Computer-aided techniques for interpreting multispectral data acquired by Landsat offer economies in the mapping of land cover. Even so, the actual establishment of the statistical classes, or "signatures," is one of the relatively more costly operations involved. Analysts have, therefore, been seeking cost-saving signature extension techniques that would accept training data acquired at one time or place and apply them to another. Signatures may be extended in preprocessing steps or in the classification steps that follow. In the present example, land cover classes were derived by the simplest and most direct form of signature extension: Classes statistically derived from a Landsat scene of the Puget Sound area, Wash., were applied to the adjacent Landsat scene of the Portland area, Oreg., acquired during the next 25 seconds down orbit. Many features can be recognized on the reduced-scale version of the Portland land cover map shown in this report although no statistical assessment of its accuracy is available.