This study challenges a computational implementation of Word and Paradigm Morphology with the task of modeling the semi-productive noun system of Maltese, which combines a dozen concatenative plural patterns with eleven non-concatenative plural patterns. We show that our model, trained on 6,511 word forms, generates accurate predictions about what meanings listeners understand and what forms speakers produce. Furthermore, measures derived from the model are predictive for Maltese reaction times. Although mathematically very simple, the linear mappings between form and meaning posited by our model are powerful enough to capture the complexity and productivity of the Maltese noun system.