The Theory-Setting-Testable Hypothesis Modei
suggest that a new setting could make a difference in the results as compared with prior research, testing a new proposition would be of interest because this would be a replication and extension. If the testing the job satisfaction- search proposition in an industrial setting had found confir- matory evidence, testing for the same relationship in a retail setting would be of more interest to the extent that some rea- son existed to anticipate that retail sales could be different from industrial.
A fruitful source of ideas for hypotheses and a test for consistency with theory is to array the concepts found in a theoretical statement on a ladder of abstraction. A great deal of research in marketing involves borrowing from other dis- ciplines, which is a process of moving up a ladder of abstraction to obtain a general statement, then moving down the ladder to apply the statement within a setting of interest to marketers. The borrowing process can be done properly if congruency exists between the theoretical statement in the mother discipline and the marketing proposition. Fishbein attitude theory extended to marketing is an example of prop- er borrowing. Borrowing is not proper if the basic theoreti- cal statement with its concepts is not congruent with the marketing proposition. As an example, Hunt (1991) con- cluded that marketing investigations of the psychophysics of price were flawed. In psychology, psychophysics involves the perceptual limits of human ability to discriminate between different stimuli, such as two identical packages weighing 15 and 16 ounces. In marketing, the interest was not in perceptual limits, but rather differences in price relat- ed to consumer buying preferences. For example, would the favorite brand be selected if it was priced at $1.89, compared with $1.79 for another brand? Psychological psychophysics did not concern such issues.
Research is strengthened if a discussion is provided to justify the appropriateness of the setting for testing hypothe- ses drawn from the theoretical statement. Parasuraman, Zeithaml, and Berry (1985) justified the services chosen for investigation by providing a cross section of services based on a classification of services proposed by Lovelock (1983). Cronin and Taylor (1992) did notjustify their settings (bank- ing, fast food, pest control, and dry cleaning) in terms of appropriateness for testing the SERVQUAL model. The SERVQUAL model covers only service delivery and excludes service outcomes, such as the taste and so on of food from a fast-food restaurant. Part of Cronin and Taylor's interest was in satisfaction, and food is probably an impor- tant facet of satisfaction for a restaurant (Swan and Trawick 1981).
In summary, the ladder of abstraction is a tool to assist the researcher in checking for consistency between the domain of the propositional level concepts and the theoretical con- cepts. The process is more than a simple mechanical proce- dure, in that the student must gain an understanding of sev- eral critical issues, including the logical tie between the the- oretical statement and hypothesis as supported by bridge laws.
The T Step: Proposition to Testable Hypothesis
The proposition has moved the research project down a lad- der of abstraction from the general concepts in the theoreti- cal statement to a specific research setting. However, an additional step down the ladder of abstraction is required for empirical testing. The example proposition "As job satisfac- tion with material rewards decreases, seeking sales work with another firm increases" is not directly testable because the prediction is not directly comparable to data. A research hypothesis is needed.
A research hypothesis is a predictive statement in which the concepts in the proposition are replaced by operational definitions, and the operational linkage is replaced by a sta- tistical test. The question of an appropriate operational defi- nition and statistical test involves several issues, many of which are beyond the scope of this article. Here the focus is on logical tests of an operational definition and the linkage between concepts. The form of the relationship (e.g., linear, power curve) between the operational definitions determine the appropriate statistical test (Hage 1972; Hunt 1991). Qperational definitions or measures of the concepts, such as "job satisfaction with material rewards," are needed.
Conceptual To Operational Definitions
The logical test of a hypothesis is that it must be homolo- gous with the proposition it tests, which implies that the operational definitions of each concept must be within the domain of its respective propositional level concept (Dubin 1978). It is quite possible that more than one operational level definition can be obtained for a single propositional concept, and each additional operational measure can create another hypothesis, so that one proposition can support sev- eral hypotheses (Dubin 1978). This occurs because of the wide domain of propositional concepts, and for each dimen- sion of the concept another operational definition may be possible.
The example proposition "As job satisfaction with mate- rial rewards decreases, seeking sales work with another firm increases" requires operational definitions of job satisfac- tion with material rewards and seeking sales work with another firm. Assume that the operational measures used were (1) the "pay" component of the Churchill, Ford and Walker (1974) measure of industrial sales job satisfaction (INDSALES) and (2) a self report (JQB SEARCH) by sales- people of job search activities such as reading trade publica- tions to leam of sales openings. The testable hypothesis would be the following:
Hj: As satisfaction with pay (measured by the pay component of INDSALES) decreases, job search activities are likely to increase (as measured by JOB SEARCH).
A lack of consistency between a concept's theoretical def- inition and its operational definition is a potential problem found in Brown and Swartz's (1989) study. They build their model on a distinction between two types of expectations: (1) expectations as experience based norms, performance that the service should provide, and (2) predictive expecta- tions, the performance that the consumer anticipates will be provided. Brown and Swartz (1989) chose norms as their