Motivating language: exploring theory with scale development.

Although the history of leadership research is long and complex, assumptions about leader communication skills have often been simplistic and unidimensional. While previous researchers have conducted elaborate studies about specific dimensions of leadership including abilities, behaviors, traits,

Organizational behavior theorists have begun to examine the language through which leaders express behavior to workers (Conger, 1991; Fairhurst, 1993; Fairhurst & Chandler, 1989; Jablin, 1985; Jablin & Krone, 1994; Lamude, Daniels, & Graham, 1988; Scandura & Graen, 1984; Sullivan, 1988; Waldron, 1991). Collectively, these scholars have questioned the simplifying assumptions that remain embedded in leadership study: that most leader-worker language is an exchange of uncertainty-reducing information (Jablin & Krone, 1994; Sullivan, 1988), and that all leaders communicate equally well with their subordinates (Bass, 1990; Yukl, 1989). Sullivan (1988) gave particular theoretical examples of uncertainty-reducing leader communication including the job characteristics growth-needs model (knowledge of a commensurate task match will motivate an employee), goal setting (specificity supports goal attainment), and expectancy theory (reward uncertainty is a key component in the model). The premise that all leaders communicate equally well with their workers is implied by the "average leadership style," which has been challenged by Leader-Member Exchange and other theories (Graen & Cashman, 1975; Jablin & Krone, 1994).

While these assumptions were appropriate as control mechanisms in earlier studies, they are being reexamined to support progress in the field of organizational communication. This article attempts to further this progress through the introduction, testing, and discussion of future implications for the motivating language scale. This instrument is designed to specifically measure both a leader's general oral communication skills with subordinates and his/her strategic use of spoken language variance to motivate workers. Equally important, this scale diverges from earlier management theories by proposing a three-factor model that includes a much broader range of language use than the prevailing emphasis on uncertainty-reducing speech. Finally, the motivating language scale is firmly grounded in current management and communication literature and is designed to encompass most relevant leader oral communication with workers.

Theoretical Background

The "spoken language of leadership" has often been identified as a critical influence in worker motivation and outcomes. Much leadership research implicitly emphasizes the importance of leader speech. For example, reward contingencies and performance goals are frequently expressed within leader conversation (Daft & Wiginton, 1979; Dansereau, Cashman, & Graen, 1973; Yukl, 1989). More contemporary studies have begun to explicitly define the relationship between leader oral communication patterns and employee affect, goal attainment, and career progression (Conger, 1991; Fairhurst, 1993; Fairhurst & Chandler, 1989; Jablin, 1985; Jablin & Krone, 1994; Lamude, Daniels, & Graham, 1988; Scandura & Graen, 1984; Sullivan, 1988; Waldron, 1991).

These recent findings are important because they break management research precedence in definitive ways. To begin, these studies directly investigate language as a separate entity that transmits leader behavior. Moreover, they do not assume that most leader language reduces uncertainty. Sullivan (1988) refers to prior models for manager-subordinate communication as dual factor: uncertainty reducing and people oriented, with most speech activity falling within the first group. Sullivan (1988) also noted that these categories parallel the initiating structure and consideration dimensions of the Ohio State studies, and gave several organizational behavior examples including equity theory and operant conditioning (employees need leader information about the distribution and contingency of rewards).

The second assumption to be challenged by contemporary research trends presupposes that leaders use the same language strategy, regardless of situation or subordinate receiver (Jablin & Krone, 1994; Sullivan, 1988). Jablin (1985) reported an overall increase in research that identifies the different types of leader communication strategies. He also commented on the need for a variety of companion methodological tools that would be capable of accurately measuring this data. Fairhurst (1993) and Fairhurst and Chandler (1989) effectively used discourse analysis to show that leader speech varied according to his/her unique relationship with each subordinate. Thirdly, these innovative studies also emphasize the use of message strategies that are outcome oriented. Scandura and Graen (1984) showed how positive performance and affective outcomes were significantly related to leader training in strategic variance of speech, with the specific worker placed in the role of moderator variable.

