Information on denitrification (particularly N-2) losses from dry ecosystems is limited despite their large area. Here, we present the first direct denitrification measurements for a northern hemisphere savanna, a Prosopis-dominated grassland/grove matrix in south Texas. We used the gas-flow intact soil core method to quantify N-2, N2O and CO2 losses and compared these with field measurements of N2O, NOy, NH3 and CO2. Under field-realistic soil moisture and O-2 conditions (average 17.5-20 % O-2, minimum 15 %) incubated soils produced no measurable N-2 flux (detection limit 52.2 A mu g N m(-2) h(-1)). Only in a subset of grove soils were fluxes of 70-75 A mu g N m(-2) h(-1) recorded after 102 h of incubation at 5-10 % O-2 following wetting of very dry soils. Making the assumption that potential N-2 production falls just below the detection limit (likely an overestimate given the conditions needed to generate measurable fluxes), N-2 flux rates would fall on the low end of that recorded for a tropical Australian savanna (45-110 A mu g N m(-2) h(-1)) under comparable abiotic conditions. Assuming maximum possible production rates, N-2 could comprise < 32-76 % of total soil N gas flux following soil wetting in summer. Lack of flux response to soil wetting in winter suggests that cold-season N-2 fluxes are negligible. N2O fluxes for core incubations were significantly higher than for field chambers; thus it is likely that incubations may overestimate N2O flux by reducing soil column consumption. Overall, results indicate that soil N-2 fluxes are less dominant in this savanna than in other ecosystems investigated.