The study presents two experiments on the role of morphological transparency of the word ending of three types of Italian words characterised by an arbitrary gender: a. Transparent words where the suffix allows the derivation of the grammatical gender ("-a" for feminine and "-o" for masculine, e.g. "casa" [house], "tavolo" [table]); b. Opaque words that end in "-e" (e.g. "pace" [peace], feminine or "pane" [bread] masculine); c. Incongruent words where the word ending contrasts the morphological regularity (e.g. "mano" [hand] is feminine and "tema" [theme] is masculine). The words were either presented alone (Experiment 1) or preceded by a definite article agreeing in gender (Experiment 2). Subjects were asked to perform a gender decision task. Results showed that Incongruent words took longer than Opaque words which, in turn, were slower than Transparent words. Gender decision was faster when the words were preceded by a definite article but the effect of the gender marking incongruency still persisted.