Within this stream of literature, motivating language theory (Sullivan, 1988) provides a more comprehensive model for understanding how leadership language impacts workers. This theory predicts that strategic oral communication is an important motivational tool which has positive, measurable effects on employee performance and job satisfaction. According to the principles of motivating language theory (MLT), differences in key outcomes shown by employees are the results of variance in how well managers engage in three fundamental speech acts when communicating with subordinates. Speech acts have been defined as "basic or minimal units of linguistic communication . . . where language takes the form of 'rules governed, intentional behavior'" (Searle, 1969, p. 16). Sullivan (1988) has taken the three main speech act classifications, locutionary, perlocutionary, and illocutionary, and interpreted them as follows:

1. Locutionary or meaning-making language occurs when a leader's language to a member explains the structure, rules, and values of the organization's culture. Most significant, locutionary language stimulates the member's cognitive schema to incorporate cultural norms. Frequently, meaning-making language is indirectly transmitted with metaphorical stories and rumors (Cooke & Rousseau, 1988). For instance, the phrase "Even the president pays for lunch in the company cafeteria" could be interpreted as "Freeloading is unacceptable here."

2. Perlocutionary language gives direction and reduces uncertainty. An example of perlocutionary language is assignment clarification from leader to employee.

3. Illocutionary language is empathetic, an expression of humanity, where the leader is willing to share emotion with a member. An example of an illocutionary speech act occurs when a leader praises a worker for a job well done.

It is important to note that MLT only explains subordinate responses to superior-initiated language and not the counterpart, i.e., comparable superior responses to subordinate-initiated language. In fact, a number of studies investigating "semantic information differences" or perceptual incongruities in communication between superiors and subordinates suggest that the gap is often substantial (Jablin & Krone, 1994; Schnake, Dumler, Cochran, & Barnett, 1990).

Motivating language theory incorporates a few more basic assumptions. First, language is assumed to cover most verbal expressions that can occur in leader-to-worker talk, just as speech acts do. Secondly, the effect of motivating language on worker outcomes will be moderated by leader behavior in the majority of cases. Sullivan expressly hypothesized that the effect of motivating language is contextual, with leader influence serving as the intervening variable. He also advised that future theory operationalization should incorporate measures of leader behavior into its research design (Sullivan, 1988; Sullivan, personal communication, March 3, 1992). Furthermore, research suggests that behavioral cues tend to dominate when actions and language are conflicting (Dulek & Fielden, 1990; Goffman, 1959; Ober, 1992). Simply put, communication will only get a leader so far, and speech must be congruent with behavior to be taken seriously over time.

Finally, a leader will obtain maximum benefit through powerful use of all three speech acts, though weakness in one area can be offset by strength in another. In a methodological framework, ML was theorized as a regression. model.

As yet, no measure has been developed that can be used to assess the extent to which managers use all three speech acts, an essential step in testing MLT. We will attempt to bridge this gap by developing and analyzing a motivating language scale. To support these objectives, we will explain scale validation steps including factor analysis, reliability tests, and tests for convergent/divergent validity.

Methods

Scale Development and Research Design

Initial scale development steps, including the literature review and generation of items, strictly adhered to relevant theory and widely accepted psychometric principles (DeVellis, 1991). To start, we reviewed a nomological net of existing communication and organizational literature for both theory development and question construction (Cooper, 1989; Hunter, Schmidt, & Jackson, 1982). After preliminary generation, items were reviewed by professors of communication and management and two upper-level corporate managers. Outside reviewers then provided feedback on scale clarity, congruence of items within each subscale, overall coherence of the scale, and face validity. We used feedback to improve the questionnaire. After these changes were implemented, we began testing the scale's empirical validity.

We used diverse strategies to examine ML scale validity. As with any new measure, only future replications can rigorously test validity (Kuhn, 1962; Lynch, 1982; Lynch, 1983; McGrath & Brinberg, 1983; Popper, 1965). However, within this exploratory phase of study, scale validity was assessed to the fullest extent possible (Calder, Phillips, & Tybout, 1981, 1982; Calder, Phillips, & Tybout, 1983; McGrath & Brinberg, 1983). To fulfill this objective, the motivating language scale's validity was tested through the following steps: confirmatory factor analysis and item deletion, reliability analysis, and convergent/divergent validity testing (DeVellis, 1991).

The first step in MLS validity examination was testing for congruence between observed and hypothesized factor structures (Sullivan, 1988). All factors with an eigenvalue greater than I were retained (Kim & Mueller, 1978a). Both orthogonal and oblique rotations were used, and any meaningful differences were noted.

