The greenhouse gas nitrous oxide (N₂O) can be both formed and consumed by microbial processes in the soil. As these processes fractionate strongly in favour of ¹⁴N and ¹⁸O, d¹⁵N and d¹⁸O gradients of N₂O in the soil profile may elucidate patterns of N₂O formation, consumption or emission. We present the first in situ data of such gradients for a mesic typic Haplaquod seeded with potatoes. During two years of measurements, N₂O concentrations were consistently higher and its d¹⁵N signature lower in the subsoil (48 or 90 cm depth) than in the topsoil, indicating production of N₂O in the subsoil. Differences of 30 ‰ between topsoil and subsoil at the same date were regularly observed. The highest N₂O concentration of 100385 mL m^-3 at 90 cm depth on July 1^st, 2003 was preceded by the lowest d¹⁵N-N₂O value of -43.5 ‰ one week earlier. This was not followed by a significant topsoil flux, but rather by a 150-day period general decrease of N₂O concentrations at 90 cm depth to 1723 mL m^-3 and simultaneous enrichment of d¹⁵N-N₂O to 7.1 ‰. There was a negative logarithmic relationship between N₂O concentration at 90 cm depth and its d¹⁵N signature. This indicated a d¹⁵N signature of -40 to -45 ‰ during production of N₂O in the subsoil, and subsequent enrichment during consumption. We concluded that the isotopic signature of topsoil-emitted N₂O is the resultant of various processes of consumption and production at different depths in the soil profile. It is therefore not a reliable estimator for the average d¹⁵N signature of N₂O in the soil atmosphere, nor for indirect losses of N₂O to the environment. This will further complicate efforts to draw up a global isotopic budget for N₂O.