Near-infrared reflectance spectroscopy (NIRS), a nondestructive analytical technique, may someday be used to rapidly and simultaneously quantify several soil properties in agricultural fields. The objectives of this study were to examine the influence of moisture content on the accuracy of NIRS analysis of soil properties and to assess the robustness of a NIRS multivariate calibration technique. Four hundred agricultural soil samples (<2 mm) from Iowa and Minnesota were studied at two moisture levels: moist and air-dried. The soil properties tested included total C, organic C, inorganic C, total N, CEC, pH, texture, moisture, and potentially mineralizable N. About 70% of the Iowa samples were selected for the calibration set, and the rest of the Iowa samples and all of the Minnesota samples were assigned to validation set I and validation set II, respectively. Calibrations were based on partial least-squares regression (PLSR), using the first differentials of log (1/R) for the 1100 to 2500-nm spectral range. The results for the calibration set and validation set I indicated that NIRS-PLSR was able to predict many soil properties (total C, organic C, inorganic C, total N, CEC, % clay, and moisture) with reasonable accuracy for both the air-dried (R² > 0.76) and moist (R² > 0.74) soils. The results for validation set II showed that NIRS-PLSR was able to predict some properties of soils (total C, organic C, total N, and moisture content) from a different geographic region, but other soil properties in validation set II were not accurately predicted. Although NIRS-PLSR predictions are slightly more accurate for air-dried soils than for moist soils, the results indicate that the NIRS-PLSR technique can be used for analysis of field moist samples with acceptable accuracy as long as diverse soil samples from the same region are included in the calibration database.