Computers in Human Behavior xxx (2016) 1e12

Contents lists available at ScienceDirect

Computers in Human Behavior

journal homepage: www.elsevier.com/locate/comphumbeh

Full length article

Why do people watch others play video games? An empirical study on

the motivations of Twitch users
€ blom a, b, *, Juho Hamari a, c
Max Sjo
a
Game Research Lab, School of Information Sciences, University of Tampere, FIN-33014 Tampere, Finland
b
Aalto University School of Science, P.O. Box 11000, FI-00076 Aalto, Finland
c
Tampere University of Technology, Finland

a r t i c l e i n f o a b s t r a c t

Article history: This study investigates why people choose to watch others play video games, on services such as Twitch.
Received 11 August 2016 Through a questionnaire study (N ¼ 1097), we examine five distinct types of motivations from the uses
Received in revised form and gratifications perspective: cognitive, affective, personal integrative, social integrative and tension
7 October 2016
release. Information seeking is shown to be positively associated with the amount of hours that users
Accepted 21 October 2016
Available online xxx
chose to spend on the service, as well as the amount of individual streamers they choose to watch.
Furthermore, we find that tension release, social integrative and affective motivations are positively
associated with how many hours people watch streams. We also find that social integrative motivations
Keywords:
Streaming
are the primary predictor of subscription behaviour. This study lays the groundwork for understanding
Uses and gratifications the motivations to consume this emerging form of new media in the context of online games and video
eSports streams.
Media usage © 2016 Elsevier Ltd. All rights reserved.
Let's play

1. Introduction 27 million subscribers on YouTube alone in 2014 and over 40

million at the time of writing, showing the impact a single indi-
Hundreds of millions of users choose to spend their time vidual can have on the media landscape (Grundberg & Hansegard,
watching others play video games through live internet broadcasts, 2014). One might regard streaming as yet another form of broadcast
referred to as streams, on services such as Twitch. This type of new entertainment akin to online videos, but for many users it is a more
media has both been made possible and fueled by the ever manifold and holistic communication channel than mere video
increasing bandwidth of networks, advances in video packing and media content, particularly due to the high levels of interaction.
encoding technologies, a user-generated content culture, and, ul- Due to the live-broadcasting nature of video game streaming, it
timately, by the desire to see others play video games. Today, peer- offers a unique relationship between the media creator and media
to-peer internet streaming of video games is a rapidly growing consumer, thus facilitating communication between the two. Other
form of media. Recent years have seen services doubling their user forms of new media such as YouTube have already adopted prac-
base year-on-year, with current figures reaching over a hundred tices common to social network sites (SNS) (Boyd & Ellison, 2007;
million unique monthly users (Ewalt, 2014; Needleman, 2015; Lange, 2007), however video game streaming services take these
Twitch, 2015). participatory aspects one step further as the interaction is taking
Streaming is an extremely interesting context for participatory place in real time. Video game streaming also blends two distinct
online media, spearheaded by services such as YouTube, that have mediums: broadcast media and games. While television spectating
put the traditional consumer into the role of content creator (Cha, has largely been considered to be a unidirectional activity, games
Kwak, Rodriguez, Ahn, & Moon, 2007). Content creators such as are commonly perceived as a multi-directional activity requiring
PewDiePie challenge traditional media corporations, having over active user participation. Hence, a mixture of these dominant me-
dia forms leads to an interesting context of spectating video games
with a degree of interaction, thus causing an experience that is
* Corresponding author. Game Research Lab, School of Information Sciences, more passive than playing games, but at the same time more active
University of Tampere, FIN-33014 Tampere, Finland. than consuming traditional television content.
E-mail addresses: max.sjoblom@uta.fi (M. Sjo € blom), juho.hamari@uta.fi However, it is not fully clear why peer-to-peer internet
(J. Hamari).

http://dx.doi.org/10.1016/j.chb.2016.10.019
0747-5632/© 2016 Elsevier Ltd. All rights reserved.

€ blom, M., & Hamari, J., Why do people watch others play video games? An empirical study on the
Please cite this article in press as: Sjo
motivations of Twitch users, Computers in Human Behavior (2016), http://dx.doi.org/10.1016/j.chb.2016.10.019
2 €blom, J. Hamari / Computers in Human Behavior xxx (2016) 1e12
M. Sjo

streaming gathers such large crowds of spectators, and if this one of the main areas of inquiry in communication and media
growth is a sign of a more general trend in media consumption and sciences. In this vein of research, the most prevalent theoretical
information seeking, or merely a niche form of entertainment. As development and framework is perhaps the Uses and Gratification
we do not have a clear grasp of the motives driving consumption (UG) theoretical perspective (Katz, Blumler, & Gurevitch, 1973;
behaviour, we see it as paramount to explore these motivations in Katz, Gurevitch, & Haas, 1973; Rubin, 2002; Ruggiero, 2000).
order to build a deeper understanding. Therefore, in this paper we Contrary to previous media theories such as mass society the-
seek to explore and measure why so many people are choosing to ory, UG considers media to have only a limited effect. UG states that
watch others play games over the Internet, focusing specifically on the motivation behind using a certain media is a particular grati-
the context of video game streaming which is the largest form of fication that is sought (Katz, Blumler, & Gurevitch, 1974; Ruggiero,
such online live peer-to-peer media production and consumption. 2000), and posits that the user seeks out their media of choice, as
We employ data gathered through an online survey (N ¼ 1097) and an active audience, rather than the media seeking out the user
analyse the data by employing structural equation modelling. (Abercrombie & Longhurst, 2007; Baran & Davis, 2006; Wang, Fink,
& Cai, 2008). Furthermore, UG states that the media competes for
2. Background gratification with other sources than only those related to media
(Katz, Blumler et al., 1973). Large individual differences can exist, as
Video games have had a certain social spectating element to UG considers users as individuals rather than a larger mass (Katz
them from their inception. In the early days of arcade games, et al., 1974). Within UG, needs are often classified in five cate-
people would gather around the person playing the game to see gories (Cognitive, Affective, Personal Integrative, Social Integrative
how they were doing and to cheer them on (Newman, 2004), and, and Tension Release), as presented in Table 1 (Katz, Gurevitch, at al.,
later, LAN gatherings encouraged face-to-face interaction (Jansz & 1973; West & Turner, 2010).
Martens, 2005). When games moved from the arcades to living UG considers media users to be aware of their own usage, and
rooms, players were no longer subject to the stares of strangers therefore able to provide researchers with an accurate idea of their
when playing their favourite games. With the emergence of video media consumption habits and motivation (West & Turner, 2010).
game streaming, we argue that a part of this social experience has UG has been used in a wide range of different communication
been brought back to the gaming culture. While video games have research contexts, such as television (Krcmar & Greene, 1999;
been the subject of decade's worth of studies within multiple fields Schmitt, Woolf, & Anderson, 2003), personal communication
of research (for a recent overview see Quandt et al., 2015), real-time (Ishii, 2006), and also in the notion of multitasking (Wang &
video game streaming is a novel development that has potential to Tchernev, 2012). Another area where UG has been heavily used is
contribute to the growing area of games research within the in- the online context (Ko, Cho, & Roberts, 2005; LaRose & Eastin,
formation sciences, media and communication domain. 2004; Papacharissi & Mendelson, 2010; Whiting & Williams,
Streaming typically refers to conveying media content in a way 2013), including online games (Sherry, Lucas, Greenberg, &
that it is simultaneously consumed by the receiver, as opposed to Lachlan, 2006; Wu, Wang, & Tsai, 2010), Facebook (Joinson,
‘downloading’ where the received media content is saved for later 2008), video streaming (Cha, 2014; Chiang & Hsiao, 2014) and
consumption. Hence the term streaming is more concerned with Twitter (Chen, 2011; Johnson & Yang, 2009). In this study, we aim to
the delivery method of the medium rather than the exact form of look at the motives for watching video game streams through the
the medium (Gelman, Halfin, & Willinger, 1991). Internet streaming lens of UG.
has existed for a long time and in various forms. However, it is The nature of video game streaming as a new and yet unex-
important to distinguish between streaming as a technological plored medium makes it an interesting target of investigation from
solution, and the cultural phenomenon of live broadcasts of user- the UG perspective, because it enables comparisons to be drawn
generated content, also commonly referred to as streaming. between it and other, more traditional, forms of media. Watching
others play is a highly alluring subject for study, especially since it
2.1. Characteristics of video game streaming remains rather unintuitive for many people as to why watching
video games would afford meaningful gratifications (especially
In previous media research, media have been classified on the over playing games oneself). Anecdotally, it has been assumed that
basis of their fidelity and participatory nature. Media types watching others play does not provide the same thrills and affor-
requiring a higher degree of participation have been coined as cool, dances for escapism as playing games by oneself, as the spectator
while types of media where the information is presented in more has less agency over the events of the game. On the other hand,
abundance and requiring less participation have been called hot watching others play may provide social gratifications that are
(McLuhan, 1994). In research centred on video game streaming, commonly absent in a normal single-player experience. Indeed,
mixed types of media have been shown to be of importance for this social dimension has been shown to be of importance in pre-
viewers (Hamilton, Garretson, & Kerne, 2014). The mix of media vious research within video game streaming (Hamilton et al., 2014).
allows for a highly interactive experience, with the hot media Much of the previous research related to spectating video games
(video content) serving as a facilitator for interaction via the cool has concentrated on competitive gaming, commonly called eSports
media (chat functionalities). The popularity of streamed game titles (Hamari & Sjo €blom, 2017). Research within eSports has indicated
has been shown to vary greatly over time (Kaytoue, Silva, Cerf, the importance of knowledge acquisition, escapism, social inter-
Meira, & Raïssi, 2012). While the most popular games continue to action, sharing emotional connections and the competitive atmo-
be established eSports, the release of new game titles partially ac- sphere as important motivators (Cheung & Huang, 2011; Hamari &
counts for large fluctuations in the viewer distribution. Addition- € blom, 2017; Lee & Schoenstedt, 2011; Scholz, 2012; Weiss &
Sjo
ally, stream popularity was shown to follow a highly skewed Pareto Schiele, 2013; Weiss, 2011). Within the limited research on video
principle, with 10% of individual streamers accounting for 95% of game streaming that has been conducted, social interaction,
viewers (Kaytoue et al., 2012). learning and entertainment have been shown to be important as-
pects of spectatorship (Hamilton et al., 2014; Kaytoue et al., 2012;
2.2. Uses and gratifications Shaw, 2013). However, none of the previous research conducted
has aimed to quantifiably measure relationships between spectator
The question of why people consume different forms of media is gratifications and the use of streaming services. Therefore, in this

