Advances in Consumer Research Volume 6, 1979 Pages 340-346
MEASURING THE BASES OF SOCIAL POWER
John L. Swasy (student), University of California, Los Angeles
The French and Raven conceptualization of social power and previous operationalizations of the bases of social power are reviewed. The development of an instrument to measure perceived social power is presented.
The topic of social power is quite complex and has been described by many different sociological and psychological theories. Psychoanalytic theory, field theory, exchange theory, as well as the persuasion and attitude change literature all deal with aspects of the influence one person has on another. For the consumer behavior areas of family decision making and attitude change, interpersonal influence processes are of particular importance.
Because social power has been studied from so many perspectives, it's understandable that a wide variety of definitions and operationalizations have resulted. The terms "influence," "power," "decision making," "authority'' and other terms have often been used interchangeably. In the family power structure literature, for example, the wide variety of definitions and measures has led some researchers to question the usefulness of the power concept (Safilios-Rothschild, 7970). A review of the social power literature indicates that one basic problem is the lack of a proven scale for measuring social power.
This study attempts to provide such a scale. Specifically it will review the French and Raven conceptualization of social power, previous operationalizations and then present the development of an instrument to operationalize social power.
Previous works have offered a variety of definitions of social power and influence (Pollard and Mitchell, 1972). A commonly cited definition is that of French and Raven (1959) and Cartwright and Zander (1968) who define influence as "a change in cognition, attitude, behavior or emotion of one person which can be attributed to another agent." Social power is "the potential influence of one person over another" (Cartwright and Zander, p. 316). Thus, power is the total possible change which one social agent can cause in another person's attitude, behavior, beliefs, etc.
This conceptualization follows from the early work of Kurt Lewin (1951), who posited a model of human behavior based upon forces within one's life space. In this model, behavior was explained as the resultant force caused by tensions and needs which originated within (or acted upon) the life space. The social influence of person A over person B was defined in terms of the force A could bring to bear on B and the resistance B could offer. Mathematically Lewin defined the Power of A over B as the quotient of the maximum force that A could (or possibly could) induce on B and the maximum resistance that B could offer. Cartwright (1959) reconceptualized power not as the quotient but as "the maximum strength of the resultant force which A can set in that direction at that time" (Cartwright, 1959, p. 193). Thus, according to this conceptualization, the extent to which agent A has control over B's behavior will depend on the magnitude of the force which he can bring to bear on B and upon B's resistance.
While Lewin's and Cartwright's representation of social power and influence strongly mirror the physical sciences' notion of force vectors, they do include several important psychological aspects. First, for one person to "have power over B" does not necessarily mean that A is having an observable effect on B. If other situational and internal forces balance A's force, then no change will occur. Secondly, it is not necessary for A to act in order to have power over B. B's anticipation of A's reaction is sufficient for A to have power.
While the outcome of an influence exchange between two persons was very important to these early theorists, the process of influence and the reasons for social power were also described. Cartwright notes that two factors determine A's power over B. These are the resources of A and the needs or "motive base" of B. Thus, "an act of A must tap a motive base in order for it to activate a force." (Cartwright, 1959, p. 205) These motive bases were described by French and Raven (1959) and Raven (1965) as being composed of six different types--reward, coercive, referent, legitimate, expert and informational.
Bases of Social Power
The Reward power of an influencer (A) over person B is based on the ability to mediate positive outcomes and to remove or decrease negative outcomes received by B. The strength of A's power is thus dependent on the size of the mediated outcome and the probability that A can mediate the reward. French and Raven also suggest that the amount of influence from this power base will depend on the subjective probability of reward for conformity, minus the subjective probability of obtaining the reward for nonconformity. In addition, with reward power, A's influence is usually limited to those domains where A has the ability to reward. Finally, since A's knowledge of B's compliance is necessary in order for B to be rewarded, B will want to relate his compliance to A and maintain A's surveillance of B's behavior. This surveillance by A is also important for another power base, coercion.
