The problems of analysing tumbaga alloys are well known (Scott,1995: 504); loss of copper from the surface occurs both by natural corrosion and, in the case of more finished objects, by deliberate treatment to improve the surface colour. Furthermore, the metal microstructure of tumbaga castings is frequently inhomogeneous (see paper by Meeks, this volume). Because of these problems, the choice of where to analyse the object is very important. The aim must be to analyse an are a which is most likely to reflect the original composition of the alloy at manufacture. The surface metal is least likely to do so, and mar not even accurately reflect the finished appearance of the newly manufactured object after it has suffered many years of corrosion. Any method which analyses the surface, and that includes most socalled non-destructive methods, can only produce semi-quantitative results at best. Table 1 gives the range of compositions obtained from analysis of different areas of two small Muisca tunjos (photographs 4 and 5) to illustrate the variations in composition which can occur.