€ blom, M., & Hamari, J., Why do people watch others play video games? An empirical study on the
Please cite this article in press as: Sjo
motivations of Twitch users, Computers in Human Behavior (2016), http://dx.doi.org/10.1016/j.chb.2016.10.019
 €blom, J. Hamari / Computers in Human Behavior xxx (2016) 1e12
M. Sjo 3

Table 1
UG need types (West & Turner, 2010, p. 398).

Need type Description Media examples

Cognitive Acquiring information, knowledge, Television (news), video (“How to Install Ceramic Tile”), movies (documentaries or films based on history
comprehension e.g., The Other Boleyn Girl)
Affective Emotional, pleasant, or aesthetic experience Movies, television (sitcoms, soap operas)
Personal Enhancing credibility, confidence, and status Video (“Speaking With Conviction”)
integrative
Social Enhancing connections with family, friends, Internet (e-mail, chat rooms, Listservs, IM)
integrative and so forth
Tension release Escape and diversion Television, movies, video, radio, Internet

study we are employing UG as a theoretical guideline to under- impacted by an increased level of cognitive motivation. By watch-
standing and modelling the relationships between gratifications ing for a larger amount of hours, we hypothesise that the learning
and media consumption. experience crucial for cognitive motivation will be facilitated. We
expect the other three types of usage to also be positively impacted
by an increase in cognitive motivation.
2.3. Research model and hypotheses We expect hours watched and streamers subscribed to be
positively affected by an increased level of personal integrative
Based on the theory framework of UG, we investigate the rela- motivation. In the context of video game streaming, we predict a
tionship between five types of motivation and four distinct types of certain level of social interaction will be required to achieve a level
usages related to the consumption of video game streams. These of personal integrative fulfilment, especially as we are focusing on
usage types are: hours watched, streamers followed, streamers recognition received through the usage of the service. Hence, social
watched, and streamers subscribed to. Hours watched indicates aspects such as streamers followed are expected to be positively
how many hours per week users consume streaming content. It is affected by an increase in personal integrative motivation. We
possible to follow particular broadcasters (or streamers as they are expect a small positive correlation with subscription as we see it
commonly referred to), thus allowing them to be more easily found impacting upon received recognition on a certain level. Thus we
by users. Streamers watched measures the amount of unique hypothesise that an increased level of personal integrative moti-
streamers watched per week for an individual user. Users may pay a vation will predict an increased level of usage (Hypothesis 3).
fee of $5 to subscribe to an individual streamer, and this is our Social integrative motivations have been shown to positively
fourth type of usage investigated. A subscription must be renewed impact the usage of online media and services in previous research
monthly to retain benefits, and additionally, one user may sub- (Chen, 2011; Hamilton et al., 2014; Pai & Arnott, 2013; Scholz, 2012;
scribe to not just one, but also a number of streamers. The common Sherry et al., 2006; Whiting & Williams, 2013). We expect the same
benefits for subscription include emoticons exclusive to sub- to be true within the context of video game streaming, and that
scribers, a visual indicator of subscription status in the chat facility, increased levels of social integrative motivation will predict an
and the possibility to take part in events or raffles aimed solely at increased level of usage (Hypothesis 4). We especially expect hours
subscribers. watched and streamers followed to be positively impacted by an
Fig. 1 presents the research model based on these five main increase in social integrative motivation. Subscription furthers so-
constructs and their relationship to the usage dependant variables. cial connections and fosters a sense of belonging within the video
The five main UG constructs are depicted as latent variables. For game streaming community on many levels. As such, this is also the
clarity, the four types of usage have been grouped into one in our motivational factor where we consider subscription to be the most
visualisation of the research model, as the relationships between impacted by an increase in motivation level, as many of the benefits
the UG needs and four types of usage are identical to more general acquired via subscribing are tied to social aspects of the service
notions of usage. (Oestreicher-Singer & Zalmanson, 2013).
Based on previous research which indicates the importance of The notion of tension release and escape having an impact on
entertainment for media usage (Cheung & Huang, 2011; Hamilton use is a topic brought up in previous motivation research within
et al., 2014; Hanson & Haridakis, 2008; Papacharissi & online communities (Courtois, Mechant, De Marez, & Verleye,
Mendelson, 2010), we consider hours watched to be particularly 2009; Hanson & Haridakis, 2008; Lin, 2002; Papacharissi &
affected by an increase in affective motivation. As the entertain- Mendelson, 2010; Whiting & Williams, 2013). We hypothesise
ment aspect of affective motivation is surely provided by a certain that increased levels of tension release motivation will predict an
subset of streamers rather than all of the streamers, we expect increased level of usage (Hypothesis 5). In particular, we expect that
streamers followed and streamers watched to be positively affected hours watched will be positively impacted by an increase in tension
by increased levels of affective motivation. We expect a smaller release motivation, and that streamers watched and streamers
effect on subscription than other types of usage as, compared to the followed will also be positively impacted. We expect a large cor-
other three forms, subscription is unlikely to offer enough of a relation with hours watched, as the tension release motivation is so
payoff compared to the cost involved for a person driven by affec- closely linked to achieving a sense of escape from everyday life.
tive motivations. Hence, we propose that increased levels of af- However, we expect a small association with subscription, as it
fective motivation will predict an increased level of usage should not be heavily impacted by tension release motives.
(Hypothesis 1).
Learning and information seeking has been shown to be an in-
tegral motivator for usage in several online media contexts 3. Methodology
(Hamilton et al., 2014; Papacharissi & Mendelson, 2010; Whiting &
Williams, 2013). We expect increased levels of cognitive motivation 3.1. Sampling
to predict an increased level of usage (Hypothesis 2). Of the four
types of usage, we argue that hours watched will especially be We piloted the study with 19 respondents and launched the

€ blom, M., & Hamari, J., Why do people watch others play video games? An empirical study on the
Please cite this article in press as: Sjo
motivations of Twitch users, Computers in Human Behavior (2016), http://dx.doi.org/10.1016/j.chb.2016.10.019
4 €blom, J. Hamari / Computers in Human Behavior xxx (2016) 1e12
M. Sjo

Fig. 1. Research model.