Coercive power rests on B's belief that A will punish him for not complying. The strength of the coercive power depends on the magnitude of punishment times the difference between the probability of punishment for nonconformity and the probability of punishment for conformity.
Often, reward and coercion are viewed as opposite ends of a one-dimensional concept. The primary distinction between these two power bases rests on the influencee's attraction to the influencer. With reward, an increase in attraction should result and lead to a higher dependency on the influencer. However, with coercive power the influencee's attraction toward the influencer should decrease and cause the influencee to avoid further encounters with the influencer.
The third basis for social power is referent power, which results largely from the influencee's feelings of identification with the influencer and desire to maintain similarity with the influencer. [Kelley's "comparison function" (1952) of a reference group is consistent with this power basis. However, the "normative function" of a reference group is analogous to the reward and coercion power bases discussed above. French and Raven note: "...to the extent that 0 (influencer) mediates the sanctions (i.e., has means of control over P (influencee) we are dealing with coercive and reward power, but to the extent that P avoids discomfort or gains satisfaction by conformity based on identification regardless of O's responses we are dealing with referent power." (French and Raven, 1959, p. 162).]
The fourth basis for social power, legitimate power, stems from internalized values of the influencee which dictate that the influencer has the right of influence and that the influencee is obligated to obey. These internalized values may reflect cultural values and norms, group norms, or role prescriptions. The influencee's underlying feelings of "oughtness" toward the compliant behavior and "has the right" toward the influencer characterize this power base.
Because this power base is often derived from many complex societal and personal values, the range of legitimate power A has over B will vary across situations. To the extent that A's legitimate power is derived from a general value or belief of B, A would be expected to have power across many situations. However, if the power rests on limited role characteristics, it is unlikely that A will have power outside the role situation.
One paradoxical type of legitimate influence is that of the "dependent" and "powerless." In some situations a person who has little power can emphasize his powerlessness and influence the more powerful. For example, a person may appear "helpless" in order to get another to perform a task. Schopler and Bateson (1965) suggest that the "power of dependence" is a form of legitimate power since it involves the acceptance of an influence role relationship. Thus, the more powerful is obligated to perform the duties associated with the powerful role position and submit to the demands of the influencee.
The expert power basis stems from the influencee's attribution of superior skills or knowledge to the influencer. The degree of expert power is a function of the amount of knowledge the influencee has and the degree to which the knowledge or skills of the influencer are appropriate for a given situation.
The final power basis, information, differs from the previous bases in that it is "independent" of the influencing agent (Deutsch and Gerard, 1955). That is, this power stems from the "logic," "reasoning" or importance of the communication provided by the influencing agent and independent of the communicator. The content of the communication alone leads to changes in belief structures, behavior, attitudes, etc. In most situations it is difficult to independently distinguish expert and information social power. In fact Collins and Raven (1969) suggest that informational influence may follow only after some degree of expert power is perceived by the influencee (p. 184).
Although there has been considerable research on social power and the French and Raven bases, there has been little to improve the operationalizations of these concepts.
One method of operationalizing social power is by creating an experimental treatment situation depicting the desired social power. The effect of the social power treatment on various dependent variables can then be tested. Busch and Wilson (1976) employed this method in a study on the effect of life insurance salesmen's expert and referent bases of social power on subject's trust in the salesman, overall attitude toward life insurance and several behavioral measures. Low and high expertise power was created by presenting the salesman as either "above average" or "excellent" on six attributes of selling ability and training. The referent conditions of low and high similarity were created by using Byrne's attitude similarity procedure (Byrne, 1961, pp. 713-5). In this method, the influencer is presented as either being similar or dissimilar to the subject on a list of attitude topics.
