Abstract. Fire frequency exerts a fundamental control on productivity and nutrient cycling in savanna ecosystems. A single fire event often increases short-term nitrogen (N) availability to individual plants, but repeated burning causes ecosystem carbon and N losses and can ultimately decrease soil organic matter and N availability. However, these effects remain poorly understood due to limited long-term biogeochemical data. Here, we leveraged one of the longest running prescribed burn experiments (established in 1964) to evaluate how fire frequency and changing vegetation composition influenced wood stable N isotopes (δ15N) across space and time. We developed multiple δ15N records across a burn frequency gradient from precisely dated Quercus macrocarpa tree-rings in an oak savanna at Cedar Creek Ecosystem Science Reserve, Minnesota, USA. Sixteen trees were sampled across four treatment stands that varied in temporal onset of burning and burn frequency, but were consistent in overstory species representation, soil characteristics, and topography. Burn frequency ranged from an unburned control stand to a high fire-frequency stand that burned in four of every five years during the past 55 years. Because N stocks and net N mineralization rates are currently lowest in frequently burned stands, we hypothesized that wood δ15N trajectories would have declined over time in all burned stands, but at a rate proportional to fire frequency. We found that wood δ15N records within each stand were remarkably coherent in their mean state and trend through time. A gradual, temporally synchronous decline in wood δ15N occurred in the mid 20th century in the no-, low-, and medium-fire stands, whereas there was no trend in the high-fire stand. The decline in the three stands did not systematically coincide with the onset of prescribed burning. Thus, we found limited evidence for variation in wood δ15N that could be attributed directly to long-term fire frequency in this prescribed burn experiment in temperate oak savanna. Our wood δ15N results may instead reflect decadal-scale changes in vegetation composition and abundance due to early to mid 20th century fire suppression.