final survey on February 26th, 2015. At launch, the end date was psychometric items used a 7-point Likert scale (1 indicating
specified as the 21st of March, but this was later extended to the “strongly disagree” and 7 indicating “strongly agree”). All of the
23rd of March. As a participatory incentive, we offered the chance psychometric items, along with their origin, can be found in
to win one of six video games from the online store Steam, worth Appendix A.
50 USD or EUR, and a raffle was conducted among valid survey To measure affective motivations, we used the perceived enjoy-
responses after the survey had concluded. The survey was pre- ment scale of Venkatesh (2000) and van der Heijden (2004, pp.
dominantly distributed through social media and social news sites 695e704). The phrase “the system” was replaced with “Twitch”
such as Reddit, Twitter and Facebook, as well as a few other forums as applicable, and one additional item was added following the
dedicated to games. phrasing convention. The cognitive motivations construct consisted
To help filter out invalid responses we omitted respondents who of two sub-constructs: information seeking about game products and
incorrectly answered a check question as well as entries with learning game strategies. For information seeking about game
missing data and finally obtained 1091 valid responses to the sur- products, the usefulness scale taken from van der Heijden (2004, pp.
vey, representing a 3.2% decrease in respondents from the raw data. 695e704), originally used in the hedonic information seeking
Amongst the valid responses, the average age was 22.9 years context, was used as a base and modified accordingly. To build our
(M ¼ 22, SD ¼ 5.9). Female respondents comprised 7.7% of our data. learning game strategies scale, items from the information seeking
Of our respondents, 93.2% reported that they had registered an scale by Papacharissi and Rubin (2000) along with an item taken
account for the Twitch service, had used the service for an average from the van der Heijden (2004, pp. 695e704) usefulness scale
of 22.1 months (M ¼ 21, SD ¼ 14.6) and 38.7% had acted as a were used. These items were modified to fit the context of video
streamer at some point. The demographic details of the re- game streaming and learning strategies. For personal integrative
spondents are displayed in greater detail in Table 2. motivations, we used the recognition by peers scale from
Hernandez, Montaner, Sese, and Urquizu (2011). Within social
3.2. Measurement integrative motivations, the companionship scale introduced by
Smock, Ellison, Lampe, and Wohn (2011), and the shared emotional
Respondents watched an average of 11.0 h per week (M ¼ 7, connection scale used by Chavis, Lee, and Acosta (2008) were used.
SD ¼ 12.1) and an average of 5.6 different streamers per week Relating to tension release, the scales of escapism, relaxing enter-
(M ¼ 4, SD ¼ 5.0). Furthermore, respondents followed an average of tainment, and habitual pass time previously introduced by Smock
26.4 streamers (M ¼ 10, SD ¼ 52.1) and subscribed to an average of et al. (2011) were used.
0.6 streamers (M ¼ 0, SD ¼ 2.5). Table 3 shows the distributions of
dependent variables. 3.3. Validity & reliability
We looked at psychometric factors that can explain the behav-
iour patterns seen among our respondents, and in this section, we The model-testing was conducted using component-based PLS-
present the psychometric scales used for our nine constructs. All SEM (Partial Least Squares Structural Equation Modelling) which is

€ blom, M., & Hamari, J., Why do people watch others play video games? An empirical study on the
Please cite this article in press as: Sjo
motivations of Twitch users, Computers in Human Behavior (2016), http://dx.doi.org/10.1016/j.chb.2016.10.019
 €blom, J. Hamari / Computers in Human Behavior xxx (2016) 1e12
M. Sjo 5

Table 2
Demographic distribution of survey.

Factor (unit) Value Factor (unit) Value

Gender (%) Male 92.3% Employment (%) Student 57.12%

Female 7.7% Full-time 22.45%
Age (years) Average 22.94 Part-time 8.49%
Median 22.00 Unemployed 10.31%
SD 5.87 Income ($) <10 000 56.48%
Education (%) None 0.18% 10 000e29 999 21.81%
Primary level 8.67% 30 000e49 999 11.41%
Secondary level 52.19% 50 000e69 999 5.11%
Upper level 38.96% 70 000e89 000 2.10%
90 000 up 3.10%

Table 3
Dependant variable grouping (range & percentage of whole).

Group Hours watched Streamers watched Streamers followed Streamers subscribed to

1 0e2 (21.6%) 0e2 (19.8%) 0e2 (20.2%) 0

2 3e5 (22.5%) 3 (18.7%) 3e6 (20.0%) 1þ
3 6e10 (25.0%) 4e5 (30.2%) 7e15 (21.9%)
4 10.5e20 (17.1%) 6e9 (15.0%) 16e39 (19.0%)
5 21þ (13.8%) 10þ (16.3%) 40þ (18.9%)

considered to be more suitable for prediction-oriented studies such watched, as well as 21.5% for streamers followed and 17% for
as the present study (Anderson & Gerbing, 1988; Chin, Marcolin, & streamers watched. For subscriptions, the model only accounted for
Newsted, 2003). Convergent validity was met since the AVE, CR and 3.7% of the variance (Fig. 2). Table 5 displays the results for each of
Alpha measures exceeded the recommended thresholds (Fornell & the five types of motivation in relation to the four types of usage.
Larcker, 1981; Nunnally, 1978). Discriminant validity was met, as From the results we can see how the initial hypotheses are sup-
the square root of the AVE of each construct was larger than its ported. In the following paragraphs we examine the results, using
correlation to any other construct (Chin, 1988; Fornell & Larcker, the same notation for statistical significance as used in Table 5
1981; Jo € reskog & So € rbom, 1996), and each measurement item (*p < 0.05, **p < 0.01, ***p < 0.001).
had the highest loading with its corresponding construct. The re- For affective motivations, three of our four hypotheses were
sults of these validations are displayed in Table 4. The validity of the supported, with positive relationships seen between the motiva-
scales and their individual items can be found in Appendix B. tion and hours watched (H1a b ¼ 0.144**), streamers watched (H1b
In order to reduce the likelihood of common method bias, we b ¼ 0.134**) and streamers followed (H1c b ¼ 0.152***). The rela-
randomized the order of the measurement items on the survey to tionship to subscription was above the threshold level of 0.05 for
limit the respondent's ability to detect patterns between the items statistical significance and thus we could not establish a relation-
(Cook, Campbell, & Day, 1979). Common method bias refers to a ship (H1d b ¼ 0.045).
situation where there is “variance that is attributable to the mea- For cognitive motivations, we found positive relations with
surement method rather than to the constructs the measures repre- hours watched (H2a b ¼ 0.089**) as well as streamers watched
sent” (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003, p. 879). The (H3b b ¼ 0.075*), thus supporting two of our hypotheses. Our re-
sample size (N ¼ 1097) satisfies several different criteria for the sults for streamers followed (H2c b ¼ 0.007) and subscription (H2d
lower bounds of sample size for PLS-SEM (Anderson & Gerbing, b ¼ 0.028) did not exceed the threshold for statistical significance.
1988; Chin & Newstad, 1999; Westland, 2010). Coupled with the low correlation, this goes to show that no relevant
relationship was to be found.
The personal integrative motivations are of particular interest,
4. Results as we found an opposite relationship to our hypotheses both for
hours watched and streamers watched (H3a b ¼ 0.177***, H3b
The model accounted for 25.8% of the variance for hours

Table 4
Fornell-Larcker criteria (main constructs in bold - values marked with asterisk are expected to correlate strongly with mother construct).

AVE CR Alpha AFF COG: PROD COG: STRAT COG PI SI: COMP SI: SEC SI TR TR: DIST TR: ESC TR: RELAX

AFF 0.749 0.922 0.887 0.865

COG: PROD 0.773 0.931 0.901 0.431 0.879
COG: STRAT 0.761 0.927 0.895 0.486 0.438 0.872
COG 0.550 0.907 0.882 0.539 0.864* 0.831* 0.742
PI 0.715 0.909 0.867 0.474 0.423 0.310 0.436 0.846
SI: COMP 0.743 0.896 0.826 0.496 0.394 0.283 0.402 0.546 0.862
SI: SEC 0.612 0.887 0.840 0.631 0.452 0.326 0.463 0.685 0.614 0.782
SI 0.537 0.902 0.875 0.643 0.477 0.345 0.489 0.696 0.856* 0.934* 0.733
TR 0.541 0.921 0.904 0.722 0.455 0.406 0.509 0.455 0.643 0.557 0.660 0.735
TR: DIST 0.663 0.887 0.829 0.641 0.410 0.364 0.458 0.389 0.559 0.479 0.573 0.902* 0.814
TR: ESC 0.723 0.886 0.808 0.487 0.363 0.289 0.387 0.406 0.596 0.458 0.572 0.817* 0.603 0.850
TR: RELAX 0.788 0.917 0.865 0.731 0.405 0.393 0.471 0.395 0.526 0.510 0.576 0.875* 0.680 0.591 0.887

