Measure Validation in Consumer Research: a Confirmatory Factor Analysis of the Voluntary Simplicity Lifestyle Scale

Deborah Cowles, Arizona State University
Lawrence A. Crosby, Arizona State University
ABSTRACT - The development of a measurement tradition in consumer research will necessitate a stronger emphasis on replication. The validity of standardized instruments must be established through repeated application of scales in different contexts and among different population groups. Leonard-Barton's Voluntary Simplicity Lifestyle Scale was administered to a sample of California and Colorado voters. Examination of the factorial validity of the scale supports a three dimensional structure which is based on theory. The values constructs of material simplicity, self-determination, and ecological awareness adequately account for the observed inter-item correlations
[ to cite ]:
Deborah Cowles and Lawrence A. Crosby (1986) ,"Measure Validation in Consumer Research: a Confirmatory Factor Analysis of the Voluntary Simplicity Lifestyle Scale", in NA - Advances in Consumer Research Volume 13, eds. Richard J. Lutz, Provo, UT : Association for Consumer Research, Pages: 392-397.

Advances in Consumer Research Volume 13, 1986      Pages 392-397

MEASURE VALIDATION IN CONSUMER RESEARCH: A CONFIRMATORY FACTOR ANALYSIS OF THE VOLUNTARY SIMPLICITY LIFESTYLE SCALE

Deborah Cowles, Arizona State University

Lawrence A. Crosby, Arizona State University

ABSTRACT -

The development of a measurement tradition in consumer research will necessitate a stronger emphasis on replication. The validity of standardized instruments must be established through repeated application of scales in different contexts and among different population groups. Leonard-Barton's Voluntary Simplicity Lifestyle Scale was administered to a sample of California and Colorado voters. Examination of the factorial validity of the scale supports a three dimensional structure which is based on theory. The values constructs of material simplicity, self-determination, and ecological awareness adequately account for the observed inter-item correlations

INTRODUCTION

Consumer behaviorists have exhibited a renewed interest in the role that research plays in the theory development process. Increased attention to the relationship between research and theory has spawned "considerable interest in methodological issues related to theoretical constructs" (Brown and Gaulden 1980, p. 241). Perhaps foremost among these concerns is the apparent lack of a tradition of measurement research and instrumentation in the marketing and consumer behavior disciplines. Bagozzi (1980) has observed that while marketers readily acknowledge in their journals the importance of measurement, they seldom examine in these same journals the conceptual underpinnings of measurement procedures and relate them to the purposes for which they were constructed. This indifference is remarkable, considering that measurement and theory "are inseparably bound in a logical sense" (Bagozzi 1980). Theory development requires theory testing. More than a decade ago, Reeler and Ray (1972) observed that tests of existing theories in marketing and consumer behavior have been marred because they were not preceded by adequate measure validation.

Some authors have pointed to replication as a way to solve at least some of the measurement problems plaguing marketing and consumer research, including the prevailing "disregard" (Ray 1979, p. l) for measure validation. Consumer behaviorists, in particular, have been vocal in their call for replication (Jacoby 1978; Brown and Coney 1976; Engel, Blackwell and Kollat 1978), on the grounds that spurious conclusions hamper theory development. "There is a strong necessity for replicating our findings using different subject populations, test products, etc. The name of the game is confidence in our findings" (Jacoby 1978, p. 93). Replication lies at the heart of generalization of any body of knowledge (Kollat, Engel and Blackwell 1970). Generalizability, in turn, is fundamental to theory's capability to predict phenomena.

MEASUREMENT VALIDATION IN PSYCHOLOGY

Not all of the social sciences have been so neglect in building a foundation of replication and instrumentation. In contrast with marketing, psychology can boast a long tradition of measurement research. Indicative of this heritage is the science T S regular publishing of compendiums of measurement tools. The Eighth Mental Measurements Yearbook (Buros 1978), for example, is a lengthy, two-volume collection of measurement instruments covering a wide range of topics (achievement batteries, multiple-aptitude batteries, personality, intelligence, education, etc.). Other similar repositories for published tests and measurement tools exist in the field of psychology (Straus and Brown 1978, Johnson 1976). Compendiums like these serve to facilitate measurement validation, as well as other aspects of measurement research, via exact replication and related studies.

