This paper examines the potential to distinguish land cover types in digital images acquired at several different sensor view angles with respect to a fixed area on the Earth's surface. Images recorded by an airborne multispectral scanner over an area of arable farmland are used to generate four such multipleview-angle (MVA) datasets: each consists of data obtained at six sensor view angles in a single spectral waveband: green (0.52-0.605 mu m), red (0.63-0.69 mu m), near-infrared (0.76-0.90 mu m), and middle infrared (1.55-1.75 mu m), respectively.The data are initially presented in the form of single-band MVA false-colour composite images. These are used to illustrate the extent to which different surface materials can be distinguished visually in MVA data. The concept of MVA (cf.,multispectral) feature space is then introduced and the separability of different land cover types within it is explored. It is suggested that single-band MVA data contain two main components of statistical variance directional and spectral . Their relative contributions to the total statistical variance in single-band MVA data is assessed using linear correlation analysis and principal components analysis (PCA). It is shown that while the spectral component tends to dominate in all wavebands, particularly in the near-infrared, the directional component nevertheless provides an important means of distinguishing certain cover types. The implications of these findings are discussed in relation to the parameters used in current bidirectional reflectance distribution function (BDRF) models and the development of 'angular indices' for vegetation monitoring (cf., traditional multispectral vegetation indices).