This Revolution is for Display Purposes Only: graffito, London Bridge

Medium vs. Message in the `'Blogosphere'

[ Home page | Random stuff ]

Web logs are presently very popular. Some commentators believe that they represent a new type of discourse which will come to supplant conventional journalism. Others believe that they represent a fad fuelled by authors engaged in a meaningless popularity contest.

What do web loggers write about? A cursory examination of popular web logs might lead you to the impression that web loggers mainly discuss other web loggers, web logs, and web logging in general. Is this impression accurate?


I sacrificed an hour or two of Saturday afternoon (17th May 2003) to answering this question. I read a large number of web log posts, drawn from the following sources:

  1. 49 recently-updated web logs from the aggregator; only web logs which were primarily English-language and had identifiable links to archived stories (`permalinks') were used. This represents a random but self-selecting sample of web logs.
  2. The web logs of four of Andrew Orlowski's `A-list' of important web loggers:

In the first category, I inspected the most recent 5 posts from each web log listed. For the `A-list' web loggers I read slightly fewer than 50 of the most recent posts. These were classified into four categories, as follows:

Medium Message
Original content about web logs etc., written by the post's author content about anything other than web logs, written by the post's author
Linked mostly links to or quotes from others' writings about web logs etc. mostly links to or quotes from others' writings about anything other than web logs

Obviously there is an element of subjectivity here, and a post which falls into more than one category cannot be classified unambiguously. In ambiguous cases I have preferred to record posts as original not linked (for instance where a post consists of a mix of quotes and commentary), and as message rather than medium (for instance where a post talks about both web logging and another topic). Obviously both problems arise partly from counting posts rather than words, column-inches or some other more continuous measure, but doing otherwise would be intractable for a brief exercise. In any case you can inspect the raw data and re-classify it yourself if you wish.


Tables show number of posts; percentages are of the total number of posts.

Recently updated web logs from

Bar chart of data in table below
Medium Message
Original 12 4.82% 105 42.2%
Linked 25 10.0% 107 43.0%

Ben Hammersley

Bar chart of data in table below
Medium Message
Original 13 28.3% 10 21.7%
Linked 14 30.4% 9 19.6%

Dave Winer

Bar chart of data in table below
Medium Message
Original 6 13.0% 5 10.9%
Linked 20 43.5% 15 32.6%

Doc Searls

Bar chart of data in table below
Medium Message
Original 6 13.0% 8 17.4%
Linked 17 37.0% 15 32.6%

Tim O'Reilly

Bar chart of data in table below
Medium Message
Original 2 4.35% 12 26.1%
Linked 4 8.70% 28 60.9%


A random sample of web loggers mainly discuss message (85.5%) rather than medium. Their posts are split approximately evenly between original content and regurgitating the work of others. `A-list' web loggers are more likely to regurgitate the work of others, and to discuss web logging and associated subjects rather than topics of more general interest.

By comparison, the traditional news media stick mainly to message and their content is largely original (although many publications rely in part on wire service reports). For instance, the BBC News site typically has a tiny number of medium stories (in this context, stories about news, news gathering, journalism and related topics) and a preponderance of message stories about actual news events.

Much web logging by `A-list' web loggers appears to be addressed exclusively to other web loggers, since non-web-loggers are unlikely to be interested in posts about the medium. This problem does not affect the web logs in the random sample.

Copyright (c) 2003 Chris Lightfoot. All rights reserved.