This study examined the attributes of the oil signal embedded in the near infrared (NIR) reflectance spectra of intact canola seed. The oil/spectrum relationship was examined with the aid of the following techniques: comparison of raw log 1/R spectra, investigation of absorption difference spectra and investigation of oil/wavelength correlations. The 1100-2500 nm reflectance spectrum of canola oil contains three major segments, 1150 to 1250 nm, 1650 to 1800 nm and 2250 to 2380 nm, that denote absorption of the constituent. The spectrum of intact canola seed closely resembles the spectrum of oil. However, the strength of oil absorption in the seed is significantly reduced, particularly at the longer wavelengths. Examination of the NIR spectra revealed that the oil signal embedded in the reflectance spectrum of intact canola seed was primarily affected by the spectral characteristics of the seed. Signal correction tools-standard normal variance, detrending and the second order derivative-applied to the data showed an unreliable efficiency in minimising scattering defects in the spectra. The absorption difference spectra displayed variable patterns in the manifestation of differences in the oil concentration between canola samples. These differences could be observed either in all or only in some of the oil-specific segments of the spectrum. Correlograms illustrated a very complex spectrum/oil relationship, displaying a large number of wavelengths strongly correlated to the oil across the entire 1100-2500 nm spectrum, including a considerable number of wavelengths, which displayed chance correlations. Therefore, NIR analysis of oil in the intact canola seed preferably should involve the full spectrum information and should involve the use of latent variables, rather than the original wavelength space.