Just coming off teaching a 3 day social network analysis (SNA) master class at Boston College (with my fellow PhD student Inga Carboni and professors Steve Borgatti and Bill Torbert), I have networks on the mind. Just before the workshop started, I was finally able to diagram a very large network data set I have been working on for a long time (with the help of programmer pal Stephen Cook and fellow Borgatti student Rich DeJordy). The other night I finally got back to it.
Frequencies of 2o most versioned riddims at reggae-riddims.com.
Ever since reading Howard Becker’s (1982) book Art Worlds, I have been interested in using network methods to operationalize Becker’s claim that art worlds can be seen as networks of cooperative relations among individuals and orgnizations producing and consuming artistic products (Becker, 1982:x). While network thinking has been a central part of this theory from the beginning, SNA methods have only recently been applied to large data sets on cultural markets. For example, the now familiar Oracle of Kevin Bacon site uses the Internet Movie Data Base (IMDB) to create an interface that allows you to play the Gen X parlor game “Six Degrees of Kevin Bacon” (e.g., trying to find an actor that is connected to Kevin Bacon in more than 6 links via co-occurrences in movies with other actors). If you play, you will find that it is hard. That is because Hollywood is a Small World. However, even this site does not really let you visualize the network of ties among actors and movies. Scholarly work on cultural markets is even less likely to do so. This is a shame because SNA is so useful for visualizing patterns in large social systems that are otherwise hard to see.
Using Ucinet and Netdraw to diagram (below) the data at Reggae-riddims.com Becker’s (1982) idea comes to life in a new way. In this art world, it is common practice for multiple artists to sing over instrumental “versions” of popular songs (which are called riddims). One of the things that makes this art world particularly interesting from a network perspective is that the same riddims are often used over and over by different artists and producers and released by “competing” labels under different titles. As a result, we can examine cooperative patterns and see what artists appear on what riddims recorded by what producers, etc. It also helps me to find another setting that operates more like a creative commons than a traditional market (as this distinction is also emerging my dissertation). (Note the sneaky nod to the ethnographers in my title which alludes to Gertz’s 1978 piece “The Bazaar economy: Information and search in peasant marketing”, American Economic Review 68(2):28-32.)
The first thing I did was to generate a frequency list (above) of all the riddims (e.g., the number of times each had been done) and plot the results. Not surprisingly, a small number of riddims get versioned a lot and most get versioned only once or twice. This kind of picture is typical of cultural industries (e.g., a few composers do most of the scores in Hollywood, a few actors do most of the movies, etc.) and is related to the idea of long tails (for anyone who caught the Wired buzz a while back). To keep track of these ideas, check out Chris Anderson’s site the Long Tail and his upcoming book. Clearly, the art world of reggae versions is another example of this ubiquitous phenomena. For yet another (that relates to my love of formal field methods in anthropology), see the excellent word count flash site which lets you interact with a frequency list of english words.
Network map of 20 most versioned riddims (blue nodes=riddims, red nodes=artists)
I used this frequency list to select the 20 most versioned riddims for further analysis. (The data set was so large — e.g., more than 10,000 lines of data), that it became hard to run on my weak machines, let alone visualize. Therefore, for practical reasons, I decided to cut the set down to the top riddims for a first look. The diagram below depict ties between artists (red) and the top 20 most versioned riddims (in blue). It uses an algorithm (called Gower Metric Scaling) that basically puts riddims that were done by many of the same artists near each other on the map.
I have just begun to analyze this data, but already, interesting patterns are emerging. For example, you can see that there are basically four “neighborhoods” of artists. The first cluster (in the middle of the diagram) are all doing many of the same riddims. The second cluster (middle left) is a group of artists associated with the Real Rock riddim. (This is interesting in itself as the Real Rock is the most versioned riddim, so you would expect it to be in the middle of the network.) Another cluster appears in the bottom right around the Stalag riddim. Finally, there is a group of artists (bottom left) that did the Stalag and Real Rock, but not much else. This is already pretty interesting, as it demonstrates the power of network analysis to identify emergent communities in artistic markets (exactly what Becker would have predicted, btw).
While the diagrams tell us something, they are really just the beginning. You see, network analysis is really good at picking out hidden patterns in social systems. But to understand the meaning behind the pattern (e.g., the why), you still need to do good ethnographic work. As far as I can tell, that is still lacking in many quarters of the academy. Moreover, because of the increasing balkanization of the academy, network analysis and ethnography currently occupy separate spheres (despite sharing common intellecual roots).
Luckily, my pal Wayne is doing his dissertation on Jamaican music and I hope we can combine our structural and interpretive approaches to provide a new look at some important cultural practices at the intersection of music and the creative commons.