The legume content of the diet of groups of resident cattle (LDR) continuously grazing a number of grass-stylo pastures was determined by the faecal carbon isotope ratio (�13C) technique on various occasions. At the same time the pastures were sampled by non-resident steers fistulated at the oesophagus (OF steers) and the legume content of their extrusa (LEO) was determined. Overall, the linear relationship between LEO and LDR had a coefficient of determination (r2) of only 0.127, indicating that LDR could not be reliably predicted from LEO. The LEO of morning samples was often markedly different from LEO of afternoon samples. The implications of the results for diet selection work on grass-legume pastures using OF animals are discussed. Researchers need to recognise that quality attributes of extrusa collected from OF animals do not necessarily reflect those of the diet of resident test animals.