Although the operationalizations employed in this study appeared at least on the surface, to be reasonable, the authors reported only marginal differences in referent power between the high referent-low expert and iow ref-erent-high expert conditions. This comparison was made using manipulation check measures for referent power which were two 7 Point items("how much they (subject) would like the salesman if the met him" and "would they enjoy working with him in a research experiment.") [The manipulation check for expert power was a one item, nine point scale--"I feel that the knowledge and competence of the salesman who is being studied here is"--with end points of "slightly below average" and definitely above average." The reliability and validity of this item was not discussed.] Thus, while these items have reportedly high reliability, (0.85, Byrne, D. and D. Nelson, 1965), some questions arise regarding the validity of using measures of attraction as surrogates for referent power in a situation where several power bases are operant. From this one might suggest that either the concept of referent power, the method of operationalization, or the manipulation check measures be revised.
Another method of operationalizing social power in an experimental design is via a cartoon representation (Leet-Pellegrini and Rubin, 1974, Raven, 1973). This consists of a series of cartoons illustrating an influence situation. In the Leet-Pellegrini and Rubin study, the first cartoon panel showed a young adult standing in front of a doorway at a city street corner. The second panel depicted a police officer saying to the young adult, "Would you please move away form here." The third panel to be selected operationalized different bases of power (e.g., (Reward) "If you do, I will keep it in mind--there are ways that I can help you," (Legitimate) "I'm asking you to do this as a part of my job.") While this cartoon method provides an alternative to the film presentation of Busch and Wilson, the need for manipulation checks remains. Such checks were not presented in the Leet-Pellegrini and Rubin study.
While a thorough review of common experimental treatment operationalizations of the other social power bases is beyond the scope of this paper, it should be noted that they seldom are more sophisticated than that of those mentioned above. Declarations by the experimenter regarding the influencer's ability to control monetary payments or electrical shocks, or attributed knowledge and experience to the influencer are also typical.
Another perspective on the measurement of social power is provided by Raven, Centers and Rodrigues (1975). In this study, the authors sought to determine the differential susceptibility to a social power base as utilized by one's spouse. Following the introductory statement:
Now, there are many cases where your wife/husband asks you to do something and you do it, even though you may not see clearly why it should be done. We are interested in finding out why you might do as your wife/husband asks, so I will give you some possible reasons and would like you to tell me how likely each of these reasons is.
The subjects rated the following statements on a three point scale (very unlikely, somewhat likely, very likely):
1. Because if you did so then she/he would do something nice for you in return,
2. Because if you did not do so, he/she might do or say something which would be unpleasant for you in return,
3. Because he/she knew what was best in this case and so I did what he/she asked you to do,
4. Because you felt that she/he had a right to ask you to do this and you felt obligated to do as she/he asked, and
5. Because you felt that given you were both part of the same family, you should see eye-to-eye on these matters.
While the face validity of these items appears to be high, the authors presented no information on the reliability and validity of these measures.
This brief review has illustrated the most common methods of operationalizing the French and Raven bases of social power and their shortcomings. The effectiveness of the experimental treatments is often not verified by manipulation check measures. Also, in cases where the manipulation check is not consistent with the treatment, one may question either the treatment or the manipulation check measure unless reliable and valid check measures are employed. The use of rating and ranking measures is also subject to deficiencies. Often, content and other types of validity for the items are assumed. In addition, single item measures are subject to many sources of error and do not allow for the assessment of reliability.
Thus while these approaches provide convenient operationalizations of social power, they do not answer issues of measurement reliability and validity. There is a need for more valid and reliable measures of social power.
For this study a Likert approach was selected. Some of the assumptions which underlie the Likert procedure are: (1) the concept being measured is unidimensional; (2) the intervals between adjacent responses are equal; (3) the intervals are equal across items; and (4) a "positive" direction can be determined for each item (Runkle and McGrath, 1972, p. 314).
The first assumption will be addressed in the analysis section. The second and third assumptions are commonly made in Likert scaling situations and do not appear to be unreasonable for this study. The fourth assumption also seemed reasonable since trained judges were used to select the scale items.
