between the subjects' judgements of the applicability of the 
various criteria to their recall exemplars. The analysis 
identifies criteria that are near to each other and 
successively links ‘them into clusters. Initially, each 
criterion is considered a cluster of one, then successive 
cluster linkages are made, and finally all criteria are 
joined into a single cluster. A hierarchy is imposed in 
that four linkages will be made on the first pass, three 
linkages on the second pass, two linkages on the third, and 
one on the fourth. 
The numbers on the dendrograms represent the greatest 
differences between criteria in the two clusters being 
linked. For example, consider Figure 1. On the first pass, 
the analysis identifies individual criteria (i.e. one-item 
clusters) that were valued similarly by the subjects. 
‘Familiarity' linked with 'Knowledge', 'Possession' with 
'Assertion', and so on. When 'Familiarity' and 'Knowledge' 
were linked to form a two-item cluster, the mean absolute 
difference between judgements of applicability was .30. 
This is expressed numerically on the dendrogram, as well as 
graphically. The longer the horizontal line linking two 
clusters, the more distant the clusters were when linked. 
on the second pass, when 'Aesthetics' linked with the 
'Familiarity-Knowledge' cluster, ‘Aesthetics’ was .35 
judgement units from 'Familiarity'. The cluster linking