I think there may be some confusion by some on the definition of "bounce rate". At one time I think it WAS naively defined as going into a web site, viewing only one page and exiting (perhaps with some time variable functioned in. As I understand it, in most cases Goog can't reliably tell how many pages on your site a person views (unless you used Goog analytics for your site! or they have google toolbar). Going to another page on your site, could be just as bad as leaving, since it may mean they did not find what they were looking for and simply gave you another chance. This is too easy to manipulate by site owners by doing page refreshes or leading people on with enticements.
Therefore they can only evaluate what they CAN see and that is: whether a person RETURNS to G and SELECTS _+ANOTHER+_ SEARCH RESULT USING THE SAME SEARCH TERMS!
In response to the apparent concern by some in this thread, at least _I_ firmly believe... because it makes the most logical sense and any other treatment by G would be simply silly, if they get to your site searching on "green widgets", find the info they wanted, then backed out to g and search on "red doodads" before clicking on another search result... then it should not count as a bounce. Whereas, if they DO click on another result without changing the search terms, it should count as a bounce.
This of course, has many exceptions, as someone previously mentioned where, for instance, someone is of the type (like me) that likes to comparison shop before finally returning and buying.
My suggestion is to search on "search engine" click on G, then go back and click on the next SE. Easy to do to your competitors too, then click on your own link, wait, and go back and search for another keyword ad infinatum ;-)