Emotion homophily in social network site messages
First Monday

Emotion homophily in social network site messages by Mike Thelwall

Social network sites (SNS) like MySpace seem to play a role in friendships and wider relationships for many people. Emotion expression can be important in relationship maintenance but little is known about the role of emotion in SNSs, other than positive comments being widespread in MySpace. Is emotion typically reciprocated, and do Friends express and/or receive similar levels of emotion expression to each other? Based upon an analysis of over two million MySpace public comments associated with 2,990 pairs of U.K. and U.S. Friends using a sentiment detection program, statistically significant evidence was found for a weak correlation between the strength of positive emotion exchanged between Friends and received by Friends. This is consistent with two separate hypotheses: members tend to Friend others with similar levels of public emotion expression, or the expression of emotion in MySpace is contagious. The results may help to identify non–optimal behaviour and at–risk individuals in SNSs.


Research questions




Emotion is now recognised as an important aspect of many areas of our lives. Aside from the obvious relevance of feelings like happiness and sadness to personal well–being, appropriate perception and communication of emotion is important for maintaining human relationships and friendships and not just intimate relationships. Particularly for women, friends are also a source of emotional support, one that can help individuals to cope in difficult times (Stoppard and Gunn Gruchy, 1993). Recently, however, an increasing amount of communication between friends seems to occur online via e–mail, chatrooms and social network sites. Given the importance of emotion in many friendships and fears that online communication is somehow poorer than face–to–face interaction (e.g., BBC, 2009), it is important to understand as much as possible about the properties of online emotion expression and support in friendship, so that appropriate actions can be taken if the important social institution of friendship is under threat or, conversely, if new opportunities are evident.

Many researchers have already investigated the role of emotion in online communication, particularly from the perspective of social psychology and for older communications media like e–mail and chatrooms (e.g., Derks, et al., 2008; Radić–Bojanić, 2006). These studies have typically used interviews or questionnaires to discover perceived communication patterns and emotion factors, such as personality types (Fullwood and Martino, 2007). It seems that no previous published study has investigated the role of emotion in SNS Friendship communication, although some have investigated aspects of emotion in SNSs (Banczyk, et al., 2008). Moreover, no previous study has used a large–scale analysis of communications between friends to investigate the role of emotion, rather than self–reports in interviews or questionnaires. This study partly fills these gaps through an investigation of the strength and polarity of sentiment expressed in messages associated with pairs of MySpace Friends. In particular, it assesses the extent of emotion homophily in public MySpace comments. Whilst traditional studies of homophily in social networks are frequently motivated by identifying the extent to which social, ethnic or educational background influences the choice or strength of relationships (McPherson, et al., 2001) and some studies have studied online homophily (Nowak and Rauh, 2006; Thelwall, 2009; Yuan and Gay, 2006), this is apparently the first large–scale study of emotion homophily and the first study of emotion homophily online.




SNSs are Web sites that allow members to maintain a profile page containing a photograph and personal information and to connect with other members as “Friends”, with lists of Friends being publicly displayed on profile pages (boyd and Ellison, 2007). Friends can normally communicate in several ways, including with private and public messages. Of particular interest here are the more public messages, such as MySpace comments and Facebook wall postings. Although online environments for communication sometimes seem to be communities that are isolated and to some extent insulated from the off–line world, particularly in the case of newsgroups and chatrooms, this does not seem to be the case for SNSs like MySpace and Facebook. These seem to embed within the lives of members (e.g., Wellman, et al., 2003), with Friends typically being also off–line friends or acquaintances (or musicians and celebrities) (boyd, 2006). For example, U.S. teens have used MySpace to socialise with friends after school (boyd, 2008).

It seems that there is much communication between friends in SNSs, although typical pairs of Friends exchange few public messages (Thelwall and Wilkinson, 2010). It seems likely that this is changing the friendship circles of members, for example by allowing them to maintain contact at a distance (e.g., whilst at college) and possibly also increasing the number of friends and acquaintances that are in at least occasional contact (Joinson, 2008). An important unknown, however, is the extent to which the nature of friendship, particularly close friendship, is changed by SNSs. For example, is it a substitute for off–line contact or a supplement? Does it change the nature of discussions between friends, for example by being differently emotional or supportive, or by being informed by facts revealed on profile pages?

Emotion is important in the daily management of friendships and romantic relationships (Weiss, 1998), for instance when showing empathy or giving social support (Hoffman, 2008). There are gender differences in the tendency to show and detect emotion in social situations, with women tending to be more sensitive to emotion and to display more positive emotion and perhaps also more negative emotion of some types (Shields, 2002). This seems to be related to gender differences in expectations of social roles (Alexander and Wood, 2000).

