0
and that for things not owned appears in Table 5. (See
Appendices J and K for complete listings.) Table 4
represents 992 items, or 83% of the total, and Table 5
represents 1090 items, or 91% of the total.
For all items and categories in Tables 2 to 5, the null
hypothesis of uniform distribution across the 10 recall
positions was tested with the Kolmogorov-Smirnov Test of
Goodness-of-Fit using the SPSS sub-program NPAR TESTS (Hull
& Nie, 1981). This test is exact even for small
distributions and is reportedly more powerful than the
chi-square test (Conover, 1980). Exemplars and categories
of exemplars with non-random recall positions are indicated
by asterisks in their respective tables. Descriptively, the
mean recall positions indicate the central tendencies of
those non-random distributions.
Also, measures were made of the extent to which the
recall tasks produced similar listings. For the
uncategorized data, 12 of the 26 most frequently listed
items were the same for both listings. The correlation for
frequency of recall for these 12 items was positive, but not
significant (r = .26, p > .0l1). However, the correlation
between the two recall listing for the 72 categories was
significant (r = .45, p < .0l), indicating that about 20% of
the variance in category frequency was due to common