The Virtual Fishtank seems to fit in perfectly with Frank Oppenheimer's vision of what a science museum should aim to teach. The fish tank has a particularly lofty goal-to expose to visitors the subtleties of the workings of real-world, decentralized, dynamic systems. This seems particularly in line with what Oppenheimer saw as the cumulative effect of all of the Exploratorium's exhibits, as the fish tank doesn't even pretend to realistically model the actual behavior of schools of fish. A glance at the whimsically designed "Sharky" and "Flash Gorton" shows a more mechanical than natural design, clearly implying that the fish tank is more "virtual" than "reality." What, then, is the point? The point is what Oppenheimer describes-a way of looking at the world, of finding order and simplicity among what seems to be extremely complex behavior.

Robert Semper says of exhibit design that "the development of the exhibit aesthetic is critical." (Semper) On first stepping into the fish tank, I was struck by the immersive environment that it generates. This was more true at the Computer Museum, where the exhibit occupied a larger, more enclosed space with only a single entrance. The combination of soothing aquatic music, the dimly lit space, and the whimsical graphics on and off the screen create an enveloping environment. I found this to be less true at the Museum of Science, where there are several entrances and noise from adjacent exhibits intrudes upon the space. Even so, the consistent whimsy between the signage and on-screen graphics lead to a cohesive aesthetic experience.

Because the exhibit is computer-based, it is natural to consider the user interface and how it relates to the use of the exhibit. The computer screens allowing the user to control schools' behavior and create their own fish are inviting and relatively simple, generally consisting only of sliders or toggle switches. However, it did not seem clear exactly what was being adjusted. Children using the stations that allowed them to create a fish would say things like "let's make him thinner!" to indicate their design. In these cases it appeared that they did not realize that they were actually altering the fish's behavior. The stations which modified schools' behavior were also not entirely clear, relying on imprecise visuals to indicate changing rules. Indeed, one might wonder if the user interface is too inviting-I saw a number of children sit down at one of these stations and change all the sliders which modified the school's behavior, keeping them occupied sometimes for several minutes-without ever looking up at the screens to see what effects their changes had on the behavior of the fish in the tank. There were, however, others who clearly understood the effect their changes had. One boy who was probably about ten years old, was quite excited to show me how he could modify the sliders and change what the fish would do. He was even able to identify long-term effects on the tank's ecosystem; for example, after making the sharks extremely hungry, he said confidently that soon "there will be no other fish left!"

Robert Semper was also clear that "the user of the exhibit, not the designer, should be in control of the learning activity." (Semper) The Virtual Fishtank gives visitors many opportunities to influence the activities in the tank, from subtle changes in the school's behavior to the clear ability to have the diver become unconscious from lack of air. The feeding station was a big hit for younger children, although I don't think any of them made the connection that turning the wheel was causing food to appear at the top of the tank. Once again, the connection is perhaps a bit too subtle: having the pipe from the feeding station appear to merge with the pipe on the screen which is spewing food falls well outside of the realm of simple observation. Even subtler was the fact that one of the panels allows visitors to wave at the fish and have the fish react. In each of these cases it seems that a simple sign (outside the computer screen) would have helped clarify the expected interaction.

The action of constructing a fish was clearly a big hit with visitors of a wide age range. Children were quick to identify with "their" fish, tagging it with their first initial and excitedly following its progress in the fishtank. However, there seemed to be several problems with immediacy in this portion of the exhibit. It was occasionally not clear to the children when their fish emerged from the pipe, and I often saw kids become confused as to where their fish had gone when it disappeared between panes of the fish tank or when it was eaten by another fish. This often led to a lessened interest in further activity, seemingly feeling cheated by the disappearance of their fish. This seemed to undermine the connection that they made with the exhibit. The problem with immediacy was also apparent as visitors used the stations that modified the schools' behavior. Because the behavior in the tank is sometimes slow-paced (compared to what is expected of the typical computer game) it was not obvious to some visitors that they were having an observable effect.

This also points to a problem in how the exhibit was being used. The focus of many visitors seemed to be using the stations to modify behavior or to create a fish, rather than simply observing the interactions in the tank itself. This seemed unfortunate, because during my 90-minute or so visit to the fish tank I observed a great deal about the fishes' behavior without ever touching any of the stations (save to recalibrate a couple of their touch-screens.) Many visitors, upon entering the exhibit, made a beeline for an available schooling station without taking the time to understand what was going on in the tank. The fact that the tank was difficult to see while at a station did not help this understanding. Once again, the raised, expansive setting from the Computer Museum seemed to be a better arrangement, allowing these users to easily see the activity in the tank.

I was disappointed to see that few people used the "Diving Deeper" stations, and when they did their interaction was usually about twenty seconds long, quickly being drawn back to the larger and flashier tank itself. The notable exception was the traffic jam station, which I saw a number of people use for a longer period of time. It seems that the difference in this station was partially the ease of use of the controls and largely the familiar subject matter. The other stations' subjects were less familiar, such as how Starlogo turtles would cluster themselves based on simple rules. The fish tank itself also clearly derives much of its power from the familiarity of the subject matter.

Indeed, the point of the Virtual Fishtank is not so much to look at something new but to look at familiar things in a new way. It does not do a great deal of what might be called "traditional" teaching but rather provides an environment in which visitors can see the effects of their modifications to a dynamic system. Though it is not perfect, it succeeds in some important aspects, such as creating an immersive, visitor-directed experience and providing easy accessibility to its visitors.