Scholars like Frederic M. Lord of the Educational Testing Service have devoted most of their lives to the study of mental tests (Wainer and Messick, 1983). That is not to say, however, that these scholars have solved all the problems associated with psychological measurement, in particular the issue of measurement validation. "The determination of validity . . . is not so much a process as it is a program. The measurement of the psychological properties of individuals is a complicated affair, to be accomplished laboriously, and seldom, if ever, with the degree of precision desired" (Ghiselli 1964, p. 368).

Nevertheless, scholars in the field of psychology pursue this "broad problem area" (Ghiselli 1964, p. 368), evidently with a commitment to precision not known co the marketing discipline as a whole. Even a cursory review of the major psychology journals at any time would produce more than a handful of perfect and imperfect replications-tests of established measurement instruments, across time, various population groups and settings--designed to examine the validity of those tools (e.g., Gahar and Bale 1982; Ibrahim 1982; and LaFromboise 1983). Marketing and consumer behavior Journals are all but void of such examinations. Calls for replications, lamentations over the sorry state of measure validation in marketing, and suggestions with respect to how one can improve measurement validation appear more frequently in the marketing literature.

IMPROVING MEASURES OF CONSUMER BEHAVIOR CONSTRUCTS

Psychometric procedures for developing better measures of marketing and consumer behavior constructs have been well described by Churchill (1979) and do not require restatement. Worthy of note, however, are two of the principles which underlie these procedures. One principle is that most measures of consumer behavior phenomena, and indeed the phenomena themselves, are seldom uni-dimensional. This means that in developing standardized instruments, correspondence rules linking theoretical dimensions to test components and subscales must be clearly established. This can be done only if the theory is sufficiently explicit about the definition and domain of the construct. Care must also be taken not to segregate the conceptual stage of the research from its test (Bagozzi 1980). The specification of theoretical constructs should include a measurement model, and it is this model which should be empirically tested. The initial concern is usually to verify the factorial validity of the proposed measure, with convergent and discriminant validity being established later.

Another principle is that measure validation cannot be accomplished in a single study. As in all research, apparent regularities involving measures and measurement items may be the result of chance or spurious relationships. There is also a need to establish that the measurement model holds across the full range of the underlying variables. If measures are found to behave differently when tested across time and across population groups, this suggests that either the measure is invalid or the concept, poorly understood. It is possible that a measurement instrument taps different constructs when applied at different times or to different population groups. Thus, repeated tests of the factorial, convergent, and discriminant validity of tests need to be made.

These two important principles are best illustrated with a specific example from the consumer behavior literature.

VOLUNTARY SIMPLICITY LIFESTYLE SCALE

An examination of the Voluntary Simplicity Lifestyle (VSL) Scale (Leonard-Barton 1981) is especially appropriate for the purposes of this paper for a number of reasons. First, the scale has not been finalized; Leonard-Barton admits potential scale shortcomings and has suggested a number of areas that may need improving. Thus, the current research shows the value of measurement research to the theory development process. Second, Leonard-Barton's scale represents only one of a number of ways that have been suggested to model voluntary simplicity (VS) (Ensley 1983).

Despite the potential value of this scale and other similar instruments, some scholars have questioned the validity of lifestyle concepts and measures in general (Mehrota and Wells 1977; Wells 1975; Wells and Cosmas 1982). Lastovicka observed that while there is an "overabundance of lifestyle traits" (1982, p 138), there have been few attempts in the literature to examine the validity of the proposed traits. An examination of the scale (ant the theoretical model it implies) enables one to consider more closely the relationship between theory and research, specifically, the relationship between theoretical constructs and the tools that are supposed to measure them.

The use of the VSL scale to explore the value of measurement validation to consumer research is appropriate also because the scale focuses on extremely dynamic characteristics of our society. Marketers more often than not face the challenges of a fast-changing marketplace while conducting research. Finally, the development of not only a theory (model) to explain voluntary simplicity, but also a measurement tool (scale) to predict (identify) the lifestyle are potentially matters of great concern to marketers. "Growth in the number of adherents to VS (voluntary simplicity) could dramatically affect current marketing practices" (Ensley 1983, p. 385).