If the observed factor structure matched theory, then poorly loading items would be deleted from the scale. A question would be retained only if it had a factor loading of .70 or greater. A loading of this magnitude would indicate a minimum of 49% shared variance between indicant and factor. An item would also be rejected if any other factor loading was greater than .50 (or 25% shared variance). If the latter case occurred, the question was not considered to load cleanly on one factor, hence the indicant's meaning would be ambiguous (Kim & Mueller, 1978b).

ML reliability was assessed after the poorly loading items were deleted. We felt that the motivating language scale should have a high level of reliability since this measure sets upper bounds on scale validity. Based on this relationship, we chose a minimum reliability of .80 for the MLS and its subscales. We also tested the scale's multivariate (total) and univariate (for each subscale) reliabilities due to ML's multivariate nature. In addition, univariate and multivariate reliabilities were assessed via Cronbach's alpha tests and the omega statistic, respectively (Zeller & Carmines, 1980).

High reliability of other scales used in this study was not as critical because these scales had established records of both validity and reliability (Hoppock, 1935; Scandura & Graen, 1984). Nevertheless, the exploratory nature of this research required reliability evaluation for these scales to screen for potential problems. Low scale reliability, in this research setting, would make nonsignificant results difficult to interpret. Therefore, a minimum reliability of .60 was required for all scales used to test convergent or divergent validity (Churchill, 1979).

Next, we examined convergent and divergent validities through LISREL analysis. A multivariate method of analysis is required to test for differential validity since ML is theorized to be a multivariate measure of leader communication. Sullivan's (1988) model predicts that the three latent components of motivating language are actually manifest from a second order factor of a leader's ability to strategically use language (also called motivating language by Sullivan). In turn, this second order factor should be positively related to, but distinct from subordinate communication satisfaction.

Proper use of convergent and divergent validity can provide evidence that motivating language use is in fact different from a subordinate simply being happy with the way his or her boss communicates. Bacharach (1989) stated that differentiation between two constructs can be demonstrated by a divergent relationship with a third construct. For this study, we chose subordinate communication competence as the divergent variable. This construct is significantly related to communication satisfaction at the .01 level (r = .25), but should have no significant relationship with a leader's motivating language use (Monge, Bachman, Dillard, & Eisenberg, 1982; Sullivan, 1988).

Confidence in scale validity would be supported if LISREL analysis showed a significant relationship between motivating language use and communication satisfaction, and no significant relationship between motivating language use and communicator competence. The LISREL model used to test convergent and divergent validity is presented in Figure 1.

Sample and Procedures

We collected data in two stages. The first stage was primarily intended to aid in determining internal validity of the motivating language scale. At this stage the ML questionnaire was given to a group of adult students taking a management information systems class at a mid-sized eastern university. All students classified themselves as having full-time work experience and as working toward B.S. degrees. The respondents produced fifty usable surveys.

We solicited feedback from this group on the clarity of the ML scale and for initial reliability tests. The majority of respondents worked in first-level line or staff jobs and represented a wide range of organizational experiences and ages. For example, a number of respondents were also in supervisory and managerial positions. In addition, respondents' ages ranged from 19 to mid 40s. Racial composition was overwhelmingly Caucasian, and this sample was approximately 40% female and 60% male. Since reliabilities were in the .90 range and feedback was very positive, no changes were made to the scale, and this sample was included with the second sample group for factor analysis and reliability tests.

The second data collection stage began by administering relevant questionnaires to a sample of supervisors and their subordinates in the nursing department of a large public hospital. To test convergent and divergent validity, data were collected from supervisor and subordinate pairs. Subordinates were asked to fill out a questionnaire rating their supervisor's skill in using motivating language (the ML scale) and their own satisfaction with their supervisor's communication (Putti, Aryee, & Phua, 1990). As a comparison, supervisors were asked to rate their subordinates' communication competence.

A usable n of 151 supervisor-employee paired respondents was obtained. This sample size provided an adequate n for statistically robust testing (Hair, Anderson, Tatham, & Grablowsky, 1987). Job classifications within this department ranged from upper managerial positions (department heads) to hourly workers (janitorial staff).

Relevant sample demographics can be summarized as the following descriptive statistics. The majority of subordinate respondents were female (68.9%), between the ages of 26 and 55 (with the highest occurrence in the 36-45 age group, 38.8%), and college educated (65.4%). The supervisors were also predominantly female (84.6%) and were generally older and better educated than the subordinates. The most frequent age range was 45-55 years (61.5%), with 61.5% of managers holding master's degrees.