€ blom, M., & Hamari, J., Why do people watch others play video games? An empirical study on the
Please cite this article in press as: Sjo
motivations of Twitch users, Computers in Human Behavior (2016), http://dx.doi.org/10.1016/j.chb.2016.10.019
6 €blom, J. Hamari / Computers in Human Behavior xxx (2016) 1e12
M. Sjo

Fig. 2. Results (*p < 0.05, **

p < 0.01, ***
p < 0.001).

b ¼ 0.105**). For the relationship between personal integrative 5.1. Theoretical implications
motivations and streamers followed, our hypothesis was supported
(H3c b ¼ 0.091*), however, the relationship between the need type With the rise of user-generated content as one of the main forms
and subscription was found to not be statistically significant (H3d of contemporary media, as discussed earlier (Cha et al., 2007), we
b ¼ 0.033). find the results of this study to offer deeper insight not only into
For social integrative motivations, our hypotheses for hours broadcast media consumption, but also into how consumers
watched (H4a ¼ 0.132**), streamers watched (H4b b ¼ 0.120*), approach games as a medium. The rise of user-generated content
streamers followed (H4c b ¼ 0.213***) and subscriptions (H4d and live broadcasting of smaller scale video productions, as seen in
b ¼ 0.150**) were all supported. Notably this is the only need type the case of video game streaming on Twitch, shows no signs of
that showed a statistically significant correlation with subscription. slowing down. This would indicate a larger paradigm shift in how
Finally, for tension release, our hypotheses for hours watched we consume media in our society. The social interaction aspect is, in
(H5a b ¼ 0.319***), streamers watched (H5b b ¼ 0.217***) and particular, facilitated by the fact that content is being broadcast in
streamers followed (H5c b ¼ 0.080*) were supported. Once again real-time, in contrast to more traditional broadcast media such as
we found the relationship between the need type and subscription television. Games have traditionally been considered to be separate
(H5d b ¼ 0.001) not to be of statistical significance. from other forms of media such as television and film (Dovey &
Kennedy, 2006). However, through the penetration and diffusion
5. Discussion & conclusions of broadcasted game streams, games are becoming a more mani-
fold media and consequently seeping onto areas of media tradi-
In this study we sought to unravel the motivations for watching tionally separate from games. The results of this study also indicate
others play video games on the internet, and to determine which of the simple fact that games fill a very real role for gratifications not
those motivations would predict how much people watch, and how only through play, but also through spectating. This can clearly be
many streamers they watch, follow and subscribe to. On a general seen from the associations between motivation factors and hours
level, our results reveal that all five classes of gratification (cogni- watched.
tive, affective, social, tension release, and personal integrative) When examining the results more closely, tension release was
were significantly associated with the main outcome variables seen as being the strongest positive predictor of how many hours
related to how many hours and how many streamers individual users watched streams. This finding supports our initial hypothesis
users watch. Interestingly however, we find that personal integra- and is in line with previous research concerning social media
tive gratifications are negatively associated with these outcomes. (Whiting & Williams, 2013), YouTube (Hanson & Haridakis, 2008),
For streamer subscriptions (which entail costs), we find that the Facebook (Papacharissi & Mendelson, 2010), online services (Lin,
gratifications investigated in the study hold less explanatory power, 2002), eSports (Hamari & Sjo €blom, forthcoming 2017) and
and the only gratification that seems to explain subscribing Internet use (Courtois et al., 2009). Moreover, tension release seems
behaviour is that concerning social integrative motivations. to be an important motivator when it comes to the other viewing

€ blom, M., & Hamari, J., Why do people watch others play video games? An empirical study on the
Please cite this article in press as: Sjo
motivations of Twitch users, Computers in Human Behavior (2016), http://dx.doi.org/10.1016/j.chb.2016.10.019
 €blom, J. Hamari / Computers in Human Behavior xxx (2016) 1e12
M. Sjo 7

Table 5
Results (*p < 0.05, **
p < 0.01, ***
p < 0.001).
2
R Hours watched Streamers watched Streamers followed Subscriber

0.258 0.17 0.215 0.037

b P CI b P CI b P CI b P CI

Affective 0.144** 0.001 0.058e0.224 0.134** 0.002 0.048e0.216 0.152*** 0.000 0.068e0.235 0.045 0.298 0.041e0.130
Cognitive 0.089** 0.007 0.023e0.155 0.075* 0.037 0.007e0.147 0.007 0.851 0.062e0.074 0.028 0.453 0.099e0.042
Personal integrative 0.177*** 0.000 0.248to0.107 0.105** 0.008 0.180to0.030 0.091* 0.021 0.016e0.169 0.033 0.407 0.044e0.112
Social integrative 0.132** 0.004 0.047e0.220 0.120* 0.020 0.021e0.222 0.213*** 0.000 0.126e0.302 0.150** 0.002 0.055e0.242
Tension release 0.319*** 0.000 0.240e0.400 0.217*** 0.000 0.128e0.300 0.080* 0.048 0.001e0.153 0.001 0.984 0.095e0.093

indicators, as it has a positive impact on not just the number of between affective motivation and subscription. As previously dis-
hours watched, but also on the numbers of streamers watched and cussed, the tangible benefits offered by subscription serve as a
streamers followed. This demonstrates that users seeking to fulfil a limiting factor to the types of needs that can be met through sub-
need for escape and diversion watch a larger amount of streams. scription. In this context we could expect to find some factors that
The fact that there is no significant relationship between tension would contribute to the need for enjoyment and entertainment. For
release and subscription is, perhaps, unsurprising, as the act of example, we might consider the potential extra content of partic-
subscription does not necessarily directly affect the needs of ten- ipatory events for subscribers to contribute something positive to
sion release. Very few of the tangible benefits acquired from sub- the overall entertainment obtained by a user, and by subscribing
scribing to a streamer help the user achieve escape or diversion. the user would then get more opportunities that could lead to
Due to this lack of concrete incentives, it seems quite natural that increased levels of entertainment. However, our results show only a
there is no significant relationship between them. As mentioned small correlation that was not statistically significant, so we have to
earlier in this study, the most common tangible benefits obtained conclude that this hypothetical correlation is not observable in our
from subscribing are access to streamer-specific chat emoticons sample.
and obtaining a visual indicator next to the user name in the chat Cognitive motivations show a small positive association with
functionality. Additionally, some streamers offer access to hours watched. This positive association supports our original hy-
subscriber-only content, such as the opportunity to play with the potheses of cognitive motivations, and is further supported by
streamer. In the context of the online music service Last.fm, activity previous research within the fields of video game streaming
in the community was seen as a more important indicator of sub- (Hamilton et al., 2014), eSports (Hamari & Sjo € blom, forthcoming
scription than the consumption of music (Oestreicher-Singer & 2017) and social media (Papacharissi & Mendelson, 2010;
Zalmanson, 2013). Even though there are significant differences Whiting & Williams, 2013). We can see that in the quest for
in subscribing to a service and subscribing to an individual knowledge and information, a user is driven to increased use of the
streamer, there are still similarities that help offer an explanation service, even if the increased usage is not large. One reason might
for this phenomenon. If community activity is an important pre- be that the value obtained subsides after a short while. It could be
dictor, it stands to reason that the needs governed by the tension that viewers driven by cognitive motivation seek fulfilment for
release category would not directly affect subscription. their need for information through other sources rather than
In regard to affective motivations, we find three statistically streams. As we asked respondents about their attitudes towards
significant associations. Firstly, there is a positive association be- learning about game strategies and information seeking about
tween affective motivations and hours watched. This is in accor- game products, these are needs that can also be met through other
dance with our hypothesis and also supported by previous research channels. Perhaps the live aspect of video game streaming is
into streaming (Hamilton et al., 2014; Shaw, 2013), eSports (Cheung detrimental to the learning experience compared to video re-
& Huang, 2011), social media (Papacharissi & Mendelson, 2010; cordings of similar content which are available through services
Whiting & Williams, 2013), and video sharing websites (Cha, such as YouTube. On the other hand, the live experience and social
2014; Hanson & Haridakis, 2008). The moderate size of the corre- interaction available between viewer and streamer allows for a
lation suggests a person might find the enjoyment they are looking level of personalisation that is not possible with pre-recorded
for after a certain amount of usage of the service, after which material. Previously, learning about games has been shown to be
additional hours become superfluous. In the case of YouTube, the a major reason for starting to watch streams (Hamilton et al., 2014).
entertainment aspect was important for comedy news watching Based on this research and also the interviews conducted at the
(Hanson & Haridakis, 2008), and it can be argued that video game start of this study, we expected learning to have one of the more
streaming falls in a similar area of hedonic consumption. The influential correlations, but this was not indicated by our results.
positive association of affective gratification on how many A small positive association is observed for streamers watched.
streamers users watch was equally of moderate size. This could Users for whom cognitive motivations are important might watch
indicate that watching a small number of different streamers, only a selected amount of streamers that offer the information and
rather than seeking out a larger number of streamers achieves the guidance that they are looking for. Interestingly, we observe no
enjoyment sought from the service. Another explanation could be significant association for cognitive motivations on streamers fol-
that the person in question has trouble finding a larger number of lowed, and it is notable that this was the only motivation type that
streamers that produce content that they think is enjoyable. There did not show a significant association with streamers followed. This
was also a positive association with streamers followed, and as seems to indicate that viewers looking to learn might not follow
streamers followed and streamers watched are quite closely any more streamers than other users do. We might expect a posi-
related, it is not surprising to see that these associations were close tive association here as not all streamers will be equally enlight-
to each other in size. A person for whom affective motivations are ening, and it would stand to reason that viewers who have
important might very well seek out the same particular streamers cognitive motivations would follow streamers that they feel teach
time after time, both watching them and choosing to follow them. them something, as this in turn would help them meet their needs.
As with tension release, no significant relationship was found As with the previous motivations, we do not find a significant

