The primary objective of this study was to investigate the spatial variability of N-15 natural abundance (delta-N-15) in a non-N2-fixing plant (Triticum aestivum L.) on two fields. Two center intersecting 50-point perpendicular transects were sampled on two separate fields, on Klassen (Udic Boroll) and Kruger (Typic Boroll) soils in Saskatchewan, Canada. Autocorrelation and semivariograms were used to assess the spatial structure of delta-N-15. Data from the Klassen field indicated that variation in grain delta-N-15 was random, with no spatial structure at the scale of investigation (2-50 m). The pooled mean grain delta-N-15 value and 95% confidence band was 5.43 +/- 10 parts per thousand, with a coefficient of variation of 8.9%. Observed variability in delta-N-15 was primarily accounted for by analytical error in the Klassen field. The range in delta-N-15 values for Klassen was 2.80 parts per thousand, and the maximum difference between samples separated by 2 m was 2.00 parts per thousand. These data indicate that, even in a low-variability field, site-to-site variation may be significant. Thus, N2-fixing plants should be grown in close proximity to the reference crop for quantifying N2 fixation. Spatial variability in delta-N-15 for the Kruger field was significantly greater than in the Klassen field. Anisotropy was noted in delta-N-15 in the Kruger field, with a pure nugget effect observed for one transect; for the other, a spherical model described the semivariance pattern with a range for spatial dependence of 16 m. The coefficient of variation and the minimum sample size required for estimating delta-N-15 were corrected for the influence of serial correlation. Mean grain delta-N-15 values for the two Kruger transects were 4.45 +/- 0.13 parts per thousand, and 4.29 +/- 0.11 parts per thousand. It is hypothesized that the greater variability in grain delta-N-15 in the Kruger field resulted from a shorter lag time between fertilizer-N application and sampling, that is 1.3 yr at Kruger, and 4 yr at Klassen.