Abstract: Methods are described to assess horses' selective grazing ability that includes choosing, sorting and the adaptive value of this behaviour. Choosing ability was tested by the experimenter presenting pairs of cut plant species that were then alternated at each presentation until the test horse had taken three bites of one of the plant pair. The results were analysed in relation to five measures of choosing behaviour: (1) the strength of the choice; (2) correspondence between first bite and the final choice; (3) constancy of the choice over a number of trials; (4) the comparison of the horses' ranking of the species over a number of trials; (5) the constancy of the linear arrangement of the plants over a number of trials. Sorting ability was tested using two methods. A mixture of two plant species was presented either in a clamp or loose in a trough. Results were based on number and weight of plant residues. The adaptive value of the behaviour related to the bitterness of toxic plants. This bitterness was represented in testing by quinine sulphate and a poisonous Senecio species. An extremely bitter substance “Bitrex” was also used in this context but was totally accepted by the horses. The horses' reactions to these substances were monitored using a behavioural score chart. The results from 12 horses revealed that the horses differed individually in their grazing ability. On this basis, the horses were classified as efficient, semi-efficient, or inefficient grazers. This finding has practical implications in deciding which horses may safely graze on pastures infested with toxic plants.