€ blom, M., & Hamari, J., Why do people watch others play video games? An empirical study on the
Please cite this article in press as: Sjo
motivations of Twitch users, Computers in Human Behavior (2016), http://dx.doi.org/10.1016/j.chb.2016.10.019
8 €blom, J. Hamari / Computers in Human Behavior xxx (2016) 1e12
M. Sjo

association with subscribing, the reasoning behind this might be might be that while people for whom recognition is important
similar to that which we have presented for the previous motiva- might watch only a small number of streamers, they still follow a
tion types. The act of subscribing does commonly not offer tangible larger amount. This might result in the fact that although they
cognitive benefits. Certain streamers might have days when the watch fewer hours per week than other users, they might want to
chat facility is only open to subscribers, and the streamers then have a larger selection of appropriate streamers to watch. Following
interact more actively. One might argue that this can facilitate a a slightly larger amount of different streamers that suit their needs
teacher and student relationship that would be beneficial for a then allows these users to pick out those they wish to watch during
viewer driven by cognitive motivations, and hence they would be the limited time they spend on the service.
inclined to subscribe. Especially for cognitive motivations, the We do not find a significant relationship between personal
notion of perceived value might be of importance. Economic value integrative motivations and subscription. This is somewhat unex-
has been shown to have a significant impact on the willingness to pected, as compared to other types of motivations, subscription
pay for social networking sites (Vock, van Dolen, & de Ruyter, 2013). offers some benefits which are more closely related to personal
If a person looks to streamers to supply them with information that integrative needs. We argue that by subscribing to a particular
might save them money, then paying for the service might seem streamer, the person in question then belongs to a more exclusive
counterintuitive. However, the extra value gained by subscribing ‘social club’ of sorts. Belonging to this clique, the person would then
might not exceed the value obtained, and thus create a negative receive positive reinforcement from the other subscribers. Perhaps
value proposition for a person driven by cognitive needs. the lack of association here goes to show that this is not the case,
Opposite to that which was expected, personal integrative mo- and that the subscriber community for a given stream is not of great
tivations show a moderate negative correlation with hours importance when it comes to receiving recognition.
watched. As the scale instrument used to measure personal inte- We find that social integrative motivations correlate with hours
grative needs was solely focused on recognition, we attempt to watched on a moderate level. This supports our initial hypothesis
explain our findings based on the need for recognition in the per- and stands to reason, as using the service for more hours per week
sonal integrative context. This result is partly in line with previous clearly gives a person more opportunities to interact with the rest
research in the context of online learning, specifically between the of the Twitch community. This result is further supported by pre-
user and their peers (Hernandez et al., 2011). Even though video vious research in video game streaming and eSports (Hamilton
game streaming is mostly a hedonistic activity, we feel that drawing et al., 2014; Scholz, 2012), social media (Chen, 2011; Pai & Arnott,
parallels to online learning is feasible when examining motivations 2013; Papacharissi & Mendelson, 2010; Whiting & Williams,
related to recognition, as there can still be a learning component 2013), and playing video games (Sherry et al., 2006). These added
present. The negative relationship between personal integrative opportunities for interaction can then translate into positive ex-
motivations and hours watched seems to indicate that the level of periences which fulfil the needs encompassed by social integrative
recognition obtained from using the service does not fulfil under- motivations. We might expect the quality of interaction to be of
lying needs. Hence, individuals for whom these needs are impor- importance for a person for whom social integrative motivations
tant use the service less than others, and these needs are likely are significant, but we also notice that the pure number of
fulfilled through other channels. This may be explained by the fact streamers watched is of importance. However, the correlation is
that the social dimension of a channel changes with its size still not very large, signifying that a large number of streamers
(Hamilton et al., 2014). A stream with a larger amount of viewers watched is not necessarily valuable in itself.
also means more people participate in the chat functionality. For The correlation with streamers followed is slightly larger than
popular streamers, the chat activity may become chaotic, and thus the previous two, indicating that people for whom social integra-
eliminate any chance for civilized discourse. As much of obtaining tive motives are important, appreciate the ability to follow
respect from other viewers and streamers relies on being able to streamers. Having a larger group of potential streamers to choose
use the chat facility to display a certain level of understanding of from gives more flexibility in choosing ones the person finds
the game or subject at hand, this presents a clear problem. The scale particularly interesting. This added flexibility might then translate
used in this study did not particularly distinguish between the to a higher level of social involvement, as the person can choose
recognition obtained from streamers and other viewers. One might streams which offer the largest social aspect from a selected list.
argue that for certain individuals these two types of recognition are Social integrative motivations are the sole types of motivation
markedly different, as has been shown to be the case when looking where we find a significant association with subscription, albeit a
at differences in motivational impact between peers and in- moderate one. This indicates that the feeling a person gains from
structors (Hernandez et al., 2011). supporting a streamer is an important factor in fulfilling their social
As with hours watched, we also find a moderate negative rela- integrative needs. By supporting a streamer and following them,
tionship between personal integrative motivations and streamers the person can develop a deeper involvement with the community
watched. As seen with hours watched, it seems that watching a and feel involved in a larger part of the communities shared ex-
larger number of different streamers does not contribute positively periences. As subscription activity shows up as an automated
to the level of perceived recognition obtained. The negative rela- message in the chat facility, we observed that other viewers often
tionship might also be a sign that the person enjoys a very partic- offer positive encouragement to new and renewing subscribers.
ular set of streamers, one that is quite small and not easily The impact of social factors on choosing to subscribe to a service
supplemented by new streamers. A person seeking recognition has been shown to be relevant for other online services
might have found a few select streamers that offer a high level of (Oestreicher-Singer & Zalmanson, 2013), thus strengthening our
interaction, as well as communities that offer support and view that social integrative motives are highly relevant to sub-
encouragement to other community members. scription. Services offering paid subscriptions seldom offer users
Surprisingly, we find a small positive correlation between per- the possibility to contribute directly to a single content producer,
sonal integrative motivations and streamers followed. This is and there is therefore potential for a higher level of social gratifi-
interesting, as for the other types of motivations studied we saw cation to be obtained through subscription. In the context of video
correlations of similar size for streamers watched and streamers game streaming, when a user subscribes to a streamer, he or she
followed, but here we find correlations which work in the opposite can feel good about supporting that particular streamer and
direction for the same two categories. One explanation for this perhaps enabling them to continue producing content.

