Since a couple of years the asteroid families are under the light of a
renewed interest. This is due to the simultaneous improvement of the
techniques of proper elements computation and of the methods of classi
fication. Recently, two different cluster analysis methods (Zappala et
al. 1990, Bendjoya et al. 1991) have provided two very similar lists
of asteroid families. This has brought a renewed confidence on the rel
iability of asteroid families classification. Since a new set of impro
ved proper elements is now available (Milani & Knezevic 1992), the ide
a of the present work is to test the two identification methods by mea
ns of artificially generated families in order to gain an insight into
their limitations and to possibly develop improved strategies for fut
ure analyses. Different cases have been envisaged in order to study th
e effectiveness of the methods as a function of increasing background
densities and of the spread of the families. The ability of the method
s to detect peculiar features such as filaments or to separate two clo
se families has also been tested. The number of interlopers has been d
etermined in each case in order to evaluate their relative amount. Whi
le some improvements to the adopted procedures are suggested by the nu
merical tests, the results presented in the present paper are encourag
ing and allow us to be confident about the reliability of our previous
family lists, as well as about the expected performances of the metho
ds in future applications.