The Use of Contact Sampling in Studies of Association in Vegetation

(1) The contact sampling method has been used frequently in studies of between-species association in vegetation. An earlier paper presented a probability argument for determining completely random association and a parallel statistical approach. (2) It is shown to be preferable, however, to tabulate and analyse the data with an unfolded contingency table where the ordered 'contact pairs' (i, j) and (j, i) are distinguished,-rather than with the folded (triangular) table used previously. If these ordered pairs are distinguished, a more fluent and straightforward data analysis is possible. (3) This, together with considerations of the treatment of data in which there are events of within-species 'contact pairs' (i.e. when j = i), forms the basis for recommending an 'optimal' strategy for examining tendencies for between-species association-optimal in the sense of both statistical efficacy and sampling mechanics. (4) Ecological and statistical characteristics of the recommended model are discussed in comparison with other methods. (5) The sampling scheme and data analysis for investigating within-species association are also outlined. (6) An empirical example is presented and analysed. (7) Attention is drawn to the frequent erroneous statistical treatment of data collected by recently devised 'plotless' methods such as the 'contact sampling' method. Many new techniques are not amenable to the usual 2 x 2 table methods of analysis and there has been a failure to apply properly 'independence' hypotheses. Information content and interpretational potential of data from 'plotless' sampling are discussed in the light of the required distinction between patterns and processes in vegetation.