Alfalfa ( Medicago sativa L.) is widely grown as a forage crop due to its good quality characteristics and high adaptability. However, seed yield is generally considered to be of secondary importance and is characterized by fluctuating yields with often poor seed quality. A field experiment using five alfalfa cultivars (Equipe, Iside, Lodi, Robot, Romagnola) was carried out in 1995, 1996 and 1997 at Foggia (southern Italy) to evaluate the effects of four defoliation practices (H1: crop mown during early plant growth; H2: plant desiccation by chemical agent at the end of February; H3: crop mown at early flowering; H4: never cut) and two irrigation treatments (I: irrigation applied from April to beginning of seed filling; NI: non-irrigated control) on seed yield, seed yield components and seed quality (as determined by seed germination with and without accelerated ageing (AA)). The relationships between yield components (stems per m 2, pods per stem, seeds per pod, 1000-seed weight) were determined by path-coefficient analysis. Irrigation significantly increased seed yield; on average doubling the control yield over the three seasons. However, the potential seed yield (calculated from seed yield components) was, on average, five times the actual seed yield. The two mowing treatments produced consistently higher seed yields than either desiccated or untreated swards. Cultivar differences were evident for seed yield, with Equipe having the highest value (40% higher than the mean of the other cultivars). Irrigation improved the yield primarily because it led on average a greater than four fold increase in the density of stems (the most influential yield component). By contrast, seeds per pod and 1000-seed weight increased in the absence of irrigation. Defoliation treatments had little effect on stems per m 2, seeds per pod and seed weight, whereas pods per stem were reduced by desiccation. Path analyses calculated across irrigation treatments and years revealed that stems per m 2 had the largest positive direct effect on alfalfa seed yield under each harvest management (path-coefficient values ≥0.89), and its indirect effects on seed yield via other traits were negligible. Seed quality, as measured by germination percentage both before and after AA, remained relatively consistent across both defoliation and irrigation treatments and cultivars, but was anomalously low in the 1997 irrigated crops. Overall, the highest seed yields were obtained under irrigated conditions when crops were mown during early growth or at early flowering. The potential seed yield of alfalfa varieties is sufficient to guarantee a profitable seed harvest. However, the harvest efficiency of the combine-harvester was low (20% of the potential seed yield); thus, more appropriate harvest techniques should be used.