Measures

In addition to the motivating language scale, two other scales - the communicator competence scale and a communication satisfaction with immediate superior scale - were used to test external validity. (All scales are reprinted in Appendix 1.)

Of these, the communicator competence scale developed by Monge, Bachman, Dillard, and Eisenberg (1982) was used to test divergent validity. To test convergent validity, we used the satisfaction with immediate superior communication subscale from the International Communication Associations Organizational Communication Relationship questionnaire (ICA-OCR). This subscale was chosen because of the ICA-OCR's wide acceptance among researchers, high levels of reliability (Putti, Aryee, & Phua, 1990), and proven differentiation from other cognitive constructs, such as job satisfaction (Pincus, 1986).

Results

The initial evaluation of the motivating language scale is quite encouraging. Overall, factor analysis results indicate a strong and stable factor structure congruent with theory. Both oblique and orthogonal rotations produced the same factor structure and similar loading patterns. Therefore, to remain congruent with theory, we used orthogonal rotation for validation purposes. As predicted, three factors have an eigenvalue greater than 1, and these three factors account for more than 75% of total data variance. Sixteen of the original 24 questions met retention standards (as presented in the Sample and Procedures section). Remaining questions, that did not meet minimum factor loading requirements, were excluded from analysis. All retained items cleanly loaded on the expected factors. (Scale questions, along with their factor loadings, eigenvalues of the factors, and percentage of variance explained by factors, are presented in Table 1.)

Next, we tested scale reliability. ML subscale reliabilities are high, ranging from .97 for empathetic language to .92 for direction-giving. Overall, the motivating language scale has a multivariate reliability (omega score) of .97. These results indicate a high degree of stability for ML and are well above the level expected for even established measures (Churchill, 1979). The other scales appear to have adequate reliability levels for their use in this study. The reliability scores are .91 for subordinate's communication competence and .95 for communication satisfaction. These reliabilities are reported in Table 2, along with scale means, standard deviations, and correlations.

The final step in evaluating the MLS was testing convergent and divergent validity. LISREL results supported the convergent and divergent validity of the motivating language scale. T-test statistics showed that the motivating language second order factor has significant and positive loadings with subordinate communication satisfaction and the three latent ML factors, but it has no significant relationship with communicator competence (Arkin & Colton, 1962; Gonick & Smith, 1993). Also, the link between communicator competence and motivating language is less than half of the next highest loading, and one-fourth of the loading between motivating language and communication satisfaction. The relevant loadings and t-values are presented in Table 3.

[TABULAR DATA FOR TABLE 1 OMITTED]

The LISREL analysis also shows that data from this study have an adequate fit with the theorized motivating language scale. The chi-square analysis, a measure of the model's overall goodness of fit, is 624 with 1,375 degrees of freedom. While the p-value is less than .01, Joreskog and Sorbom (1989) caution that the p-value should not be the final arbiter of a model's goodness of fit. Rather, they suggest that researchers should examine the ratio of the degrees of freedom to the chi-square statistic. As this ratio is approximately 2.20, the model's fit can be considered to be well within the acceptable range for the initial scale development process (Wheaton, Muthen, Alwin, & Summers, 1977).

[TABULAR DATA FOR TABLE 2 OMITTED]

Discussion and Implications

Results from data analysis support the reliability and validity of the motivating language scale. Development of this measure has meaningful contributions for both research and practical applications. For researchers, the MLS opens up new avenues for the study of leadership. MLT describes a role for spoken language that is potentially richer than most roles found in traditional leadership theories. For academics and managers alike, motivating language has a strategic, goal-oriented focus that should be evident in measurable outcomes. Of course, realization of these implied benefits hinges on future research.

Table 3

Standardized Parameter Estimates and t-Values for Motivating
Language and Latent Factors

                   Parameter Estimates       t-values

Communication
Satisfaction             .858                10.928(*)
with Superior

Communicator
Competence               .235                 2.56

Direction-Giving
Language                 .915                12.313(*)

Empathetic               .950                 6.576(*)
Language

Meaning-Making
Language                 .593               14.173(*)

* Significant at the .01 level

Researchers can use the MLS to test and operationalize the motivating language theory model. In particular, the MLS can be used to investigate the predicted positive relationship between motivating language and key dependent variables such as performance and job satisfaction. Subsequent studies can also measure the contextual relationships between motivating language, desirable outcomes, and appropriate leader behavior models that incorporate variance into leader-subordinate communication hypotheses.