€ blom, M., & Hamari, J., Why do people watch others play video games? An empirical study on the
Please cite this article in press as: Sjo
motivations of Twitch users, Computers in Human Behavior (2016), http://dx.doi.org/10.1016/j.chb.2016.10.019
 €blom, J. Hamari / Computers in Human Behavior xxx (2016) 1e12
M. Sjo 9

5.2. Practical implications This goes to show that merely having a chat is not enough for many
viewers that are highly dedicated, demonstrating the need for
Based on the broader observations and theoretical implications streaming service developers to further integrate tools and services
presented in the previous section, we would like to further our for social interaction into the core activity of stream spectating.
discussion of gratifications within video game streaming through This strong positive effect of community does, however, not
some recommendations related to practical implications. carry over to personal integrative motivations and receiving
From a game development standpoint, the fact that games are recognition. The negative association found for building one's
being turned from sole playing experiences into spectator experi- reputation and watching streams would indicate that there is work
ences increases the amount of facets that should be taken into to be done within that domain for service developers. While not all
account in game design. Content creators are increasingly turning forms of media can expect to obtain high levels of gratifications
to games as a new media of conveying a message to their audience across the board, we argue that by focusing on the problems behind
in the form of both video game streaming and videos on services this negative association, one could improve the situation. As dis-
such as YouTube. These game videos serve as a framework for a cussed previously, a problematic dynamic seems to exist between
host of new actors in the media landscape, following a more general the amount of viewers and the potential ability to interact in a
paradigm shift away from large scale productions and towards meaningful way for receiving recognition. That is to say, when
user-generated content. Developers are advised to keep in mind channels increase in size, communication through chat becomes
that their products are thus employed in a variety of communica- more problematic as there is more communication noise generated
tive purposes beyond playing. Streaming has already been shown by the increased number of chat messages. Interestingly, building
to have a significant impact when it comes to game sales reputation has been shown to positively affect continued use for
(Hernandez, 2016), and by taking into account the spectator video sharing sites such as YouTube (Chiang & Hsiao, 2014), so
element of games, companies can achieve a competitive advantage perhaps through the usage of the recommendation systems used in
over each other if they can capitalize on this development in a services like YouTube, streaming services could also increase the
timely manner. How to make a game more appealing for spectators personal integrative gratifications obtained from watching streams.
is no easy task, and not an activity this paper aimed to answer,
however, we suggest to investigate particularly the social aspects
related to both gameplay and game spectatorship as also prompted 5.3. Future research directions
by the results of the present study.
As was noted earlier in this paper, cognitive motivations, such as The topic of this study offers many potentially fruitful avenues
information seeking, was not significantly associated with how for further research. While this study investigated the motivational
much people watch streams, while at the same time respondents types commonly used in UG research, there remain factors worthy
did report receiving related gratification from stream consumption. of study in order to build a deeper understanding of the video game
This might imply that viewers seek to watch streams for other streaming phenomena. For example, comparisons of motivation
reasons but might receive cognitive gratification as a by-product. levels between various types of demographic factors, such as
That is to say, users primarily use other forms of media to fully gender, country of residence, education and income could be
fulfil these cognitive needs. Video sharing services such as YouTube valuable. Investigating differences between service usage habits
feature a host of videos detailing how to play games, and perhaps a could also prove valuable. For example, we could identify differ-
major contributor is the fact that users can pause videos, rewind ences in player motivations in different types of game genres. Game
and watch sections multiple times. This facet is challenging when it genres are inherently different, as some games are highly
comes to live broadcasting, and perhaps rather than only offer ar- competitive, while others offer more of a free-form playing expe-
chives of live streams, streaming services could look to the possi- rience. We believe this not only impacts on the playing experience,
bility of incorporating tools for making communication between but is also reflected in the viewing experience. Subscriptions are a
streamers and spectators more seamless. This might then, in turn, significant indicator of service usage and indicate a willingness to
facilitate learning in the form of something more akin to a teacher- pay for content, however, this study was not able to obtain a high
student relationship. Perhaps another angle to approach this, for level of prediction when it comes to subscription motivations.
stream services, is to highlight particularly educative streams in an Therefore, it would also be of great interest to further investigate
effort to make it easier for users to find them. the types of motivations that drive subscription behaviour.
Perhaps the strongest results of the present study highlight the Furthermore, the personal integrative motivations investigated
fact that social factors are an immensely important aspect of the in this study were limited to recognition. In future research it might
consumer experience of streaming. The results show that feeling a be beneficial to investigate the inclusion of more than one scale for
sense of community in the watching experience not only increases identifying different aspects of personal integrative motivations,
how much people watch streams, but perhaps more importantly, especially given the results produced by this study, where we found
was also the strongest determinant of following streamers and a negative association between recognition and hours watched.
subscribing. Therefore, it seems clear that game developers,
streaming platform developers and streamers alike would be
Disclosure statement
strongly advised to attempt to increase the degree to which the
viewers experience communality and sense of belonging. This is
No competing financial interests exist.
particularly true for streamers, as attracting highly dedicated
viewers that then transition into subscribers, generating revenue, is
something of great interest for the streamer in question. Currently, Acknowledgements
social aspects on streaming services are mainly facilitated through
chat functionalities, following and subscribing, although many The research has been carried out as part of research projects
streamers take to using ancillary services, such as social media and (40009/16, 40111/14) funded by the Finnish Funding Agency for
private discussion groups, to maintain their streaming community. Innovation (TEKES).

€ blom, M., & Hamari, J., Why do people watch others play video games? An empirical study on the
Please cite this article in press as: Sjo
motivations of Twitch users, Computers in Human Behavior (2016), http://dx.doi.org/10.1016/j.chb.2016.10.019
10 €blom, J. Hamari / Computers in Human Behavior xxx (2016) 1e12
M. Sjo

Appendix A. Psychometric items

Item Related UG need type Scale origin AVE CR Mean

Enjoyment (ENJ) Affective Venkatesh, 2000 0.75 0.92 5.42

ENJ_1: I find using Twitch to be enjoyable.
ENJ_2: Using Twitch is exciting.
ENJ_3: I have fun using Twitch.
ENJ_4: Using Twitch is entertaining
Information seeking about game products (IS_PROD) Cognitive van der Heijden, 2004 0.77 0.93 4.57
IS_PROD_1: Using Twitch, I can better decide which game I want to play than in the past
IS_PROD_2: Using Twitch, I am better informed about new games I consider playing
IS_PROD_3: Using Twitch, I find games I would not otherwise have found.
IS_PROD_4: Using Twitch, I can better decide whether I want to play a particular game or not
Learning about game strategies (LSTRA) Cognitive Papacharissi & Rubin, 2000 0.76 0.93 5.3
LSTRA_1: Watching Twitch, I am better informed about new game strategies
LSTRA_2: Watching Twitch helps me get information on learning to play games.
LSTRA_3: Watching Twitch helps me look for information on game tricks.
LSTRA_4: Watching Twitch helps me see what game tactics are out there.
Recognition (REC) Personal integrative Hernandez et al., 2011 0.72 0.91 4.1
REC_1: I like when other Twitch users take my comments into account
REC_2: I feel good when my comments prove to other Twitch users that I have knowledge about the game being played.
REC_3: I try that my comments improve my reputation among other Twitch users.
REC_4: I like when streamers on Twitch take my suggestions into consideration.
Companionship (COMP) Social integrative Smock et al., 2011 0.74 0.90 3.57
COMP_1: Using Twitch, I don't have to be alone.
COMP_2: I use Twitch when there's no one else to talk or be with
COMP_3: Using Twitch makes me feel less lonely
Shared emotional connection (SEC) Social integrative Chavis et al., 2008 0.61 0.89 3.81
SEC_1: It is very important to me to be a part of the Twitch community.
SEC_2: I spend time with other Twitch community members a lot and enjoy spending time with them.
SEC_3: I expect to be a part of the Twitch community for a long time.
SEC_4: Members of the Twitch community have shared important events together.
SEC_5: Members of the Twitch community care about each other.
Escape (ESC) Tension release Smock et al., 2011 0.72 0.89 4.18
ESC_1: Using Twitch, I can forget about school, work, or other things
ESC_2: Using Twitch, I can get away from the rest of my family or others
ESC_3: Using Twitch, I can get away from what I'm doing.
Distraction (DIST) Tension release Smock et al., 2011 0.66 0.89 5.03
DIST_1: Using Twitch is a habit, just something I do.
DIST_2: When I have nothing better to do, I use Twitch.
DIST_3: Using Twitch passes the time away, particularly when I'm bored
DIST_4: Using Twitch gives me something to do to occupy my time.
Relaxation (RELAX) Tension release Smock et al., 2011 0.79 0.92 5.38
RELAX_1: Watching Twitch allows me to unwind.
RELAX_2: Watching Twitch relaxes me
RELAX_3: Watching Twitch is a pleasant rest

