Good questions. It turns out that Marxism as "science" has been very much on my mind lately as I am reading Michael Perelman's manuscript dealing with "punctuated equilibrium," an attempt to understand economics from the perspective of Stephen Jay Gould's evolutionist theories.
In one of his early chapters, Michael makes a point that is absolutely essential. He says that economists have tended to look at things like depressions as if they were natural disasters, like tornadoes. There is a science which can predict tornadoes. It is called meterology. But there is no science that can predict depressions.
The main reason for this, in my view, is that human agency can not be predicted as cloud movements or the temperature can. Mass psychology is unpredictable. If in the East Asian crisis there is no militancy and workers struggles to recapture the economic gains of recent years, then capital will continue to accumulate at workers' expense. If the Russian workers in 1917 had not taken matters into their own hands, then the capital accumulation cycle would have continued.
All that Marxism can do is highlight tendencies. CLR James, one of my favorite Marxists, anticipated the emergence of modern black nationalism. He did this by studying the class composition of northern black populations and examining the Garvey movement without prejudice. This is what Marxists should be involved with, not looking into a crystal ball trying to predict political or economic collapses.
Louis Proyect (http://www.panix.com/~lnp3/marxism.html)