A Likert scale development format was used for this study and included the following major steps: (1) A large pool of belief statements which reflect different characteristics of a social power situation was generated; (2) Judges rated these items to select those which clearly indicate a particular type of power: (3) Influence situations were generated and judged for use as standards for scale development. For each power basis, two situations were selected which depicted that social power base; (4) Subjects assumed the role of the influencee in a situation and responded to multiple items reflecting each basis of power; (5) Items reflecting each social power basis were analyzed to determine the most reliable items for the final scale; (6) Power base scores across situations were compared as a check on predictive validity.
The items generated by the author, depicted characteristics of either an influence situation in general, a particular relationship between the influencer and influencee, abilities or resources of the influencer, etc. One hundred and fifty items were generated, approximately 25 for each French and Raven power type.
While this process does not ensure a thorough sampling of the population of the social power characteristics for each power basis being measured, it is felt that the large number of initial items provides an acceptable representation of social power characteristics.
These items were given to six judges [I wish to thank George Belch, James Bettman, Harold Kassarjian, Richard Lutz, Michael Munson, and Bertram Raven for their assistance and helpful comments.] familiar with the French and Raven typology. Each item was judged as to whether or not it was an indicator of each of the French and Raven types of power. A judge could designate the item to represent more than one power type.
An item was designated by the author as an acceptable indicator if:
1. A minimum of five of the six judges classified the item as an indicator for the same power type; and
2. The item also was not classified as an indicator in the other categories a total of 3 or more times (either within another category or over several categories.)
This procedure required reasonably high interjudge consistency, but still tolerated a small degree of inevitable variability present in judging tasks of psychological concepts. Eighty-five items were selected by the judges as acceptable indicators, 12-16 items for each type of power.
Scenarios were used to represent influence situations. [Because of the costs associated with using actual situations, a scenario approach was utilized. This approach allowed for the development of clear examples of each power basis. While using this approach lessens the external validity of the results, it does allow other researchers to easily and rather inexpensively test the findings.] (See Figure 1 for examples.) In this case, the subjects were instructed to assume the role of the influencee in a given scenario, and then indicate their agreement or disagreement to each of the 85 belief statements selected by the judges.
An important aspect in the development of the scenarios was the need to generate situations which clearly depicted only one type of social power. This was necessary to ensure that the later scale results could be validated with a known situation treatment (i.e., the "expertise" power score would be highest in the expertise situations.) The second issue was the need to generate two situations depicting each type of power. This replication of situation treatments was intended to ensure that the scales developed from this research would have some degree of generalizability.
Situations were generated by the author and reviewed by the same panel of judges. The six judges rated thirty-six situations for each type of influence on a 1 to 5 scale having end points of "does not depict this type of power" and "clearly depicts this type of social power." For each type of power these ratings were summed across judges. Thus each situation had six scores (one for each type of social power) ranging from the lowest possible (6) to the highest (30). The final situations were selected on the basis of being a clear representation of one social power type. For all the situations except for the expert situations, the selections were quite direct. For the expert situations, it was difficult to generate a situation free of informational power. The final expert situations used for this study were agreed upon by the judges to be clear examples of expertise, but also possibly examples of information power. This confounding is understandable considering the close theoretical meaning of these two types of social influence. Situation #4, an example of legitimate power, deserves comment. This situation depicts the "power of the powerless." In this type of situation persons who have few resources in the sense of power and influence, have a "legitimate" right to influence those who are more powerful. Thus either because of societal or personal norms, the more powerful person is obligated to conform to the request of the "powerless" (e.g., a prosperous person giving money to a beggar).
SELECTED SITUATION SCENARIOS
The final instrument consisted of one scenario and the 85 test items which were arranged in random order. [Perhaps a better procedure would have been to word half the items in the negative to protect against possible response set biases. Since efforts were taken to encourage responsible participation by the subjects and no response patterns were observed in the data, it is felt that the procedure used is acceptable.]
Subjects for this research were undergraduate accounting majors at UCLA. A total of 321 were used in the final analysis. Approximately 25 other subjects were eliminated from the study because of incomplete questionnaires. These incompletes did not occur in any particular class or any one influence treatment. This relatively homogeneous population of subjects restricts the generalizability of the results.
