Matrices of binary or count data are modelled under a unified statistical framework using finite mixtures to group the rows and/or columns. These likelihood-based one-mode and two-mode fuzzy clusterings provide maximum likelihood estimation of parameters and the options of using likelihood ratio...