We simulated bird surveys using recorded bird songs to assess factors affecting detection probability in grassland bird point counts. We used mixed effects logistic regression models to estimate effects of those factors and to estimate and visualize the variation in the area around the observer where birds can be perceived (the perception area). We simulated surveys with 8,926 binary opportunities for detection in Minnesota grasslands in 2011 and 2012. Species, distance to the observer, wind speed and direction, observer, and density of vegetation all affected detection of recorded bird songs. Species had a strong effect; the size of the predicted perception area around the observer differed by an order of magnitude among species. Wind also had a strong effect on detection. As wind speed increased, probability of detection downwind of the observer was reduced and the perception area around the observer became smaller and more asymmetrical. The effective distance at which an observer is more likely to detect a bird than not detect it may differ among species and angles to the wind, even within the same survey. Eight of 10 species had low probability of misidentification (≤0.03), but Grasshopper Sparrow (Ammodramus savannarum) and LeConte's Sparrow (Ammospiza leconteii) were frequently misidentified (probability = 0.09–0.24 among observers), contributing to a low rate of correct detection for those species. We recommend collecting point-count data within distance bands so that data can be analyzed based on the effective radius for each species and standardizing surveys across wind conditions to reduce variation in detection probability.