VOLUNTARY SIMPLICITY: THE CONCEPT

An in-depth review of the VS literature is not required for the purposes of this paper. Recent reviews (Ensley 1983, Leonard-Barton 1981, Leonard-Barton and Rogers 1980) report agreement among scholars as to the building blocks of the lifestyle: material simplicity (nonconsumption-orientation patterns of use); human scale (a desire for small-scale institutions and simpler technologies); self-determination (desire for greater control over personal destiny); ecological awareness (recognition of the interdependency of people and resources); personal growth (a desire to explore and develop the "inner life")

Scholars do not agree, however, about what comprises the best method for identifying voluntary simplifiers. The measurement tool which is the focus of this article, developed by Leonard-Barton (1981), is essentially a behavioral scale. Leonard-Barton maintains that while certain "humanistic values" are an important part of the concept, "as any student of human behavior knows, there is often a large gap between an attitude and an act. . . Behaviors are, therefore, probably better indicators of public support for a voluntary simplicity lifestyle than verbal responses to survey questions" (p. 244).

The scale, which employs survey methodology, asks respondents to report the frequency with which they engage in behaviors that are "typical of self-proclaimed advocates of this scaled-down lifestyle and that were also suggested in literature on the topic . . ." (Leonard-Barton 1981, p. 244). Lastovicka (1982) has criticized the reliance of lifestyle research on self-report data as a "serious problem in lifestyle trait measurement" (p. 138). Another potential problem in the development of a behavioral scale that would be usable over time is the rate of diffusion of behaviors included in the scale. Table l shows that the so-called VSL behaviors have diffused considerably in just a few years. While the percentage of respondents giving a positive response in each category increased as much as 47 percent, it is unlikely that the number of voluntary simplifiers has increased to the same extent. In short, the behaviors do not necessarily indicate VS. Leonard-Barton (1981) defends her approach, maintaining:

"an individual's high score on any one of these six factors by itself does not indicate an interest in voluntary simplicity . . . However, if an individual engages in many of the 18 behaviors . . . then we may assume that some sort of coherent (although often unrecognized) philosophy underlies these diverse acts" (p. 246).

But Ensley (1983) asserts that VS "is essentially an attitude, and adherents emphasize that no particular behavior is required" (p. 388).

TABLE 1

ITEMS IN THE VOLUNTARY SIMPLICITY INDEX IN ORDER OF RATE OF ADOPTION

The Behavioral Scale

The scale developed by Leonard-Barton (1981, p. 244) "evolved through three stages, each revision being tested on a different California population. All three versions of the scale were factor analyzed; factors that emerged were reported by Leonard-Barton to be "quite robust" for all versions, across samples. Leonard-Barton characterized the six factors that emerged in the data collected from the most recent version (n=812 California homeowners) as: conservation through biking; self-sufficiency in services; recycling resources (metals, glass); self-sufficiency through making goods; recycling of durable goods (clothes, furniture); and closeness with nature.

"As Cattell (1961) and others demonstrate, factor analysis is the primary tool for systematically identifying and measuring (lifestyle) traits" (Lastovicka 1982, p. 126). Nevertheless, factors and traits are not the same. A factor is linked to a trait via a correspondence rule. In the case at hand, Leonard-Barton has employed exploratory factor analysis to identify six dimensions which are given a trait-like interpretation. While Leonard-Barton does not specifically use the word "trait", her discussion of the factors is consistent with the accepted definition of traits as "generalized action tendencies" (Allport 1966, p. 3). Figure l shows the implied relationship between scale items, resultant factors, and VSL. While the author acknowledges that at least two of the five "building blocks" of VSL are not represented (human scale and personal growth), no explanation is proffered concerning the conceptual meaning of the six empirically derived factors/traits.

Throughout the social sciences, there is growing recognition that the approach of exploratory factor analysis is seriously flawed. To quote Long (1983, p. 12), "The exploratory factor model's inability to incorporate substantively meaningful constraints, and its necessary imposition of substantively meaningless constraints, has earned it the scornful label of garbage in/garbage out (GIGO) model." The use of exploratory factor analysis in the Leonard-Barton studies seems particularly unfortunate given there is fairly specific theory regarding the value underpinnings of VSL. Leonard-Barton can be criticized for failing to integrate her conceptualization with the empirical test. Figure 2 presents an alternative VSL model that is more theory-based. Behavioral constructs which comprise the alternative model are not the result of factor analysis, rather they are suggested by VSL theory as presented in the literature. According to this new conceptualization, individual ideas in the Leonard-Barton scale should be viewed as indicators of behavioral tendencies which are values driven. The undersampling of some behavioral constructs (i.e., two "building blocks" are not represented) and the oversampling of others can be directly traced to the decoupling of the theory and its test.

