Rangeland management plan is one of the major means of management and utilization of rangelands in Iran. The formulation of these plans for the rangeland users should be studied from the ecological and socio-economic points of view. One of the main envisaged activities in almost all the management plans is the grazing system. An experiment was used to study the effects of three deferred grazing systems (15, 30, 45 days delay) and control treatment (Non- grazed) on vegetation parameters in semi-arid rangelands of Jashlubar in Semnan province. The experiment was conducted using a completely randomized block design (RCBD) with three replications over 6 years (2006-2011). In each experimental unit, data from three life forms of vegetation (shrub, forbs and grasses) were collected along a 30 m transect within ten fixed quadrates (0.5 ´ 0.6 m2) appropriate to vegetation sizes. In addition, forage productions of two life forms (forbs and grasses) were collected over 5 years. Data were analyzed using SAS software and means comparison was made based on Duncan’s method. The results showed the significant effects of deferred grazing systems on the growth of shrubs in terms of canopy cover. However, there were no significant differences between treatments for canopy cover percent of forbs and grasses. The lowest shrub canopy cover was obtained in 45-day delay of grazing. There were also significant effects of deferred grazing systems (15, 30 and 45 days delay) on forage production of both forbs and grasses (P<0.05). Result of means comparison showed that the best delay time for the rangeland utilization of this area considering the annual precipitation is the 15-day delay with the highest forage production for forbs and grasses and the highest cover percent for shrubs.