I've seen this before, but there's a minor surge of interest on the SOCNET listserv on the network analysis of legal citations, so I thought I would mention it here.
The approach dates back at least four decades (e.g., Nagel 1964), but became more sophisticated in the mid-80s (e.g., Caldeira 1985, 1988). Yet most efforts before 2000 were still manually coded, and therefore limited in scope and scale.
The transition to e-documentation and the rising popularity of social network analysis is beginning to revive the approach (e.g., Fowler etal, 2005; Smith 2005).
CITE-IT, an NSF-funded project I work on here at the University of Maryland, is another such effort. Using a range of semi-automated approaches, we have developed a database of every federal level decision on a single area of law (regulatory takings), dating back to 1978*. We are also able to identify all cites to precedent between these decisions, noting both the citing and cited courts. This has allowed us to generate maps and metrics of all citations, controlling for those to higher or lower courts, or to "lateral" courts in other jurisdictions.
We are currently developing automated text mining tools to pull additional information from each decision (e.g., plaintiffs, defendants, judges, lawyers), from which we plan to contruct an additional actor-level database.
* We are also in the process of expanding the database to include the full history of the concept - dating back to at least 1918.
ADDENDUM: Almost forgot - we also have an annotated and linked bibliography on citation analysis.