FIGURE 1

CONFIRMATORY FACTOR ANALYSIS (CFA) MODEL IMPLIED BY VOLUNTARY SIMPLICITY INDEX (Leonard-Barton 1981)

FIGURE 2

PORTION OF ALTERNATIVE MODEL TESTED DURING CURRENT RESEARCH

Despite the above-mentioned problems, Figures 1 and 9 provide competing motels or hypotheses concerning the structure of VSL. These models were compared, insofar as data would permit, in terms of their ability to account for the observed inter-item correlations in new data. The data were obtained as part of a voting behavior study which included Leonard-Barton's VSL scale as one of its measures.

METHODOLOGY

Data

Voter, legislator, and consumer behavior concerning "bottle bills" is an environmental issue of current public and research interest. The data for the analysis came from a study of voter behavior concerning bottle bills appearing on the November 2, 1982 ballots in California and Colorado. A quota sample of consumer panel households in these states was supplied by a large field research organization. Households were selected for inclusion by matching household characteristics with known census data for each state. A random selection procedure was then used to select an eligible adult voter within each household.

The questionnaire, which was preceded by a pre-alert postcard, was mailed to l,000 panel members, 500 in each state. The Leonard-Barton VSL scale was one of several measurement instruments included in the questionnaire. A standard inducement was offered to panel respondents. A total of 662 surveys were returned (66.2 percent response rate); 424 of the returned questionnaires included complete responses to all VSL scale items.

Analysis

The VSL models portrayed in Figure l and 2, suggest that voluntary simplicity is a multidimensional unobserved, or latent, variable the presence of which to varying degrees can be measured by specific behaviors. Confirmatory factor analysis via the computer program LISREL (Joreskog and Sorbom 1977; Sorbom 1974) was used to assess the factorial validity of the models for a population other than those from which Leonard-Barton had developed the scale. Technically, these are both higher-order factor models which are estimated by Submodel #3 in LISREL. The endogenous latent variables (n1, n2 . . .) are assumed to be partial functions of a global tendency toward VSL. Errors in the equations for these latent variables (41, 42 . . .) acknowledge there are other unspecified influences on the VSL responses. The fitting function for unweighted least squares (ULS) was used to obtain parameter estimates. The ULS estimator has been shown to be consistent without making assumptions about the distribution of the variables (Bentler and Weeks 1980). Estimation via ULS "involves minimizing the sum of squares, in much the same way that ordinary or unweighted least squares for regression analysis minimizes the sum of the squared residuals" (Long 1983, p. 57). Because ULS estimators are scale-dependent the scales of the observed variables were standardized by analyzing a correlation matrix

Long (1983) notes that while it can be advantageous not having to make distributional assumptions about the observed variables, a major limitation is that "there are no statistical tests associated with ULS estimation of the confirmatory factor model" (p. 57). One must rely on an overall goodness-of-fit index, as well as an examination of parameter values, variances and covariances. Despite this limitation, a major advantage to using the confirmatory factor analysis approach over the exploratory factor analysis approach is that "a hypothesized factor structure can be specified as a pattern of zero and nonzero factor loadings. The adequacy of the model (and the theory for which it stands) as representation of the original data matrix can then be empirically assessed" (Aneshensel et al, p. 388). The technique also allows one to compare the hypothesized model with alternative specifications of the model in which parameter constraints are modified, a most important feature for theory builders.

RESULTS

The confirmatory factor model suggested by the Leonard-Barton research (1981) and shown in Figure l, was tested to determine its tenability. Goodness-of-fit indices and ULS parameter estimates for a modified version of that model are found in Table 2. Modifications to the original model, which were indicated by LISREL program output, were made within the bounds of practicality and theory. These modifications took the form of an allowance for some correlated error among the indicators. The results show that the Leonard-Barton model is relatively robust for this population, with an adjusted goodness of fit of .956. Table 3 provides goodness-of-fit indexes and ULS parameter estimates for a modified version of the alternative model, which is at least as robust as the model suggested by Leonard-Barton. The adjusted goodness of fit index for the alternative model was .962. Table 4 compares the two models' estimates of indicator and construct reliability.