Prior to the distribution of the test instrument the subjects were asked to participate in a social science research project and were told that their participation was completely voluntary. They were also informed that if they felt that they could not perform the task in a conscientious manner, to please return the questionnaire unmarked or incomplete. The required time for completion of this task ranged between 15 and 20 minutes.
In Likert test development an index of the items' discriminating power serves as the evaluation criterion. Items with high discriminating power are selected for the final form. This scaling procedure implies a "single common factor" model (Green, 1954). A factor analytic scale analysis was chosen to construct the final scales. (McKelvy, 1976). This approach is quite suitable for "subjective areas where external criterion variables have not yet been identified or where the emphasis is on discovering and defining new theoretical constructs and accompanying operational measures." (McKelvy, 1976, p. 2.) The lack of a clear external criterion is particularly characteristic of this research.
The factor analysis was performed for each set of items (basis of power) across all situations. In addition to providing the factor loadings for each item, this measurement set analysis program provides Cronbach's (1951) alpha coefficient for each set of items, starting with the two highest loading items, then the three highest, etc. This calculation was made for all sets of items having at least a commonality of 0.35 for the smallest loading item. While this cut-off is arbitrary, it appears to be reasonable for this study. The calculation of alpha allows for an assessment of the relationship between test length and internal consistency. While some research situations may require alternative criteria for item selection (i.e., a wide content of scale items), the issue of internal consistency is still important.
The major guidelines for using this program in a uni-factor solution include: Items should be written to measure a well-defined construct; the number of cases should be at least three times the number of items; and the data should be multivariate normal and homoscedastic. Except for the third assumption, these requirements were met. The effect of the departure from the third assumption is that the alpha coefficients are, at best, estimates.
This section will first briefly review the overall "goodness of fit" for the uni-factor scales. Then each scale will be briefly reviewed. Table 1 presents the information on the factor analysis of each scale. The "total common variance with principal factor solution" indicates the degree of overlap between the test items, regardless of whether the 'overlap' occurs in a one, two, or n-dimensional space. Thus, these values suggest that even though the items were designed and rated by judges to be measuring the same concept, the "common" portion of the items account for at best 52.9% of the item scores' variance.
FACTOR ANALYSIS DATA FOR THE ITEM ANALYSIS
A "total common variance with the uni-factor solution" column in Table 1 indicates the degree to which one dimension accounts for between 24 and 35% of the total variance among the test items (e.g., a one dimensional scale accounts for about 25% of the "total" variance). Considering only the common variance, the proportion attributable to the single factor ranges between 50.0 and 67.6% (e.g., a one dimensional scale accounts for about 50% of the "true" variance). These figures suggest that the assumption of a uni-dimensional concept is a reasonable one.
For the six social power (SP) scales discussed below, the factor loadings and alpha coefficients for items with communality greater than .35 are presented in Table 2.
While an appropriate reliability level depends on the variety in item content desired and the reliability of other constructs in a given research situation, a commonly accepted reliability level for most basic psychological research is 0.70 (see Nunnally, p. 226).
The coercion SP items, as expected, centered around potential harm and punishment from the influencer. The item loadings were consistently high and an alpha level of over 0.80 was obtained with the first three items.
For the expertise SP scale the item loadings were consistently high and the alpha values were equally acceptable. Trusting the influencer's knowledge and experience were the primary elements in this factor.
The legitimate SP scale items had fairly high factor loadings. However the best alpha level which could be obtained with items having an estimated communality greater than 0.35 was 0.59. One possible explanation for this is the limited range in item variances which resulted for some situations. Since the "power of the powerless" situation (#4) is a rather unique example of legitimate power, many of the legitimate SP scale items were inappropriate. Thus, having only one situation to represent the higher legitimate power value and 11 situations having small values could cause these lower intercorrelations. Since the same item generation and judging procedures were used for this SP scale as the others, it is believed that the items are indicative of legitimate power and that the lower alpha levels for this scale are primarily due to the limited array of situations used here.
