The use of ordination (Legendre and Legendre 1998) and cluster (classification) analysis (Krebs 1989) based on similarity index analysis to quantify faunal data originally stems from ecology (e.g., Lichstein et al. 2002). The methods have been applied to palaeontological data for some time (Etter 1999). Ordination methods attempt to order the sampled faunal data in a continuum according to a given set of guidelines. In the ordination techniques applied here, the ordering is calculated based on values of the similarity between faunas, which are in turn calculated using simple algorithms (Jongman et al. 1995). In contrast to cluster analysis, ordination seeks to accentuate the intergradational aspects of the sample data (Etter 1999). The dimensions of the multivariate taxa space are therefore reduced meaningfully, graphically displaying relationships between the sampled units. Ordination provides a method of modelling the pattern of the index value, and the results can be compared to e.g., palaeomagnetic data to aid plate reconstructions.

Clustering or classification is a straightforward method of visualizing association data, however, the confidence of the nodes are highly dependent on data quality, and levels of similarity for cluster nodes is dependent on the similarity index used. The mean linkage (UPGMA) method is chosen here, because clusters are joined based on the average distance between all members in the two groups. Krebs (1989) noted that mean linkage is superior to single and complete linkage methods for ecological purposes because the other two are extremes, either producing long or tight, compact clusters respectively. There are, however, no guidelines as to which mean-linkage method is the best (Swan and Sandilands 2001).

Mean-linkage and ordination (PCO and NMDS) analysis was performed on brachiopod data from all Ordovician and Silurian epochs and stages. Irrespective of the specific choice of ordination and cluster methods and the different way these calculate faunal associations or taxon-space organization of samples, using several methods allows for consistent cross-validation of results (Harper et al. 1996). As a further check on results, analyses were performed at both epoch and stage levels of temporal resolution. Here some examples of the cluster analysis results in the form of dendrograms are presented at epoch level of resolution. Likewise presented are examples of the ordination diagrams at stage level of resolution.