DISCUSSION

Given the model in Figure 2 is incomplete, yet performs as well as the Leonard-Barton model, it would appear to be a better approach to the development of a VSL scale for a number of reasons. First, the constructs, as configured, more accurately reflect VSL theory as it is presented in the literature. Second, as Leonard-Barton points out, the meaning of a specific behavior can change over time and across boundary lines. The assignment of these or other behavioral indicators to theoretical (vs. behavioral) constructs would seem to offer more flexibility in theory/scale development. Finally, the model is more explicit in portraying the relationship between and among variables and constructs, a characteristic of great value to theory builders.

The current research has shown that Leonard-Barton has made a significant contribution to VSL research and theory development. Her scale, which was developed with a California-only population, yielded meaningful results when applied to a different California/Colorado population. However, as Leonard-Barton recommends in her 1981 article, future research should include a "further refinement of the index, including tests for the applicability of items to different geographic locations" (p. 250), and "expansion of the 18-item index to cover interest in holistic health, improved nutrition, and greater personal happiness" (p. 250) - perhaps as a means of addressing the VSL components not included in the scale.

Curiously, however, Leonard-Barton at the same time calls for a streamlining of the scale, and suggests three different ways to achieve this: by factor analysis (selecting the highest loading variables); by multiple regression (dropping those behaviors that explain little variance in the regression line); and by using rate of adoption. Although a more streamline scale is not necessarily indicated, an analysis of the rate of adoption of VSL behaviors, seems appropriate for any VSL research. Indeed, it seems unlikely that researchers will be able to capture the underpinnings of VSL via simplistic models:

. . . recognition of nomological validity as a necessary tool for testing lifestyle trait concepts emphasizes the need for better lifestyle theory. The current elementary theoretical schemes provide settings for simple tests of trait validity, but as nomological validity is more seriously examined these schemes must become more elaborate. (Lastovicka 1982, p. 138)

Although the more complex model with all five VSL building blocks has not been put to a rigorous test, the preliminary analysis of a portion of the model holds promise for future VSL research. Opportunities exist to examine how VSL, and lifestyle traits, in general, relate to the individual's value structure. Also important is the manner in which lifestyles and values combine to influence consumer and voter decision-making in specific choice situations. Finally, returning to the issue of measurement in marketing research, it is clear that research tools available to marketing will only be of value if they can be shown to be reliable and valid. The literature is consistent in its appraisal that the VSL concept is potentially of great importance to marketers (Ensley 1983). If for no other reason than this, the development of a valid and reliable VSL measurement tool based on theory should be an integral part of all future VSL research.

TABLE 2

PARAMETER ESTIMATES (UNWEIGETED LEAST SQUARES) LEONARD-BARTON VSL CONFIRMATORY FACTOR MODEL

TABLE 4

RELIABILITY OF INDIVIDUAL AND COMPOSITE MEASURES (MODIFIED MODELS)

TABLE 3

PARAMETER ESTIMATES (UNWEIGHTED LEAST SQUARES) ALTERNATIVE VSL CONFIRMATORY FACTOR MODEL

REFERENCES

Allport, Gordon W. (1966) "Traits Revisited," American Psychologist, 21 (January), 1-10.

Aneshensel, Carol S., Virginia A. Clark, and Ralph R. Frerichs, (1983) "Race, Ethnicity, and Depression: A Confirmatory Factor Analysis," Journal of Personality and Social Psychology, 44 (2),385-398.

Bagozzi, Richard P. (1980) Causal Models in Marketing, New York: John Wiley & Sons.

Bentler, P.M. and D. G. Weeks (1980) "Linear Structural Equations With Latent Variables," Psychometrika, 45, 289-308.

Brown, Stephen W. and Kenneth C. Coney (1976) "Building a Replication Tradition in Marketing," in Marketing: 1776-1976 and Beyond, Kenneth L. Bernhardt, ed., Chicago: AMA. 622-625.

Brown, Stephen W. and Corbett F. Gaulden, Jr. (1980) "Replication and Theory Development," AMA Proceedings 1980, 240-243.

Buros, Oscar Krisen, ed. (1978) The Eighth Mental Measurements Yearbook. Vol. I and II, Highland Park, M.J.: The Gryphon Press.

Cattell, R.B. (1961) "Theory of Situational, Instrument, Second-Order and Refraction Factors in Personality Structure Research," Psychological Bulletin, 58 (March), 160-174.

Churchill, Gilbert A., Jr. (1979) "A Paradigm for Developing Better Measures of Marketing Constructs," Journal of Marketing Research, 16 (February), 64-73.