The SP scale for reward is composed of items reflecting the non-attainment of rewards for non-compliance and the importance of the rewards as the motivation for compliant behavior. Loadings greater than 0.5 were obtained for most of these items. To achieve an alpha level of 0.8, five items were required.
As expected, the items dealing with the "logic," "good reasoning" and "sense" of the influencer's message are the best items for the information scale. While all the items have fairly high factor loadings, the best internal consistency (0.74) is achieved by the first three items.
Consistently high loadings were obtained for the referent SP scale also. Similarity in opinions and values and the desirability and goodness of being similar to the influencer were the principal identifiers of this scale. The first three items of this scale achieved an alpha of 0.80.
In summary, the scales for the French and Raven power bases provide reasonable levels of internal consistency. The lower reliability for the legitimate power scale appears to be the result of the situations used in this study. Since the procedures used in generating and judging the scale item were the same for all the scales, it is felt that the items for legitimate power will have acceptable internal consistency in most legitimate power situations.
SOCIAL POWER SCALES
The predictive validity of the SP scales was examined by performing a one way ANOVA, with the situations as the treatments and a power score as the dependent variable. Thus, six ANOVAs were run one for each power base.
If the scales are actually measuring their respective power bases and the situations truly depict a given power base situation, then one would expect that significant differences in the power scores should occur across the situations. Also one would expect that the situations depicting a given social power base should have the highest mean score. Table 3 presents this analysis, using the three highest loading items in each scale to calculate the scores. Newman-Keuls tests of differences are also presented ( p < .05).
The significant F ratios and the higher mean scores for the appropriate situations suggest that the scales and situations were good operationalizations of the power constructs. For all six power bases, one of the appropriate situations had the highest mean score for the power scale. Except for the legitimate social power situation number 4 (the "power of the powerless"), the second situation for each power base was either second or third highest on the appropriate power score.
The results of this study indicate that reliable measures of perceived social power can be obtained with relatively few items. Using the three highest loading items for each of the six French and Raven power bases results in a scale of reasonable total length and acceptable levels of internal consistency.
While it is desirable to have a scale with high internal consistency, the validity of the scale is equally important. A high level of content validity for this scale was sought by generating and selecting items which represented clear theoretical constructs. These constructs were the French and Raven bases of social power which have been widely cited and used in much of the social power literature. Also, the use of a panel of trained judges to screen the test items, and the "face" validity of the scale items suggest that the social power scales have an acceptable level of content validity.
For the assessment of predictive validity, criterion measures were created by generating scenarios which clearly depicted a given basis of social power. Using these "known" situations as standards, the predictive ability of the scales was examined. The consistency between the scale values and situations suggest that the scales have reasonable predictive validity.
The important aspect of construct validity remains to be tested. The replication of relationships between social power and other psychological constructs which were derived in laboratory experiments would serve as good indicators of construct validity.
An area of interest to attitude change researchers which might provide evidence for construct validity would be a comparison of the SP scales with the social influence components of the extended Fishbein model (Fishbein and Ajzen, 1975). This model is based on Dulany's (1961) verbal conditioning research and suggests that the amount of social influence is dependent upon the influencee's motivation to comply with the influencer. For example, one possible hypothesis that might be tested is that as the degree of referent social power increases, so will the motivation to comply. This analysis might provide insights into the reasons why one referent is more influential than another. In understanding family decision-making, this information might be particularly useful.
ONE WAY ANOVA AND NEWMAN-KEULS TEST OF DIFFERENCES IN SOCIAL POWER SCORES ACROSS SITUATIONS
In conclusion, social power and influence are important aspects of many consumer decisions. To aid in the understanding of these aspects, this paper has attempted to provide a first step in the development of reliable and valid measures for the bases of social power.
Paul Busch and David T. Wilson, "An Experimental Analysis of a Salesman's Expert and Referent Bases of Social Power in the Buyer-Seller Dyad," Journal of Marketing Research, 13 (February, 1976), 3-11.
