Flickr system, and therefore common method bias should not
arise in interpreting our results.
To measure the effects of different tagging motivations, we
have developed a scale based on Ames and Naaman’s
qualitative work [2]. The scale includes three constructs,
representing the three categories of intended users of the tags,
as perceived by the user: Self, Family & Friends, and Public.
For each construct, we included questionnaire items
representing both the communication and the organization
functions (we did not design the study to differentiate
between these functions). All of the motivation items in the
questionnaire were presented as statements to which users
were asked to state how strongly they agree, on a scale of 1
to 7; figure 2 contains examples of such questionnaire items.
After the preliminary scale was developed, a pilot study (N =
193) was carried out to validate the scale. An exploratory
factor analysis using principle component analysis (PCA)
was carried out and resulted in a three-factor solution. Items
showing factor loading higher than 0.6 and cross-loadings
lower than 0.4 were retained, and others were dropped. The
CHI 2008 Proceedings · Search April 5-10, 2008 · Florence, Italy
1098