The masses of the three largest asteroids: (1) Ceres, (2) Pallas and (4) Ve
sta were determined from gravitational perturbations exerted on respectivel
y 25, 2, and 26 selected asteroids. These masses were calculated by means o
f the least-squares method as weighted means of the values obtained separat
ely from the perturbations on single asteroids. Special attention was paid
to the selection of the observations of the asteroids. For this purpose, a
criterion based on the requirement that the post-selection distribution of
the (O - C) residuals should be Gaussian was implemented. The derived masse
s are: (4.70 +/- 0.04) x 10(-10) M., (1.21 +/- 0.26) x 10(-10) M., and (1.3
6 +/- 0.05) x 10(-10) M. for (1) Ceres, (2) Pallas and (4) Vesta, respectiv
ely.
We also show how the fact that a statistically substantial number of pertur
bed asteroids is used in the determination of the mass of (1) Ceres and (4)
Vesta increases the reliability of their mass determination because effect
s like the flaws of the dynamical model and/or the observational biases can
cel out. In case of Ceres and Vesta, these effects have a very small influe
nce on the final result. The number of acceptable mass determinations of Pa
llas is much smaller, but can be increased in the future when the dynamical
model is improved. We indicate some promising encounters with Pallas.