constrained the respondents to make judgements on each of the 24 verbs under study and only 
on those verbs. The data from the four tasks were used to generate six expressions of the 
similarity of the verbs to own. First, following Friendly's (1977) advice on semantic proximity 
analysis, counts were made of the frequency of the recall of a verb adjacent to the recail of 
own: 
Briefly, this procedure locates subjective memory units on the basis of ordinal separation 
or proximity between pairs of words in recall protocols. A proximity measure is developed 
from the assumption that items which are coded together in subjective units will tend to 
be recalled contiguously at the time of test. This assumption is entailed by all measures 
of clustering and subjective organization. An item-by-item proximity matrix can then be 
constructed with numerical entries representing the degree to which each pair tends to 
occur in contiguous {though not necessarily adjacent) output positions over some set of 
trials. (Friendly, 1977, p. 197) 
A second, less restricted semantic proximity measure was made by counting the frequency of 
recall of a verb in the same recall cluster with own, with cluster here being defined as a 
respondent's entire recall production. A third measure was the frequency of mention of a verb 
in the explanation of the meaning of own to a child. A fourth, and more restricted measure was 
to limit this count to first-mentioned verbs. A fifth measure was the frequency with which a verb 
was clustered with own on the card sort task. The sixth measure was the scaled judgement of 
similarity to own. 
Because assumptions of equal interval units and of metric equivalence between 
respondents’ data would have been unwarrented, a rank-order transformation was done on each 
respondent’s scores on each of the six measures. As is standard practice in ordinal ranking, 
identical scores were assigned the average of the shared ordinal ranks. The computer program 
to calculate the six measures appears in Appendix F, and the program to rank-order each 
respondent’s scores appears in Appendix G. 
The results of the recall task appear in Table 10. Generally, constrained free-recall was 
not a successful research task for the population examined and the social contexts of the 
research. Because Cree respondents objected to tape recorded responses, no audio record 
of the temporal clustering of the recall responses could be made, and thus temporal clustering 
in the recall sequences could not be examined. Many older Cree would elaborate on the 
meanings and the usages of words as they recalled them. English-Canadians, on the other 
— were commonly suspicious that the recall task was an IQ test and were discomfited by 
perceived poor performance. The recall production of the Cree had a mean of 5.65 words 
(S.D. = 2.20), which was significantly less (F =2.42, p <.01) than the English-Canadians’ mean