Donn Byrne, "Interpersonal Attraction and Attitude Similarity,'' Journal of Abnormal and Social Psychology, 62 (May, 1961), 713-715.
Donn Byrne and Don Nelson, "Attraction as a Linear Similarity," Journal of Personality and Social Psychology, (June, 1965), 659-663.
Darwin Cartwright, "A Field Theoretical Conception of Power," in D. Cartwright (ed.) Studies in Social Power (Ann Arbor, Mich.: Institute for Social Research, 1959).
Darwin Cartwright and Alvin Zander, Group Dynamics (New York: Harper and Row, 1968), 266.
Barry E. Collins and Bertram Raven, "Group Structure: Attraction, Coalitions, Communication and Power," in G. Lindsey and E. Aronson (eds.) The Handbook of Social Psychology, Vol. 4 (Reading, Mass.: Addison-Wesley, 1969).
Lee J. Cronbach, "Coefficient Alpha and the Internal Structure of Tests," Psychometrika, 16 (1951).
Morton Deutsch and Harold Gerald, "A Study of Normative and Informational Influence Upon Individual Judgment," Journal of Abnormal and Social Psychology, 51 (1955), 629-636.
Don E. Dulany, Jr., "Hypotheses and Habits in Verbal Operant Conditioning," Journal of Abnormal and Social Psychology, 63 (September, 1961), 251-263.
Martin Fishbein and Icek Ajzen, Belief, Attitude, Intention and Behavior: An Introduction to Theory and Action (Reading, Mass.: Addison-Wesley, 1975).
John R. P. French and Bertram Raven, "The Basis of Social Power," in D. Cartwright (ed.) Studies in Social Power, (Ann Arbor, Mich.: Institute for Social Research, 1959).
Bert F. Green, "Attitude Measurement," in Handbook of Social Psychology, Vol. 1 (Reading, Mass.: Addison-Wesley, 1954).
Harold H. Kelly, "Two Functions of Reference Groups," in G. E. Sevanson, T. M. Newcomb and E. L. Hartley (eds.) Readings in Social Psychology (New York: Holt, Rinehart and Winston, Inc., 1952).
Helena Leet-Pellegrini and Jeffrey Z. Rubin, "The Effects of Six Bases of Power Upon Compliance, Identification, and Internalization," Bulletin of the Psychonomic Society, 3 (January, 1974), 63-70.
Kurt Lewin, Field Theory in Social Science (New York: Harper, 1951).
William McKelvy, "An Approach for Developing Shorter and Better Measuring Instruments," Working Paper #76-6, Human Systems Development Study Center, UCLA, 1976.
Jum C. Nunnally, Psychometric Theory (New York: McGraw Hill Co., 1967).
William E. Pollard and Terence R. Mitchell, "Decision Theory Analysis of Social Power," Psychological Bulletin, 78 (November, 1972), 433-446.
Bertram H. Raven, "Social Influence and Power," in I. D. Steiner and M. Fishbein (eds.) Current Studies in Social Psychology (New York: Holt, Rinehart and Winston, Inc., 1965), 371-382.
Bertram H. Raven, "A Cartoon Task for the Analysis of the Bases of Social Power," Unpublished manuscript, UCLA, 1973.
Bertram H. Raven, Richard Centers, and Aroldo Rodrigues, "The Bases of Conjugal Power," in R. E. Cromwell and D. H. Olson (eds.) Power in Families (New York: Sage Publications, 1975).
Philip J. Runkel and Joseph E. McGrath, Research on Human Behavior: A Systematic Guide to Method (New York: Holt, Rinehart and Winston, Inc., 1972).
Constantina Safelios-Rothschild, "The Study of Family Power Structure: A Review 1960-1969," Journal of Marriage and the Family, 32 (November, 1970), 539-552.
John Schopler and Nicholas Bateson, "The Power of Dependence,'' Journal of Personality and Social Psychology, 2 (1965), 247-254.