Seed production of subterranean clover (Trifolium subterraneum L.) in mixtures with lovegrass (Eragrostis curvula (Schrader) Nees cv. Consol), cocksfoot (Dactylis glomerata L. cv. Currie), phalaris (Phalaris aquatica L. cv. Sirolan), danthonia (Austrodanthonia richardsonii (Cashm.) H.P. Linder, cv. Taranna), and lucerne (Medicago sativa L. cv. Aquarius) was compared with pure and degraded (invaded by annual volunteers) annual subterranean clover pasture at 2 sites (Junee and Kamarah) in the southern wheatbelt of New South Wales. Seed yields, clover seedlings in winter, and the change in the proportion of 3 subterranean clover cultivars (Dalkeith, Seaton Park, Goulburn) when grown with and without perennials were assessed. The effect of thinning the perennials to 10 plants/m2 on clover seed set was examined at the drier site. Seed production of subterranean clover in the mixtures was depressed by up to 50% compared with the pure and degraded annual swards. Initial clover seed poduction in the mixtures was at least 60 kg/ha even in the drought year at the wetter site (Junee), and >85 kg/ha at Kamarah, the drier site (seedling establishment at Kamarah failed in the drought year). Clover seed reserves in the following 2 years progressively increased to >300 kg/ha in the perennial swards at Junee but were <100 kg/ha by the end of the third year at Kamarah. In comparison, seed reserves in the pure clover and degraded annual swards were >650 kg/ha at Junee and >350 kg/ha at Kamarah. Reducing perennial density to 10 plants/m2 at the drier site increased clover seed yield about 3-fold in the first year compared with unthinned perennial swards. The increased seed yield was due to increased numbers of burrs set and increased seeds per burr and, in all perennial pasture treatments except lucerne, increased seed size. Clover seedling regeneration in 3rd and 4th year after sowing was substantially lower in the perennial-based mixtures than annual plots, with a significant (P < 0.05) positive correlation at both sites between clover seedling regeneration and seed bank size (1996, r2 = 0.46–0.64; 1997, r2 = 0.64–0.85). Following false breaks in early autumn, clover seedling populations were substantially higher in the pure and degraded clover treatments than in most perennial treatments. The proportion of the 3 cultivars present in the seed bank at the end of the pasture phase differed between sites but the sward type only influenced the proportion at the drier site. At the medium rainfall site, the later maturing cultivar Goulburn constituted 27–54% of the seed bank and the early flowering Dalkeith 25–46%, with unsown cultivars being insignificant ( <1%). At the low rainfall site, Dalkeith was the major component (33–52%) of the seed bank but the background population of unsown cultivars constituted 11–48%, the lowest proportion being in swards without a perennial component. The proportion of Goulburn was highest (23%) in the pure sward and lowest (10%) in lucerne and phalaris. It was concluded that subterranean clover could form relatively stable mixtures with perennials in medium rainfall environments, with clover populations increasing with time. In lower rainfall environments, clover seedling populations in perennial swards may be low due to reduced seed set and decreased seedling survival following early autumn rains. In these environments earlier maturing, hard-seeded cultivars are more likely to persist in mixtures and there is more potential for unsown cultivars to constitute a greater proportion of the sward. Decreasing perennial density offers scope for improving clover seed set and survival in these environments.