Seed shattering, that is, the loss of seeds at ripening stage shortly before or during seed harvest, is strongly reducing seed yield in Italian ryegrass (Lolium multiflorum Lam.). The aim of this study was to evaluate the possibility to reduce seed shattering within breeding germplasm via recurrent phenotypic selection on spaced plants. Starting from a founder population of 300 plants not previously selected for seed shattering, two cycles of phenotypic selection for high and low seed shattering were performed based on spaced plants. The resulting five populations with different levels of selection for seed shattering (selected once or twice for low seed shattering, selected once or twice for high seed shattering, and no [random] selection) were phenotyped in plot trials with two harvesting dates (early and late). Seed shattering was highest in the population selected twice for increased seed shattering (7.92% for early and 28.44% for late harvesting), followed by the population selected once for increased seed shattering (7.34% early, 21.56% late), the non-selected population (6.12% early, 20.22% late), the population selected once for decreased seed shattering (4.13% early, 16.73% late), and the population selected twice for decreased seed shattering (2.41% early, 9.31% late). Selecting twice for low seed shattering resulted in a reduction of seed shattering by 10.91 percentage points at late harvest when compared to the non-selected population. Thus, recurrent phenotypic selection on spaced plants is appropriate to efficiently reduce seed shattering and to increase seed yield in sown plots, which is highly relevant for seed production.