I did this in my first class this fall semester--as I recall, Beth Chance suggested it--and it really was a big hit. The original fish were, I believe, the Pepperidge Farm cheddar cheese fish; the "marked" fish were some kind of pretzel-like fish. They need to be quite different in appearance from the original fish.
I recommend reading Roger Johnson, "How Many Fish are in the Pond?", _Teaching Statistics_, vol.18, no.1, Spring 1966 before doing this one. Johnson suggests several ways to capitalize on this activity, as well as supplying code for a Minitab "exec" macro for simulating the capture/recapture process. Furthermore, he points out that there is "substantial bias and variability" in the usual estimator. This was what I emphasized in my first encounter with my students this semester: an introduction to bias and variability; see Moore & McCabe, p.261 ff.
Extension 3 to the U.S. Census makes it clear that this is not just an idle game.
============================================== Bruce King Department of Mathematics and Computer Science Western Connecticut State University 181 White Street Danbury, CT 06810 (firstname.lastname@example.org)