A comparative study was made of 3 techniques applied to the study of herbivore diet selection, namely direct observation, faecal analysis and the recently developed remote control oesophageal fistula valve, using 3 animals over 4 study days. Direct observation showed a relatively high level of precision with respect to the woody forage class but a poor measurement of the grass class. The ratios of grass to dicot were similar in the diets determined by direct observation and valve fistulation, but faecal analysis over-emphasised dicots relative to the other techniques. The greatest overlap in estimated diet was between faecal analysis and valve fistulation. Overall the valve fistulation technique was considered superior to the other 2 techniques because it provided reliable estimates of diet composition that could be readily equated to range conditions at the time of ingestion.