In this article, the authors present a matrix-fracture transfer function where the statistical variation in geometric properties of the matrix blocks is considered. Several particular representations with hypothetical probability density functions (PDFs) for matrix block size distributions are presented, including: (a) the single-value distribution (the limiting case); (b) the uniform distribution; (c) the Gamma distribution; and (d) an approximate representation for arbitrary PDFs. An example using experimental data from the literature, along with the single-block based transfer function developed in this study, is presented demonstrating how the statistical procedure proposed in this text can be applied in practice. It is shown with this example that significant relative errors can be introduced when the statistical variance is ignored. Furthermore, two existing dual-porosity models, the Lim and Aziz model and the Zimmerman et al. model, are also considered using the experimental data. It is shown that considerable relative errors can be introduced with these two models when the effect of statistical variance is not taken into account.