There are many socially recognised types of emotion, such as happiness, fear, sadness and anger, but one common way of analysing emotion is in terms of the two dimensions of valence and arousal (Cornelius, 1996). Valence is the type of emotion felt: the extent to which it is positive or negative. Arousal is the general perceived level of activation or energy and it is often associated with the strength of an emotion. Hence it is reasonable to simplify a discussion of emotions in friendship to just valence/polarity and strength. Note, however, that it is possible to simultaneously experience or express positive and negative emotions (Fox, 2008): to have mixed feelings about an event.



Research questions

The objective of this research is to assess different kinds of emotion homophily between pairs of Friends in MySpace using an automatic analysis of emotion in public comments, the main available public source of text messages between Friends in MySpace or any other SNS. For practical reasons, described below, it is not possible to address the objective by analysing all public comments written by individual members but it is possible to analyse all comments written to certain members. As a result the following research questions address different aspects of emotion:

  • Do public comments exchanged between pairs of Friends contain similar strengths of positive and negative emotion?

  • For a given pair of Friends, do the public comments sent to each of them by other Friends contain similar strengths of positive or negative emotion?

  • For a given pair of Friends, when they write comments on others’ profiles, do they contain similar strengths of positive and negative emotion?




In order to test the emotion homophily hypothesis, a large sample of MySpace members was generated and the emotional content of their messages was compared with that of their friends. The following summarises the procedure.

A sample of MySpace users was generated by systematically selecting member IDs from those allocated on 3 July 2006. This gives a representative sample of all those who joined before this date. From this sample, the following were excluded: musicians, comedians, filmmakers, ex–members, people with private profiles and people declaring residence outside of the U.K. or U.S. All excluded members either did not reveal the necessary data, were commercial rather than private MySpace users, or were difficult to study due to their country being a relatively low MySpace user. For comparison, a similar sample was taken of members joining on 18 June 2007 and another sample was taken of all members that had joined on any day up to 3 July 2007. Multiple samples were used to assess how consistent the results were over time.

For each selected member, all public comments made to them were extracted from their profiles (and any additional comment pages) and all comments made (by anyone) to one randomly selected commenting friend with a profile matching the above restrictions were also recorded. This rather strange data set, illustrated in Figure 1, was chosen as this is the natural way to collect comment data. In particular, it is not possible to directly collect all comments made by a person, except by filtering them out of the set of comments made to each of their friends. Since some users have thousands of MySpace Friends it is not practical to collect a large data set of comments made by individual members.


Figure 1: Data based upon pairs of Friends A and B that exchange comments
Figure 1: Data based upon pairs of Friends A and B that exchange comments, including all comments made to A or to B but not all comments from A or from B.


An emotion categorisation program (Thelwall, et al., submitted) was used to assign a positive and a negative emotion to each comment on a scale of one (no emotion) to five (very strong emotion). The classification was based upon a predefined list of about 500 emotion–bearing words (e.g., hate = negative 4), emoticons and emphatic devices including words (e.g., very), punctuation (e.g., !!!!) and repeated letter emphatic spelling (e.g., soooo).

Homophily was identified directly and indirectly from the data described above as follows:

Direct method: For each pair of Friends A and B, the average comment emotion strength A → B was compared with the average comment emotion strength B→ A (see Figure 2). The direct method may give biased assessments of homophily if emotion in a dialog is reproduced between participants and so a second method was also used.


Figure 2: Comments used for the direct method
Figure 2: Comments used for the direct method.


Indirect method: For each pair of Friends A and B, the average emotion strength of comments to A is calculated, except for comments from B. This is compared to the average strength of comments to B except for those from A (see Figure 3). Assuming that emotion is copied between comments in a dialog (i.e., precisely the assumption that makes the direct test invalid) then the average emotion of comments to A is a good estimate of the average emotion of comments from A. Hence comparing the results from A and B approximates a direct test of homophily in emotion between A and B in a way that is not biased by copying in direct communication between A and B.


Figure 3: Comments used for the indirect method
Figure 3: Comments used for the indirect method.


Pearson correlations were used to assess the extent of homophily. A positive correlation suggests that a higher level of emotion in one member is associated with a higher level of the same type of emotion associated with their partner.




As shown in Table 1, there was a correlation between the emotion of comments associated with MySpace members and the emotion of comments associated with their friends for both positive and negative emotion in all cases except one (the smallest sample size), and whether measured directly or indirectly. For the direct method, this means that the emotion of comments sent by a member to a friend tends to be similar to the emotion of comments returned by that Friend. For the indirect method, this means that the emotion of comments sent to a member tends to be similar to the emotion of comments sent to a friend. Combining the two results, it can be deduced that the emotion of comments sent by a member tends to be similar to the emotion of comments sent by a Friend, whether or not the comments between the two Friends in question are excluded. Hence there is homophily in the emotion of comments of MySpace members and their Friends. Nevertheless, the degree of homophily is relatively weak, ranging from an average correlation of 0.227 to an average correlation of 0.308, showing that the homophily is not a strong tendency in the sense that it would be unreasonable to expect the level of emotion of MySpace Friends to closely follow each other.


