i% 
pairs using the Wilcoxon Signed-Ranks Test in a protected 
fashion, i.e. following a significant Friedman Two-Way 
ANOVA Test (Church & Wike, 1980). All computations were 
done with SPSS sub-program NPAR TESTS (Hull & Nie, 1981). 
All four Friedman Tests were statistically significant (p < 
.0l). The rank I of the criteria for the recall of 
things owned appears in Table 6, for recall of things not 
owned in Table 7, for the recall difference scores in Table 
8, and for explicit judgements in Table 9. The results of 
the Wilcoxon tests of pair-wise comparisons appear in the 
right-most columns of the tables. The criteria codes 
indicate the nearest following criterion that is 
significantly different (p = .01). For example, in Table 6, 
criterion #2, ‘Assertion’, ie not different from 
'Familiarity', but is from 'Knowledge' and all subsequent 
criteria. Criteria that are not statistically different 
from each other are not separated by a space in the tables. 
Cluster Analysis 
It was also important to examine the interrelationships 
among the criteria judgements to determine whether or not 
there might be coherences among sub-groups of criteria. In 
a parametric analysis, this might have been done by factor 
analysis or discriminant analysis. In the present study, 
cluster analysis was used because it requires no assumptions 
of multivariate normality in the distributions of the data.