Theory testing could also lead to important organizational applications such as training managers in goal-oriented language use and communicating planned cultural change. These benefits will most likely be established as a result of future longitudinal or experimental studies. Sullivan (1988) advocated this kind of inquiry for scale refinement and identification of optimal levels of motivating language. For example, comparison groups could be used to measure the training intervention effects of partial ML use (one or two of the essential speech acts) against the presence of the "full construct" (all three fundamental types of language).

Well-designed and longitudinal tests of MLT could also help identify the degree of generalizability and potential moderators. For instance, situations with a demographically biased group, such as a high percentage of one gender, could significantly alter the influence of motivating language on targeted results. Finally, future studies can help to define optimal usage of ML, both at the cumulative (all three speech acts) and subscale levels.

In conclusion, we have only begun to explore the strategic use of managerial language. Much remains to be done. The next steps are model testing and scale refinement. If the theory is supported, then additional research should investigate generalizability, the contextual influences of behavior and culture, and the effectiveness of motivating language training programs.

NOTES

The authors give special thanks to Jim Cashman and Ron Dulek at the University of Alabama and Jeremiah Sullivan at the University of Washington for their extensive contributions to this paper.

Jacqueline and Milton Mayfield share first authorship of this article. Jacqueline is an assistant professor of business communication and management. Milton is an assistant professor of research methodology and management. Both are in the College of Business Administration and International Trade at Texas A & M International University. Address correspondence to J. Mayfield or M. Mayfield at the College of Business Administration and the Graduate School of International Trade and Business Administration, Texas, A & M International University, 1 West End Washington Street, Laredo, TX 78040-9960.

Jerry Kopf is an associate professor of management and director of the Small Business Institute in the College of Business and Economics at Radford University, Radford, VA.

REFERENCES

Arkin, H., & Colton, R. (1962). Tables for statisticians, New York: Barnes & Noble.

Bacharach, S. (1989). Organizational theories: Some criteria for evaluation. Academy of Management Review, 14, 494-515.

Bass, S. (1990). Handbook of leadership. New York: Free Press.

Calder, B, Phillips, L., & Tybout, A. (1981). Designing research for application. Journal of Consumer Research, 8, 197-207.

Calder, B., Phillips, L., & Tybout, A. (1982). The concept of external validity. Journal of Consumer Research, 9, 240-243.

Calder, B, Phillips, L., & Tybout, A. (1983). Beyond external validity. Journal of Consumer Research, 10, 112-114.

Churchill, G. (1979). A paradigm for developing better measures of marketing constructs. Journal of Marketing Research, 16, 64-73.

Conger, J. (1991). Inspiring others: The language of leadership. Academy of Management Executives, 1, 31-45.

Cooke, R. A., & Rousseau, D. (1988). Behavioral norms and expectations: A quantitative assessment of organizational culture. Group & Organizational Studies, 13, 245-273.

Cooper, H. (1989). Integrating research: A guide for literature reviews (2nd ed.). Newbury Park, CA: Sage.

Daft, R. L., & Wiginton, J. C. (1979). Language and organization. Academy of Management Review, 4, 179-191.

Dansereau, F., Cashman., J., & Graen, G. (1973). Instrumentality theory and equity theory as complimentary approaches in predicting the relationship of leadership and turnover among managers. Organizational Behavior and Human Performance, I0, 184-200.

DeVellis, R. F. (1991). Scale development. Newbury Park, CA: Sage.

Dulek, R., & Fielden, J. (1990). Principles of business communication. New York: Macmillan.

Fairhurst, G. T. (1993). The leader-member exchange patterns of women leaders in industry. Communication Monographs, 60, 260-278.

Fairhurst, G. T., & Chandler, T. A. (1989). Social structure in leader-member interaction. Communication Monographs, 56, 215-239.

Goffman, E. (1959). The presentation of self in everyday life. New York: Doubleday.

Gonick, L., & Smith, W. (1993). The cartoon guide to statistics. New York: Harper-Collins.

