This paper addresses the question to what extent morphological blocking in language is a rule-based phenomenon. We argue that language users do not operate with a blocking rule, but that a form preference emerges as a result of cognitive selection mechanisms in a neural network of linguistic information. The actual target form develops its own token frequency in a probabilistic process, known as Preferential Attachment. After some time and some generations, one form will develop a nearly absolute dominance with its own local token frequency. This model implies that there is no blocking as an active negative action, but only a local lemma specific frequency, built up by a stochastic Preferential Attachment process, which favours one of the theoretically possible forms and, as a consequence, 'suppresses' the other options.