Jay Manifold cites the above here, noting that bubbles were implicated in the Chernobyl accident. As an engineer with a decade's worth of experience onsite at nuclear power plants during construction, startup and operation I must point out a big difference between Chernobyl and the BWRs and PWRs I mentioned above.
As Jay mentioned, the Chernobyl reactor had what we call a positive void coefficient. Here the voids are bubbles, and the idea was that the more bubbles there were, the more rapidly the reactivity, and thus the power, and thus the heat of the reactor would rise. Of course this would create even more voids. In fact this happened so rapidly that the Chernobyl unit 3 exploded from the violence of the sudden bubble generation.
In the US such a design could not be licensed for a commercial nuclear power plant precisely because of the above scenario. In fact, regulations insist on a negative void coefficient, such that the presence of bubbles suppresses the reaction. How do you do that?
The difference is fundamental in the design. A thermal nuclear reactor requires something called "moderation" - no moderation, no useful reaction. In a BWR or PWR (the reactors commonly found in the US, India, France, and Korea) most of this service is provided done by the liquid water in the reactor. In the stricken Chernobyl reactor, the water contributed to moderation, but most of the moderation was from graphite blocks in the reactor.
Notice I said liquid water. The bubbles of steam are water in the chemical sense, but in terms of moderation they behave very differently - steam moderates far less. So the presence of steam bubbles tends to calm the reactor down so to speak. But with the RBMK reactor design at Chernobyl, the influence of the steam bubbles was not sufficient.
I don't know the relevant figures for the RBMK design. But just to provide an example with numbers in it, suppose that the water provided 10% of the moderation, and production of a given amount of steam reduced that to 1%. Then turning the water to steam would drop the overall moderation figure to 91% of the original. If water provided 100% of the moderation, then the moderation would be 10% of the original. Again, this is not intended to represent that actual state of affairs, but just to show how there is a range of "tuning" at the design stage which can radically affect the reactor's behavior.
In licenseable US reactors this negative void coefficient behavior is built in. It is not possible for the operator to run the plant in such a way that this is not true. Thus such problems are prevented at the licensing stage in the US.
The short answer is that the Chernobyl reactors are fundamentally different in design from US plants such that they are subject to entirely new classes of problems, and the spectacular accident almost 17 years ago resulted from one of those problems. Thus its relevance to US plants is limited.