Appendix B. Discriminant validity of psychometric items,

corresponding scale values bolded

ENJ IS_PROD LSTRA REC COMP SEC DIST ESC RELAX

COMP_1 0.454 0.345 0.240 0.518 0.900 0.590 0.456 0.522 0.456
COMP_2 0.409 0.293 0.257 0.388 0.796 0.443 0.552 0.507 0.438
COMP_3 0.419 0.376 0.239 0.495 0.885 0.545 0.452 0.513 0.468
DIST_1 0.485 0.339 0.316 0.313 0.440 0.391 0.754 0.479 0.490
DIST_2 0.467 0.309 0.303 0.267 0.425 0.340 0.825 0.406 0.511
DIST_3 0.596 0.368 0.312 0.350 0.500 0.425 0.882 0.564 0.645
DIST_4 0.529 0.316 0.255 0.332 0.451 0.403 0.789 0.507 0.557
ENJ_1 0.855 0.353 0.418 0.352 0.369 0.480 0.555 0.382 0.599
ENJ_2 0.804 0.362 0.418 0.462 0.462 0.592 0.482 0.451 0.584
ENJ_3 0.908 0.384 0.417 0.417 0.466 0.585 0.584 0.420 0.686
ENJ_4 0.890 0.393 0.432 0.417 0.424 0.532 0.591 0.440 0.655
ESC_1 0.431 0.293 0.250 0.337 0.518 0.389 0.563 0.874 0.536
ESC_2 0.333 0.298 0.214 0.357 0.491 0.390 0.415 0.785 0.418
ESC_3 0.467 0.335 0.269 0.346 0.513 0.392 0.548 0.888 0.543
IS_PROD_1 0.384 0.895 0.398 0.395 0.381 0.405 0.381 0.324 0.351
IS_PROD_2 0.387 0.916 0.466 0.378 0.334 0.414 0.351 0.311 0.347

€ blom, M., & Hamari, J., Why do people watch others play video games? An empirical study on the
Please cite this article in press as: Sjo
motivations of Twitch users, Computers in Human Behavior (2016), http://dx.doi.org/10.1016/j.chb.2016.10.019
 €blom, J. Hamari / Computers in Human Behavior xxx (2016) 1e12
M. Sjo 11

(continued )

ENJ IS_PROD LSTRA REC COMP SEC DIST ESC RELAX

IS_PROD_3 0.372 0.800 0.280 0.360 0.351 0.415 0.356 0.318 0.377
IS_PROD_4 0.376 0.900 0.377 0.357 0.323 0.361 0.358 0.326 0.356
LSTRA_1 0.450 0.354 0.904 0.249 0.225 0.272 0.324 0.241 0.347
LEARN_2 0.429 0.472 0.877 0.302 0.280 0.318 0.326 0.261 0.364
LSTRA_3 0.397 0.367 0.822 0.295 0.287 0.295 0.319 0.273 0.339
LSTRA_4 0.417 0.323 0.884 0.233 0.193 0.252 0.300 0.233 0.319
REC_1 0.431 0.378 0.252 0.903 0.483 0.614 0.358 0.340 0.360
REC_2 0.353 0.307 0.268 0.814 0.429 0.507 0.280 0.339 0.298
REC_3 0.365 0.337 0.227 0.806 0.459 0.629 0.274 0.318 0.292
REC_4 0.448 0.403 0.305 0.856 0.471 0.554 0.398 0.376 0.381
RELAX_1 0.537 0.372 0.274 0.317 0.445 0.416 0.530 0.532 0.850
RELAX_2 0.688 0.348 0.368 0.362 0.482 0.460 0.634 0.533 0.915
RELAX_3 0.713 0.359 0.400 0.370 0.472 0.480 0.643 0.510 0.896
SEC_1 0.466 0.364 0.225 0.613 0.570 0.848 0.369 0.387 0.398
SEC_2 0.394 0.343 0.193 0.597 0.509 0.823 0.309 0.334 0.339
SEC_3 0.666 0.412 0.347 0.494 0.462 0.747 0.537 0.426 0.512
SEC_4 0.511 0.343 0.309 0.476 0.408 0.725 0.387 0.355 0.396
SEC_5 0.446 0.308 0.219 0.484 0.439 0.762 0.283 0.290 0.359

References Hamilton, W. A., Garretson, O., & Kerne, A. (2014). Streaming on twitch: Fostering
participatory communities of play within live mixed media. In CHI '14 pro-
ceedings of the SIGCHI conference on human factors in computing systems (pp.
Abercrombie, N., & Longhurst, B. J. (2007). The Penguin dictionary of media studies.
1315e1324). New York: ACM.
Penguin.
Hanson, G., & Haridakis, P. (2008). YouTube users watching and sharing the news: A
Anderson, J. C., & Gerbing, D. W. (1988). Structural equation modeling in practice: A
uses and gratifications approach. The Journal of Electronic Publishing, 11(3).
review and recommended two-step approach. Psychological Bulletin, 103(3),
van der Heijden, H. (2004). User acceptance of hedonic information systems. MIS
411e423.
Quarterly.
Baran, J. S., & Davis, K. D. (2006). Mass communication theory: Foundation and future.
Hernandez, D. (2016). Game creator success on Twitch: Hard numbers. Retrieved
Boyd, d. M., & Ellison, N. B. (2007). Social network Sites: Definition, history, and
6.10.2016 from Twitch: https://blog.twitch.tv/https-blog-twitch-tv-game-
scholarship. Journal of Computer-Mediated Communication, 13, 210e230.
creator-success-on-twitch-hard-numbers-688154815817.
Cha, J. (2014). Usage of video sharing websites: Drivers and barriers. Telematics and
Hernandez, B., Montaner, T., Sese, F. J., & Urquizu, P. (2011). The role of social mo-
Informatics, 31, 16e26.
tivations in e-learning: How do they affect usage and success of ICT interactive
Cha, M., Kwak, H., Rodriguez, P., Ahn, Y., & Moon, S. (2007). I tube, you tube,
tools? Computers in Human Behavior, 6, 2224e2232.
everybody tubes: Analyzing the world's largest user generated content video
Ishii, K. (2006). Implications of Mobility: The Uses of Personal Communication
system. In Proceedings of the 7th ACM SIGCOMM conference on internet mea-
Media in Everyday Life. Journal of Communication, 56(2), 346e365.
surement (pp. 1e14). ACM.
Jansz, J., & Martens, L. (2005). Gaming at a LAN event: The social context of playing
Chavis, D. M., Lee, K. S., & Acosta, J. D. (2008). The sense of community (SCI) revised:
video games. New media & society, 7(3), 333e355.
The reliability and validity of the SCI-2. In 2nd international community psy-
Johnson, P. R., & Yang, S.-U. (2009). Uses and gratifications of Twitter: An exami-
chology conference. Lisbon.
nation of user motives and satisfaction of Twitter use. In Communication tech-
Chen, G. M. (2011). Tweet this: A uses and gratifications perspective on how active
nology division of the annual convention of the association for education in
twitter use gratifies a need to connect with others. Computers in Human
journalism and mass communication. Boston.
Behavior, 27, 755e762.
Joinson, A. N. (2008). Looking at, looking up or keeping up with people? Motives
Cheung, G., & Huang, J. (2011). Starcraft from the stands: Understanding the game
and use of Facebook. In Proceedings of the SIGCHI conference on human factors in
spectator. In CHI '11 proceedings of the SIGCHI conference on human factors in
computing systems (pp. 1027e1036). ACM.
computing systems (pp. 763e772). New York: ACM.
€reskog, K. G., & So
Jo € rbom, D. (1996). LISREL 8: User's reference guide. Scientific
Chiang, H.-S., & Hsiao, K.-L. (2014). YouTube stickiness: The needs, personal, and
Software International.
environmental perspective. Internet Research, 25(1), 85e106.
Katz, E., Blumler, J. G., & Gurevitch, M. (1973). Uses and gratifications research. The
Chin, W. W. (1988). The partial least squares approach for structural equation
Public Opinion Quarterly, 37(4), 509e523.
modeling. In G. A. Marcoulides (Ed.), Modern methods for business research (pp.
Katz, E., Blumler, J. G., & Gurevitch, M. (1974). Utilization of mass communication by
295e336). Lawrence Erlbaum Associates.
the individual. In E. Katz, & J. G. Blumler (Eds.), The uses of mass communications:
Chin, W. W., Marcolin, B. L., & Newsted, P. R. (2003). A partial least squares latent
Current perspectives on gratifications research (pp. 19e34). Bevery Hills, CA: Sage.
variable modeling approach for measuring interaction effects: Results from a
Katz, E., Gurevitch, M., & Haas, H. (1973). On the use of the mass media for
Monte Carlo simulation study and an electronic-mail emotion/adoption study.
important things. American Sociological Review, 38(2), 164e181.
Information systems research, 14(2), 189e217.
Kaytoue, M., Silva, A., Cerf, L., Meira, W. J., & Raïssi, C. (2012). Watch me playing, I am
Chin, W. W., & Newstad, P. R. (1999). Structural equation modeling analysis with
a professional: A first study on video game live streaming. In WWW '12 com-
small samples using partial least squares. In R. Hoyle (Ed.), Statistical strategies
panion proceedings of the 21st international conference companion on World Wide
for small sample research (pp. 307e341). Thousand Oaks: Sage Publications.
Web (pp. 1181e1188). New York: ACM.
Cook, T. D., Campbell, D. T., & Day, A. (1979). Quasi-experimentation: Design &
Ko, H., Cho, C.-H., & Roberts, M. S. (2005). Internet uses and gratifications: A
analysis issues for field settings (Vol. 351). Boston: Houghton Mifflin.
structural equation model of interactive advertising. Journal of Advertising,
Courtois, C., Mechant, P., De Marez, L., & Verleye, G. (2009). Gratifications and
34(2), 57e70.
seeding behavior of online adolescents. Journal of Computer-Mediated Commu-
Krcmar, M., & Greene, K. (1999). Predicting exposure to and uses of television
nication, 15, 109e137.
violence. Journal of Communication, 49(3), 24e45.
Dovey, J., & Kennedy, H. W. (2006). Game cultures: Computer games as new media:
Lange, P. G. (2007). Publicly private and privately Public: Social networking on
Computer games as new media. McGraw-Hill Education.
YouTube. Journal of Computer-Mediated Communication, 13, 361e380.
Ewalt, D. M. (2014, 1 16). How big is Twitch's Audience? Huge. Retrieved 6 6, 2016,
LaRose, R., & Eastin, M. S. (2004). A social cognitive theory of Internet uses and
from Forbes: http://www.forbes.com/sites/davidewalt/2014/01/16/twitch-
gratifications: Toward a new model of media attendance. Journal of Broadcasting
streaming-video-audience-growth/.
& Electronic Media, 24(3), 358e377.
Fornell, C., & Larcker, D. F. (1981). Evaluating structural equation models with un-
Lee, D., & Schoenstedt, L. J. (2011). Comparison of eSports and traditional sports
observable variables and measurement error. Journal of Marketing Research,
consumption motives. The ICHPER-SD Journal of Research in Health, Physical
18(1), 39e50.
Education, Recreation, Sport & Dance, 6(2), 39.
Gelman, A. D., Halfin, S., & Willinger, W. (1991). On buffer requirements for store-
Lin, C. A. (2002). Perceived grati®cations of online media service use among po-
and-forward video on demand service circuits. In Global telecommunications
tential users. Telematics and Informatics, 19, 3e19.
conference (pp. 976e980). Phoenix: IEEE.
McLuhan, M. (1994). Understanding media: The extensions of man. MIT Press.
Grundberg, S., & Hansegard, J. (2014). YouTube's biggest draw plays games, earns $4
Needleman, S. E. (2015, 1 29). Twitch's viewers reach 100 million a month. Retrieved 6
million a year. Retrieved October 29, 2015, from The Wall Street Journal: http://
6, 2016, from The Wall Street Journal: http://blogs.wsj.com/digits/2015/01/29/
www.wsj.com/articles/youtube-star-plays-videogames-earns-4-million-a-year-
twitchs-viewers-reach-100-million-a-month/.
1402939896.
€ blom, M. (2017). What is eSports and why do people watch it? Newman, J. (2004). Videogames. New York: Routledge.
Hamari, J., & Sjo
Nunnally, J. C. (1978). Psychometric theory. New York: McGraw-Hill.
Internet Research, 27(2). http://dx.doi.org/10.1108/IntR-04-2016-0085.
Oestreicher-Singer, G., & Zalmanson, L. (2013). Content or community? a digital