Graen, G., & Cashman, J. (1975). A role-making model of leadership in formal organizations: A development approach. In J. G. Hunt & L. L. Larson (Eds.), Leadership frontiers (pp. 143-165). Kent, OH: Kent State University Press.

Graen, G., & Scandura, T. A. (1987). Toward a psychology of dyadic organizing. Research in organizational behavior, 9, 175-208.

Hair, J., Anderson, R., Tatham, R., & Grablowsky, B. (1987). Multivariate data analysis with readings. New York: Macmillan.

Hoppock, R. (1935). Job satisfaction. New York: Harper & Row.

Hunter, J., Schmidt, E, & Jackson, G. (1982). Meta-analysis: Cumulating research findings across studies. Beverly Hills, CA: Sage Press.

Jablin, F. (1985). Task/work relationships: A life-span perspective. In M. L. Knapp & G. R. Miller's (Eds.), Handbook of interpersonal communication (pp. 615-654). Beverly Hills, CA: Sage Press.

Jablin, F., & Krone, K. (1994). Task/work relationships: A life-span perspective. In M. L. Knapp & G. R. Miller's (Eds.), Handbook of interpersonal communication (2nd ed.), (pp. 621-675). Beverly Hills, CA: Sage Press.

Joreskog, K., & Sorbom, D. (1989). Lisrel 7 user's reference guide. Mooresville, IN: Scientific Software.

Kim, J., & Mueller, C. W. (1978a). Introduction to factor analysis. Newbury Park, CA: Sage.

Kim, J., & Mueller, C. W. (1978b). Factor analysis statistical methods and practical issues. Newbury Park, CA: Sage.

Kuhn, T. (1962). The structure of scientific revolutions. Chicago: University of Chicago Press.

Lamude, K., Daniels, T., & Graham, E. (1988). The paradoxical influence of sex on communication rules coorientation and communication satisfaction in superior-subordinate relationships. Western Journal of Speech Communication, 52, 122-134.

Lynch, J. (1982). On the external validity of experiments in consumer research. Journal of Consumer Research, 9, 225-239.

Lynch, J. (1983). The role of external validity in theoretical research. Journal of Consumer Research, 10, 109-111.

McGrath, J., & Brinberg, D. (1983). External validity and the research process: A comment on the Calder/Lynch dialogue. Journal of Consumer Research, 10, 115-124.

Monge, P. R., Bachman, S. G., Dillard, J. P., & Eisenberg, E. M. (1982). Communicator competence in the workplace: Model testing and scale development. In M. Burgoon (Ed.), Communication Yearbook 5 (pp. 505-527). Newbury Park, CA: Sage.

Ober, S. (1992). Contemporary business communication. Boston: Houghton Mifflin.

Pincus, D. (1986). Communication satisfaction, job satisfaction, and job performance. Human Communication Research, 12, 395-419.

Popper, K. (1965). Conjectures and refutations: The growth of scientific knowledge. New York: Basic Books.

Putti, J., Aryee, S, & Phua, J. (1990). Communication relationship satisfaction and organizational commitment. Group and Organization Studies, 15, 44-52.

Scandura, T., & Graen, G. (1984). Moderating effects of initial leader-member exchange on the effects of a leadership intervention. Journal of Applied Psychology, 69, 428-436.

Schnake, M., Dumler, M., Cochran, D., & Barnett, T. (1990). Effects of differences in superior and subordinate perceptions of superiors' communication practices. Journal of Business Communication, 27, 37-50.

Searle, J. R. (1969). Speech acts. London: Syndics of the Cambridge University Press.

Sullivan, J. (1988). Three roles of language in motivation theory. Academy of Management Review, 1, 104-115.

Waldron, V. R. (1991). Achieving communication goals in superior-subordinate relationships: The multi-functionality of upward maintenance tactics. Communication Monographs, 58, 288-305.

Wheaton, B., Muthen, B., Alwin, D. R., & Summers, A. F. (1977). Assessing reliability and stability in panel models. In D.R. Heise (Ed.) Sociological Methodology (pp. 84-136). San Francisco: Jossey-Bass.

Yukl, G. (1989). Managerial leadership: A review of theory and research. Journal of Management, 15, 251-289.

Zeller, R., and Carmines, E. (1980). Measurement in the social sciences: The link between theory and data. New York: Cambridge University Press.