Technical Abstract: Theoretical models predict that dryland ecosystems can cross critical thresholds after which vegetation loss is independent of initial drivers, but experimental data are nonexistent. We used a long-term (13 year) pulse-perturbation experiment featuring heavy grazing and shrub removal to determine if critical thresholds and their determinants can be demonstrated in Chihuahuan Desert grassland. We asked if cover values or spatial pattern metrics could predict vegetation change, supporting their use as early-warning indicators. We found that season of grazing, but not shrubs, mediated the severity of grazing impacts on the dominant grass species. Recovery occurred at the same rate irrespective of grazing history, suggesting that critical thresholds were not crossed, even at low cover levels. Grass cover, but not spatial pattern metrics, predicted variation in recovery rates. Our results indicate that threshold models and pattern –based early warning indicators should be applied cautiously, even in transition-prone ecosystems.