€ blom, M., & Hamari, J., Why do people watch others play video games? An empirical study on the
Please cite this article in press as: Sjo
motivations of Twitch users, Computers in Human Behavior (2016), http://dx.doi.org/10.1016/j.chb.2016.10.019
12 €blom, J. Hamari / Computers in Human Behavior xxx (2016) 1e12
M. Sjo

business strategy for content providers in the social age. MIS Quarterly, 37(2), Smock, A. D., Ellison, N. B., Lampe, C., & Wohn, D. Y. (2011). Facebook as a toolkit: A
591e616. uses and gratification approach to unbundling feature use. Computers in Human
Pai, P., & Arnott, D. C. (2013). User adoption of social networking sites: Eliciting uses Behavior, 27(6), 2322e2329.
and gratifications through a meanseend approach. Computers in Human Twitch. (2015). 2014 retrospective. Retrieved 6 6, 2016, from Twitch: https://www.
Behavior, 29, 1039e1053. twitch.tv/year/2014.
Papacharissi, Z., & Mendelson, A. (2010). 12 toward a new (er) sociability: Uses, Venkatesh, V. (2000). Determinants of perceived ease of use: Integrating control,
gratifications and social capital on Facebook. In S. Papathanassopoulos (Ed.), intrinsic motivation, and emotion into the technology acceptance model. In-
Media perspectives for the 21st century (pp. 212e230). Routledge. formation systems research, 11(4), 342e365.
Papacharissi, Z., & Rubin, A. M. (2000). Predictors of internet use. Journal of Vock, M., van Dolen, W., & de Ruyter, K. (2013). Understanding willingness to pay
Broadcasting & Electronic Media, 44(2), 175e196. for social network sites. Journal of Service Research, 16(3), 311e325.
Podsakoff, P. M., MacKenzie, S. B., Lee, J.-Y., & Podsakoff, N. P. (2003). Common Wang, Q., Fink, E. L., & Cai, D. A. (2008). Loneliness, gender, and parasocial inter-
method biases in behavioral research: A critical review of the literature and action: A uses and gratifications approach. Communication Quarterly, 56(1),
recommended remedies. Journal of applied psychology, 88(5), 879e903. 87e109.
Quandt, T., Van Looy, J., Vogelsang, J., Elson, M., Ivory, J. D., Consalvo, M., et al. (2015). Wang, Z., & Tchernev, J. M. (2012). The ‘‘Myth’’ of media Multitasking: Reciprocal
Digital games research: A survey study on an emerging field and its prevalent dynamics of media multitasking, personal needs, and gratifications. Journal of
debates. Journal of Communication. http://dx.doi.org/10.1111/jcom.12182. Communication, 62(3), 493e513.
Rubin, A. M. (2002). The uses-and-gratifications perspective of media effects. In Weiss, T. (2011). Fulfilling the needs of eSports consumers: A uses and gratifications
J. Bryant, & D. Zillmann (Eds.), Media effects: Advances in theory and research. perspective. In Proceedings of the 24th bled eConference “eFuture: Creating so-
Mahwah, NJ: Lawrence Erlbaum Associates Publishers. lutions for the individual, organisations and society”. Bled, Slovenia.
Ruggiero, T. E. (2000). Uses and gratifications theory in the 21st century. Mass Weiss, T., & Schiele, S. (2013). Virtual worlds in competitive contexts: Analyzing
Communication & Society, 3(1), 3e37. eSports consumer needs. Electronic Markets, 23(4), 307e316.
Schmitt, K. L., Woolf, K. D., & Anderson, D. R. (2003). Viewing the Viewers: Viewing Westland, J. C. (2010). Lower bounds on sample size in structural equation
behaviors by children and adults during television programs and commercials. modeling. Electronic Commerce Research and Applications, 9(6), 476e487.
Journal of Communication, 53(2), 265e281. West, R. L., & Turner, L. H. (2010). Introducing communication Theory: Analysis and
Scholz, T. M. (2012). New broadcasting ways. In IPTV e The case of the starcraft application. Boston, MA: McGraw-Hill.
broadcasting scene. World media economics & management conference. Whiting, A., & Williams, D. (2013). Why people use social media: A uses and
Shaw, A. (2013). E-Sport spectator motivation. Fairfax: George Mason University. gratifications approach. Qualitative Market Research: An International Journal,
Sherry, J. L., Lucas, K., Greenberg, B. S., & Lachlan, K. (2006). Video game uses and 16(4), 362e369.
gratifications as predictors of use and game preference. In P. Vorderer, & Wu, J.-H., Wang, S.-C., & Tsai, H.-H. (2010). Falling in love with online games: The
J. Bryant (Eds.), Playing video games: Motives, responses, and consequences (pp. uses and gratifications perspective. Computers in Human Behavior, 26,
213e224). 1862e1871.

€ blom, M., & Hamari, J., Why do people watch others play video games? An empirical study on the
Please cite this article in press as: Sjo
motivations of Twitch users, Computers in Human Behavior (2016), http://dx.doi.org/10.1016/j.chb.2016.10.019