Table 1: Basic statistics for the data sets used and correlations between average positive emotion and average negative emotion associated with pairs of MySpace Friends.
Note: Bold=Significant at five percent level, Bold+italic= one percent level, Bold+italic+underlined= 0.1 percent level.
NameInfoPairs of friends used in analysisTotal comments between friendsTotal comments not between friendsDirect method correlation (+ve, -ve)Indirect method correlation (+ve, -ve)
U.K. July 2007U.K. members joining 3 July 20061941,985115,5550.2710.2160.1720.199
U.K. June 2008U.K. members joining 18 June 20076639028,2100.262-0.0330.1370.560
U.K. March 2008U.K. members joining before 3 July 200710184641,8400.1340.2960.0460.136
U.S. July 2007U.S. members joining 3 July 20061,20611,939829,2600.1940.2360.2870.346
U.S. June 2008U.S. members joining 18 June 20072702,458183,0660.2400.3420.2870.319
U.S. March 2008U.S. members joining before 3 July 20071,15313,065909,1550.2570.2720.2210.285





The results show a weak but statistically significant level of homophily for both positive and negative sentiment directly for the first two research questions. This indirectly gives the same answer for the third research question.

A limitation of the research is that it is based upon an automatic analysis of the way in which emotion is expressed. In consequence, the similarity may be partly due to language use. For example, if Friends shared oblique ways of referencing emotion then this would not be captured well by the approach used. Nevertheless, the prevalence of detected emotion suggests that this is not a major issue.

The method does not test whether people Friend others with similar emotional dispositions, personalities or expressive styles. Alternatively, it does not test whether emotion expression strengths are copied within MySpace but one of these two options seems likely to be the explanation for the results.

In terms of applications of the findings, Friendship emotional homophily suggests that posting positive messages to Friends may be a successful strategy to encourage positive emotion. Similarly, SNS members should aim to be sensitive to the emotion conveyed by their Friends and try to reciprocate to some extent: not to do so may make them seem abnormal and hence at risk of losing Friendships. End of article


About the author

Mike Thelwall is Professor of Information Science and leader of the Statistical Cybermetrics Research Group at the University of Wolverhampton, U.K. He is also visiting fellow of the Amsterdam Virtual Knowledge Studio, a Docent at Åbo Akademi University Department of Information Studies, and a research associate at the Oxford Internet Institute. He has developed tools for downloading and analysing Web sites, blogs and social networking sites, including the research Web crawler SocSciBot and software for statistical and topological analyses of site structures (through links) and site content (through text). He has published 152 refereed journal articles, seven book chapters, and two books including Introduction to Webometrics, is an associate editor of the Journal of the American Society for Information Science and Technology and sits on eight other editorial boards.
Web: http://www.scit.wlv.ac.uk/~cm1993/
E–mail: m [dot] thelwall [at] wlv [dot] ac [dot] uk



The work was supported by a European Union grant by the 7th Framework Programme, Theme 3: Science of complex systems for socially intelligent ICT. It is part of the CyberEmotions project (contract 231323).



Michelle G. Alexander and Wendy Wood, 2000. “Women, men and positive emotions: A social role interpretation,” In: Agneta H. Fischer (editor). Gender and emotion: Social psychological perspectives. Cambridge: Cambridge University Press, pp. 189–210.

Barbara Banczyk, Nicole Krämer, and Maria Senokozlieva, 2008. “The ‘wurst’ meets ‘fatless’ in Myspace: The relationship between self–esteem, personality and self–presentation in an online community,” paper presented at the Congress of the International Communication Association (Montreal, Canada).

BBC, 2009. “Facebook criticised by archbishop,” BBC News, number 2 (August), at http://news.bbc.co.uk/1/hi/uk/8180115.stm, accessed 1 September 2009.

danah boyd, 2008. “Taken out of context: American teen sociality in networked publics,” Ph.D. dissertation, University of California, Berkeley, at http://www.danah.org/papers/TakenOutOfContext.pdf, accessed 15 April 2010.

danah boyd, 2006. “Friends, Friendsters, and Top 8: Writing community into being on social network sites,” First Monday, volume 11, number 2, at http://firstmonday.org/htbin/cgiwrap/bin/ojs/index.php/fm/article/view/1418/336, accessed 29 April 2009.

danah boyd and Nicole Ellison, 2007. “Social network sites: Definition, history, and scholarship,” Journal of Computer–Mediated Communication, volume 13, number 1, at http://jcmc.indiana.edu/vol13/issue1/boyd.ellison.html, accessed 7 May 2009.

