Lark Sparrows (Chondestes grammacus) are declining throughout most of their range. Effective management for this species is hampered because relatively little is known about nesting ecology. We studied habitat characteristics affecting Lark Sparrow nest-site selection and nest success at nine study pastures in a southern mixed-grass prairie in Oklahoma. We used a neural-network technique to discriminate between nest and random locations, and bootstrapping with 95% confidence intervals to compare habitat features of successful and unsuccessful nests. We quantified habitat features at the nest and random points during the breeding seasons of 1999 and 2000 among three grazing treatments (control, moderate, and heavy). We located 40 nests during two years of the study, for which crude nest-success was 26.3%. Most nests were located in either moderately grazed pasture (55%) or heavily grazed pasture (40%). The neural model correctly identified nest and random points 91% of the time. Percentage of structural cover, distance to nearest structural element, bare-ground exposure, and percentage of litter cover were the most important nest-site selection criteria according to the model. Simulation analysis indicated points were classified as nest sites if they were <270 cm from structural elements, <87% bare-ground exposure, <74% litter cover, and >9% structural cover. Successful nests had less bare-ground exposure (x̄ = 6.2 ± 1.9% [SE]) and more litter cover (x̄ = 18.0 ± 4.6%) compared to unsuccessful nests (x̄ = 17.5 ± 3.8% and 10.1 ± 1.6%, respectively). These results suggest that habitat management for Lark Sparrows in mixed-grass prairie should focus on creating abundant structural cover with moderate levels of litter accumulation and bare ground.