Self-regenerating annual medics (Medicago spp.)in pasture-crop rotations may benefit both cereal and livestock production through fixing nitrogen, providing a disease break for cereals, and improving livestock fodder supply (Latta and Carter, 1998). These benefits are often reduced due to annual grasses in the pasture. This study examined grass control options in a 3 year pasture-pasture-wheat rotation with a mechanical fallow and a selective grass herbicide. In a replicated experiment 375 ml/ha fluazifop (500 ml/L a.i.) applied in August 1994 and a mechanical fallow treatment, commenced in July 1995 was compared on pasture and cereal production in a low rainfall region of south-eastern Australia. Sheep grazed plots at 1.25 DSE/ha in 1994 and 4 DSE/ha in 1995. Plots were sown to wheat in June, 1996, and grain yields and protein percentages measured at harvest in November. Rain in the 3 years totalled 164, 403 and 333mm. The average is 338mm. The densities of established medic plants in 1994 and 1995 were similar (200 to 300 m^-2). There was a lower grass density in 1995 due to 1994 selective grass control. Pasture production was similar until the mechanical fallowing in July 1995. This also resulted in lower densities of medics regenerating in 1997 following the 1996 wheat crop (Table 1). The chemical grass control treatment had higher volunteer grass densities in the wheat phase of the rotation. This was reflected in a lower grain yield (Table 2). The use of costly selective grass herbicide almost 2 years prior to a cereal crop is desirable to ensure cereal root disease control, maximise legume production and extend the period of livestock fodder (Latta and Carter, 1998). Fallowing resulted in a higher grain yield, however, it also necessitates re-establishing a medic pasture following the cereal phase. Community concerns relating to soil erosion will ultimately force the abandonment of this practise irrespective of economic considerations.