Jordi Cabot reports the results of a very unscientific competition to find the best modeling notation for requirements engineering:
To make a long story short, natural language and rich images won the contest. Given the informal and in fact comedy-like nature of the panel, it makes no sense to try to conclude anything. However, it is true that for the two semi-formal notations that passed the first round, adapting on the fly the model to the new requirements was tough due to the difficulty of presenting the resulting model in an attractive form for the audience and the need to use a flip-chart in the room instead of some supporting modeling tool.
I've said before and will say again: I believe the onus is on the advocates of semi-formal modeling (e.g., UML) to demonstrate that their tools make programmers more productive, not on the rest of us to argue that they don't.