Date of Degree


Document Type


Degree Name





Conrad Berenson

Committee Members

Leon Schiffman

Edward Wolf

William Dillon

Subject Categories



The theoretical conceptualization of attitude presented in this dissertation assumed that attitude serves as a precursor of behavior. In order to validate the attitudinal construct, empirical research must provide evidence that a consistent relationship exists between attitude measured at a given point in time, and behavior measured at a later point in time than the professed attitude. Previous research has generally failed to support the consistency of the attitude-behavior relationship. The inconsistency in this relationship has been attributed to the contamination of several types of factors (i.e., methodological, personal and situational).

In this dissertation it was argued that the relationship between brand attitudes and purchase behavior was affected by two methodological factors and one personal factor. More specifically, this researcher postulated that attitude measurement (i.e., via magnitude estimation or category scaling), the conceptualization of attitude (i.e., by form of multi-attribute model used) and a personal factor (i.e., internal-external locus of control) were in large part responsible for this weak predictive relationship.

An "after only with control group" experimental design was utilized in order to examine the effects of these factors. A total sample of 375 females aged 18-49 who currently used toothpaste (residing in the New York Metropolitan area) were included in the data base, with approximately one-half of the total falling into each experimental group. Each of the groups completed a typical attitude and usage questionnaire concerning toothpaste. The questionnaires were identical in layout of questions, while the scaling technique used to complete the attitude questions differed within experimental group. Group I utilized magnitude estimation to answer the product specific attitude questions, and Group II used a traditional category scaling method (i.e., a seven-point bi-polar scale) to answer these same questions. This design allowed for the testing of all hypotheses in the study, because the variations of the multi-attribute model tested were only mathematical manipulations of data which were collected in both cells. The data on internal-external locus of control were also gathered in both groups. Respondents were recontacted by telephone approximately four weeks after the initial interview in order to obtain additional behavioral information needed for the testing of hypotheses.

The hypotheses set forth concerning the impact of certain methodological and personal factors on attitude-behavior consistency were tested via discriminant analysis. One of the methodological factors, type of attitude measurement device, did not significantly effect the predictive ability of the attitude-behavior relationship (although magnitude estimation consistently exhibited greater prediction on an absolute basis). After analyzing the impact of the other methodological factor, form of multi-attribute model utilized, it was found that greater attitude-behavior consistency existed when the model was presented in disaggregated, rather than aggregated form; while the inclusion of importance weights had no significant positive impact on prediction. Locus of control, the personal factor studied, was generally shown to have a moderating effect on the attitude-behavior relationship. In a more specific sense, externals were found to be more likely to show brand differences on product attributes which were related to social approval.


Digital reproduction from the UMI microform.

Included in

Business Commons