Elgin, Duane and Arnold Mitchell (1977), "Voluntary Simplicity: Lifestyle of the Future?" The Futurist, 11 (August), 200-209, 254-261.

Engel, James F., Roger D. Blackwell, and David T. Kollat (1978), Consumer Behavior, Third Edition, Hinsdale, Illinois: The Dryden Press. 579.

Ensley, Elizabeth E. (1983) "Voluntary Simplicity: A Segment of Concern to Marketers?" AMA Proceedings 1983, 385-389.

Gahar, Helen Anna and Robert L. Hale (1982) "Factor Analysis of the Quay-Peterson Behavior Problem Checklist Across Racially Different Exceptional Children," Journal of Psychology, 112, 287-293.

Ghiselli, Edwin E. (1964) Theory of Psychological Measurement, New York: McGraw-Rill Book Co.

Heeler, Roger M. and Michael L. Ray (1972) "Measure Validation in Marketing," Journal of Marketing Research, 9 (November), 361-370.

Ibrahim, Abdul Sattar (1982) "The Factorial Structure of the Eysenck Personality Questionnaire Among Egyptian Students," The Journal of Psychology, 112, 221-226.

Jacoby, Jacob (1978) "Consumer Research: How valid and useful are all our consumer behavior research findings? State of the Art Review," Journal of Marketing, 42 (April), 87-96.

Johnson, Orval G. (1976) Tests and Measurements in Child Development: Handbook II, San Francisco: Jossey-Bass Publishers.

Joreskog, Karl G. and Dag Sorbom (1977) "Statistical Models and Methods for Analysis of Longitudinal Data," in Latent Variables in Socioeconomic Motels, D.J. Aigner and A.S. Goldberger, eds., Amsterdam: North-Holland.

Joreskog, Karl G. and Dag Sorbom (1984) LISREL VI: Analysis of Linear Structural Relationships by Maximum Likelihood, Instrumental Variables, and Least Squares Methods, Mooresville, Indiana: Scientific Software, Inc.

Kollat, David T., James F. Engel, and Roger D. Blackwell (1970), "Current Problems in Consumer Behavior Research," Journal of Marketing Research, 7 (3) 327-332.

LaFromboise, Teresa D. (1983) "The Factorial Validity of the Adult Self-Expression Scale with American Indians," Educational and Psychological Measurement, 43, 547-555.

Lastovika, John L. (1982) "On the Validation of Lifestyle Traits: A Review and Illustration," Journal of Marketing Research, 19 (February), 126-138.

Leonard-Barton, Dorothy (1981) "Voluntary Simplicity Lifestyles and Energy Conservation," Journal of Consumer Research, 8 (December), 243-252.

Leonard-Barton, Dorothy and Everett Rogers (1980), "Voluntary Simplicity," in ACR Vol. VII, J.C. Olsen, Ed., Ann Arbor: Association for Consumer Research.

Long, J. Scott (1983) Confirmatory Factor Analysis: A Prefact to Lisrel, Series: Quantitative Applications in the Social Sciences, Sage University Paper.

Mehrota, S. and W.D. Wells (1977) "Psychographics and Buyer Behavior: Theory and Recent Empirical Findings," in Consumer and Industrial Buying Behavior, A.G. Woodside, J.N. Sheth, and P.S. Bennett, eds. New York: Elsevier North Holland.

Ray, Michael L. (1979) "Introduction to Special Section: Measurement and Marketing Research -- Is the Flirtation Going to Lead to a Romance?" Journal of Marketing Research, 16 (February), 1-6.

Sorbom, Dag (1974) "A General Method for Studying Differences in Factor Means and Factor Structure Between Groups, British Journal of Mathematical and Statistical Psychology, 27, 229-239.

Straus, Murray A. and Bruce W. Brown (1975) Family Measurement Techniques: Abstracts of Published Instruments 1935-1974, Revised Edition, Minneapolis: University of Minnesota Press.

Wainer, Howard and Samuel Messick, eds. (1983) Principles of Modern Psychological Measurement: A Festschrift for Frederic M. Lord, Hillsdale, N.J.: Lawrence Erlbaum Associates. Publishers.

Wells, W.E. (1975), "Psychographics: A Critical Review," Journal of Marketing Research, 12 (May), 196-213.

Wells and S.C. Cosmas (1977), "Life Styles," in Selected Aspects of Consumer Behavior, R. Ferber, ed. Washington, D.C.: National Science Foundation.

----------------------------------------