When using Slony, Londiste, or similar statement-based replication software, or even streaming replication plus Hot Standby, each of your slaves is going to lag a bit behind the master. If your application can support running queries against that slightly out-of-date information, it's straightforward to put a load balancing program such as pgpool-II in front of the master and a set of slave nodes in order to scale up reads.
This replication approach is also very good for maintaining autonomous remote servers. You might put servers in multiple cities and the application reads against the closest copy of the data, as a way to decrease the load on any one server.
More information about how pgpool-II ...
