We develop a method for computing the probability densities of spherical voids in distribution of galaxies on a large scale. This method is applied to Poissonian cluster models. We consider CDM biased and non-biased models; the appropriate approximation has been used in each case. We find that, in the latter case, the probability finding a void like that of Bootes within the CFA catalogue is very small. For the former case, we study the dependence of the probabilities of voids on the biasing factor. For the appropriate value of this variable, the results are in good agree with the observations.