Randolph R. Cornelius, 1996. The science of emotion: Research and tradition in the psychology of emotions. Upper Saddle River, N.J.: Prentice Hall.

Daantje Derks, Agneta H. Fischer, and Arjan E.R. Bos, 2008. “The role of emotion in computer–mediated communication: A review,” Computers in Human Behavior, volume 24, number 3, pp. 766–785.http://dx.doi.org/10.1016/j.chb.2007.04.004

Elaine Fox, 2008. Emotion science: Cognitive and neuroscientific approaches to understanding human emotions. Basingstoke: Palgrave Macmillan.

Chris Fullwood and Orsolina I. Martino, 2007. “Emoticons and impression formation,” Applied Semiotics, volume 19, number 7, pp. 4–14.

Martin Hoffman, 2008. “Empathy and prosocial behaviour,” In: Michael Lewis, Jeannette M. Haviland–Jones, and Lisa Feldman Barrett (editors). Handbook of emotions. Third edition. New York: Guildford Press, pp. 440–455.

Adam N. Joinson, 2008. “Looking at, looking up or keeping up with people? Motives and use of Facebook,” Proceedings of the Twenty–sixth Annual SIGCHI Conference on Human Factors in Computing Systems, pp. 1,027–1,036.

Miller McPherson, Lynn Smith–Lovin, and James M Cook, 2001. “Birds of a feather: Homophily in social networks,” Annual Review of Sociology, volume 27, pp. 415–444.http://dx.doi.org/10.1146/annurev.soc.27.1.415

Kristine L. Nowak and Christian Rauh, 2006. “The Influence of the avatar on online perceptions of anthropomorphism, androgyny, credibility, homophily, and attraction,” Journal of Computer–Mediated Communication, volume 11, number 1, pp. 153–178, and at http://jcmc.indiana.edu/vol11/issue1/nowak.html, accessed 15 April 2010.

Biljana Radić–Bojanić, 2006. “Fragmentation/integration and involvement/detachment in chatroom discourse,,” Skase Journal of Theoretical Linguistics, volume 3, number 1, at http://www.skase.sk/Volumes/JTL05/4.pdf, accessed 3 March 2008.

Stephanie A. Shields, 2002. Speaking from the heart: Gender and the social meaning of emotion. Cambridge: Cambridge University Press.

Janet M. Stoppard and Carla D. Gunn Gruchy, 1993. “Gender, context, and expression of positive emotion,” Personality and Social Psychology Bulletin, volume 19, number 2, pp. 143–150.http://dx.doi.org/10.1177/0146167293192002

Mike Thelwall, 2009. “Homophily in MySpace,” Journal of the American Society for Information Science and Technology, volume 60, number 2, pp. 219–231.http://dx.doi.org/10.1002/asi.20978

Mike Thelwall, Kevan Buckley, Georgios Paltoglou, Di Cai, and Arvid Kappas, submitted. “Sentiment strength detection in short informal text.”

Mike Thelwall and David Wilkinson, 2010. “Public dialogs in social network sites: What is their purpose?” Journal of the American Society for Information Science and Technology, volume 61, number 2, pp. 392-404.

Robert S. Weiss, 1998. “A taxonomy of relationships,” Journal of Social and Personal Relationships, volume 15, number 5, pp. 671–683.http://dx.doi.org/10.1177/0265407598155006

Barry Wellman, Annabel J. Quan–Hasse, Jeffrey Boase, Wenhong Chen, Keith Hampton, Isabel Isla de Diaz, and Kakuko Miyata, 2003. “The social affordances of the Internet for networked individualism,” Journal of Computer–Mediated Communication, volume 8, number 3, at http://jcmc.indiana.edu/vol8/issue3/wellman.html, accessed 15 April 2009.

Connie Y. Yuan and Geri Gay, 2006. “Homophily of network ties and bonding and bridging social capital in computer–mediated distributed teams,” Journal of Computer–Mediated Communication, volume 11, number 4, pp. 1,062–1,084, and at http://jcmc.indiana.edu/vol11/issue4/yuan.html, accessed 15 April 2010.


Editorial history

Paper received 20 March 2010; accepted 5 April 2010.

Creative Commons License
This work is in the Public Domain.

Emotion homophily in social network site messages
by Mike Thelwall.
First Monday, Volume 15, Number 4 - 5 April 2010

A Great Cities Initiative of the University of Illinois at Chicago University Library.

© First Monday, 1995-2020. ISSN 1396-0466.