j 472
report of the findings would eventually be made available to the respondents. [Data for this
study appear in Appendix E.]
Results
The analysis of the data for this study has been shaped by the limitations of the design,
by the conservative standards required for cross-cultural comparisons, and by the relatively
specific goals of the study. Typically, psycholinguistic semantic data on a set of related words
are transformed into similarity or distance matrices and subjected to factor analysis,
multidimensional scaling, cluster analysis, etc. (e.g. Coltheart & Evans, 1981; Friendly, 1977;
Rapoport & Fillenbaum, 1972; Rudmin & Berry, 1987; Takane, 1980; Wagener & Pohl, 1986). For
example, Study 2 used cluster analysis and factor analysis to examine the structure of a
semantic similarity matrix. However, such analyses are too indeterminate for use without
ancillary means of validation {Rapoport & Fillenbaum, 1972). For example in Study 2, the
multivariate analysis of the verb similarity ratings was made tenable by the relatively large
duplicate samples, by an existing psychological literature with which to relate the results and
interpretations, and by the researcher having native-speaker command of the language. The
present study had none of these advantages.
There is the further uncertainty of cross-cultural comparison of two groups reported and
hypothesized to be different in conceptualizations of ownership. Analyses must not presume
equivalence of the multivariate structure of the data or of the linear scale units used by the
respondents (Bijnen, Van Der Net & Poortinga, 1986; Malpass & Poortinga, 1986). Finally, the
goal of the study was limited to uncovering differences, if any, in the meaning of own. It was
not the goal of the study to describe the entire semantic space of ail the words of possession
in Cree and English. Thus, the analyses of the data avoid multivariate explorations and rely on
non-parametric statistics, which make minimal assumptions of linear equivalence between
respondents’ responses.
The respondents expressed their conceptions of the meaning of own on two
unconstrained and two constrained psycholinguistic tasks. The constrained free-recall and the
Rplanaiion of own to a child are considered to be “unconstrained” because respondents were
not required to make semantic judgements on all 24 of the verbs under study and were free to
introduce new verbs in their responses. The card sort and scaling tasks, on the other hand.