This week there has been quite a buzz about a study which claims to have validated Dunbar’s Number (see footnote) in terms of Twitter, for example ‘Human Brain Limits Twitter Friends to 150’ and ‘Dunbar’s Number proves that you can’t realistically follow more than 150 friends on Twitter’. The actual paper ‘Validation of Dunbar’s number in Twitter conversations’ is available on Scribd.
However, this buzz misses the point, and I think the paper also contains at least one basic weakness. The sorts of relationships that Dunbar was talking about are really quite different to Twitter followers, LinkedIn contacts, and Facebook friends. I do not doubt the basic idea that as we are connected with more and more people in social media we exhibit a hierarchy in how much attention we pay to each contact, but I am not convinced that Dunbar or the 150 is terribly relevant to most of the aspects of social media we are interested in.
Robin Dunbar was not talking about people we have regular communications with he was talking about stable social networks. A postman, a newsagent, a bus conductor have two-way networks well in excess of 150, but these are not stable social networks because the people being interacted with are not necessary in a network with each other. When Dunbar talks about stable social relationships he is talking about groups we are a member of, not the number of people we are interested in or interacting with. If I were to live in an isolated tribe of 150 people I need to do more than remember the 150 people, I need to know who is related to who, who likes who, who is employed by who etc. Dunbar’s hypothesis is that the cognitive load required to hold and manipulate all this information limits the size of a tight-knit, stable community to 150 – he is quite happy to accept we have wider, less complete connections.
In social media, people are developing new types of relationship with much lower levels of engagement, a level of communication that falls well short of the sort of interaction that Dunbar was talking about, and therefore does not have the same cognitive load. For example, there are people in my Facebook network that I would not dream of contacting in any conventional way, for example distant family members and some of the people I was at school with, but I am very happy to hear their latest news, for example finding out who has married, who has died, and about key events such as babies and graduations. This sort of relationship is clearly very different to what Dunbar was talking about.
My feeling is that what is important for the continued development of social media is how to blend differing levels of contact and engagement. Amongst the things that make Facebook so powerful is the way that news flows can be controlled, for example most recent versus top news, to facilitate different tiers of closeness to develop. In Twitter it has been the development of tools such as TweetDeck and the use of hashtags that enables people to draw their engagements from wider networks. The continued growth of social media will be partly driven by improved tools and techniques for staying in touch with casual contacts.
One change that has happened is that the growth of Facebook has somewhat changed the way the word ‘friend’ is used. Before social media the word ‘friend’ was more typically used for somebody who people made a real effort to stay in touch with, but post-Facebook the term has been stretched to cover much more ephemeral contacts. When somebody says the number of friends you can have is X, they need to be clear what they mean by friend.
Implications for marketing and market research
The most popular theory of how social media is connected is that it is an example of a scale-free network. Most people have a relatively small number of contacts and a few people have a very large number of connections. Messages (memes, jokes, ideas etc) flow through the nodes with the largest number of connections, especially when the connections link different social groups. One result of this structure is that these vital people, the nodes, tend to feel spammed whenever a brand tries to reach people via social networks, because the messages come through the nodes a very, very large number of times.
One issue for market research is that the best way to sample a scale-free network is not to sample it randomly, because different people have different impacts on the network. This is why disease centres often find they need to concentrate on the nodes in an infection, rather than treating people at random.
Note about Dunbar’s Number
Dunbar’s number was devised by Robin Dunbar, a British anthropologist and evolutionary psychologist. Dunbar suggested, on the basis of some theory, some primate studies, and some real world data, that the size of a stable social group is limited by the brain’s ability to maintain relationships. Although Dunbar indicates that the number is really quite variable, it is often taken as being 150. Where the number seems to be highly relevant is that 150 is the average upper limit for a cohesive group, such as a tribe or a village that does not have modern communications. However, Dunbar does point out that there are layers within and outside this 150 limit, i.e. we have closer contacts, which are fewer, and we have larger ones, which are more casual. One of the key features of the number in Dunbar’s study is that it relates to people interacting as a group, not to how many people an individual interacts with, so people in a stable social relationship with you, in Dunbar’s terms, should have some relationship with each other. This Guardian interview with Robin Dunbar gives some more background, in both text and as a video.
Dunbar is not suggesting that we can’t have networks of more than 150, he is suggesting that the limit to the number of people we can be heavily engaged with is limited by the capacity of our brain. Dunbar suggests that we can only have N meaningful relationships, but the definition of meaningful is highly subjective, and care needs to be taken to avoid defining meaningful in a circular way that comes back to 150.
It should also be noted that other researchers have produced quite different estimates of the maximum size of group (the Bernard-Killworth median is 231 for example), and the studies appear to suggest a wide range of values, not a magic number.
I am quite happy to believe that each person has a limited number of people who they are heavily engaged with. I am not convinced that it is necessarily 150. I think that the Dunbar number is just one suggestion, perhaps based on too little data, there are other suggestions, and it seems the number varies widely. I am also happy to believe that one consequence of these limitations is that functioning organisations have optimal sizes, for example an optimal size for a class, a school, a neighbourhood, a work team, a conference, but I am sure that the optimal sizes are different.
But, in terms of social media, as a pastime, as a source of information, as a marketing tool, and as a place to conduct market research I think the Dunbar number is largely an irrelevance.
The weakness in the Twitter study
I have a couple of minor worries about the study looking at the validation of Dunbar’s number via Twitter. The first point is that I am not convinced that most people are using Twitter as a social network, the study was based on people sending and receiving messages to each other, using @. People do this, but it is not the main use of Twitter. Much of Twitter is about asynchronous relationships and news processing, or the following of ‘stars’ such as Ashton Kucher. My second minor quibble is the way that the researchers seem to be using share of comment, rather than absolute values.
However, the big weakness is that the researchers claim to have found that 100-150 is the number of Twitter contacts that people can maintain as a reciprocal relationship and they then link that to Dunbar, claiming to validate Dunbar. But the implication of the paper is that if somebody had 150 Dunbar friends on Twitter then they ought not to have any friends who are not on Twitter. This is clearly nonsense, many of the people who we communicate with on Twitter are not inside our stable social network, they are simply contacts. It would have been possible for the Twitter study to show people had networks of 10 contacts or 2000 contacts without it either challenging or validating Dunbar, because Dunbar was not talking about those sorts of low-level communications – indeed in the interview mentioned above Dunbar appears to suggest that one requirement for a meaningful contact (the thing he thinks is limited to 150) is face-to-face contact. If the researchers had removed all the Twitter links that were not based on face-to-face contact I suspect they would have arrived at a much smaller number than 100-150, the number they suggest from their research.