1. Transitions from grass to woody plant dominance, widely reported in arid systems, are typically attributed to changes in disturbance regimes in combination with abiotic feedbacks, whereas biotic mechanisms such as competition and facilitation are often overlooked. Yet, research in semi-arid and sub-humid savannas indicates that biotic interactions are important drivers in systems at risk for state transition. We sought to bridge this divide by experimentally manipulating grass-on-shrub and shrub-on-shrub interactions in early and late stages of grassland-shrubland state transition, respectively, and to assess the extent to which these interactions might influence arid land state transition dynamics. 2. Target Prosopis glandulosa shrubs had surrounding grasses or conspecific neighbors left intact or killed with foliar herbicide, and metrics of plant performance were monitored over multiple years for shrubs with and without grass or shrub neighbors. 3. Productivity of small shrubs was enhanced by grass removal in years with above-average precipitation, a result not evident in larger shrubs or during dry years. Proxy evidence based on nearest-neighbor metrics suggested shrub-shrub competition was at play, but our experimental manipulations revealed no such influence. 4. Competition from grasses appears to attenuate the rate at which shrubs achieve the size necessary to modify the physical environment in self-reinforcing ways, but only during the early stages of shrub encroachment. Our results further suggest that at late stages of grassland-to-shrubland state transitions, shrub-shrub competition will not slow the rate of shrub expansion, and suggest that maximum shrub cover is regulated by something other than density-dependent mechanisms. We conclude that grass effects on shrubs should be included in assessments of desert grassland state transition probabilities and rates, and that desertification models in arid ecosystems that traditionally focus on disturbance and abiotic feedbacks should be broadened to incorporate spatial and temporal variations in competitive effects.