State transition from perennial grassland to shrubland or woodland is often synonymous with land degradation and desertification in dryland ecosystems. These physiognomic transitions markedly alter rates and dynamics of ecological processes and the ability of ecosystems to provide services valuable to society. Land managers and other stakeholders could benefit from research designed to identify vegetative structural properties that forecast when ecological sites might be at risk for state changes. We sought to determine such indicators via removal experiments conducted along an extant Bouteloua eriopoda-to-Prosopis glandulosa encroachment gradient at the Jornada Basin LTER site in the northern Chihuahuan Desert. Ninety plots centered upon B. eriopoda patches were established along the grassland-to-shrubland gradient. The size of and distance to all P. glandulosa individuals within a 5 m radius was recorded and aboveground biomass was estimated using site- and species-specific allometric regression (R2 = 0.89). Peak season B. eriopoda biomass was quantified in 2010. P. glandulosa shrubs within 5 m of a subset of the grass plots (n = 45) were then killed via foliar herbicide in summer 2011. Annual B. eriopoda production was then quantified in subsequent growing seasons. B. eriopoda cover along the grassland-to-shrubland gradient ranged from 26.3% (grassland end) to 3.6% (shrubland end) and declined exponentially as P. glandulosa cover and biomass increased linearly from 2.2% and < 200 g m-2 to 17.9% and > 400 g m-2. In 2013, mean (SE) B. eriopoda productivity in plots with shrub neighbors removed was 95 (6) g m-2 y-1, compared to 53 (6) g m-2 y-1 in plots with shrub neighbors intact (a 1.8-fold increase; p < 0.0001). The grass production response to shrub removal was significant in both the grassland and shrubland portions of the gradient, more so in the latter [2.7-fold increase; p < 0.0001] than in the former [1.4-fold increase; p = 0.001]. However, B. eriopoda productivity did not differ between the control and shrub removal treatments in the ecotone (shrub biomass 200 – 400 g m-2), owing to highly variable grass productivity in the control plots (range = 21 to 212 g m-2 y-1). Variability in grass ANPP may therefore be a more reliable indicator of impeding state transition than mean ANPP. Forthcoming analyses will examine how the structure (e.g. size, number, density) of the shrub neighborhood influences grass productivity.