Finding the optimal parent number for synthetic varieties has a crucial importance in forage breeding. The objective of this work was exploring this number for Mediterranean tall fescue selected for forage yield. The general (GCA) and specific (SCA) combining ability of parents, and their effects on the performance and the inbreeding depression of synthetics, were also assessed. The full-sib families from the diallel cross of 20 genotypes chosen from well performing populations were evaluated for fresh biomass over 13 harvests in Sanluri (Sardinia, Italy). The Syn 1 and Syn 2 of 15 synthetics varying in parent number (4, 8, 12, 16 or 20) and, within same number, in mean GCA of parents were evaluated for dry matter yield over 11 harvests in Lodi (northern Italy) in a greenhouse simulating the temperature pattern of a Mediterranean environment. The yield responses of Syn 2 synthetics with 2 to 20 parents with highest mean GCA were predicted from yield values of S₁ and F₁ progenies, also evaluated in Lodi. The variance of GCA effects was almost 2-fold larger than that of SCA effects. The observed vigour loss from Syn 1 to Syn 2 of the 4-parent synthetics (-6%) tended to be greater than those of higher parent number groups. The 4-parent synthetics with larger SCA effects tended to greater inbreeding depression. The comparison among synthetics with different parent number and highest GCA of their parents indicated the superiority of the 4-parent synthetic over any other in both generations (P < 0.05). The predicted yield response was maximised by the 3-parent synthetic. The results and other considerations suggest adoption of 4- to 6-parent synthetics.