As stated by previous studies, the column test seems to be relevant for the evaluation of the state of segregation of self compacting concrete (SCC) at rest. However, the available information shows a great diversity of apparatus, procedures, aggregate sizes analyzed and segregation indexes. Based on repeatability procedures and statistical analysis of SCC compositions of the French National Project on SCC (PN B@P), this study is aimed particularly at defining what objective sieve size must be chosen to discriminate a real segregation state. When the values of the various indexes proposed in the literature are calculated as a function of sieve size from 2.5 mm, the results show significant variations in the indexes, which are not always explained by a significant segregation in the coarsest aggregate sizes. When segregation does not exist, these variations may be due to a random distribution of the coarsest aggregates. An optimal sieve size Dopt ranging from 0.5.Dmax to 0.625.Dmax is suggested (Dmax, maximum gravel size in the grading curve of concrete) which should ensure correct segregation analysis. In addition trends, but no strong relationships, are noted between the index values calculated from Dopt and the associated criteria.