Journal of Family Psychology 2002, Vol. 16, No. 3, 318 –337

Copyright 2002 by the American Psychological Association, Inc. 0893-3200/02/$5.00 DOI: 10.1037//0893-3200.16.3.318

Development of an Empirical Typology of African American Family Functioning Jelani Mandara and Carolyn B. Murray University of California, Riverside This study empirically identified types of African American families. Adolescents (N ⫽ 111) were assessed on family functioning. With cluster analytic methods, 3 types of families were identified. The cohesive–authoritative type was above average on parental education and income, averaged about 2 children, exhibited a high quality of family functioning and high self-esteem in adolescents. The conflictive–authoritarian type had average parental education and income, an average of 2.7 children, exhibited controlling and rigid discipline, and placed a high emphasis on achievement. The defensive–neglectful type was predominately headed by single mothers with below average education and income and averaged about 3 children. Such families displayed chaotic family processes, and adolescents tended to suffer from low self-esteem. The typology exhibited good reliability. The implications of the typology are discussed.

The quality of family functioning has consistently been identified as one of the most important predictors of individual well-being. Studies have illustrated the effects of family functioning on reducing the risks of youth violence and aggression (Gorman-Smith, Tolan, Zelli, & Huesmann, 1996) and substance abuse (Brooks, Stuewig, & LeCroy, 1998). Family functioning has been shown to have positive effects on postdivorce adjustment (Johnson & McNeil, 1998), academic achievement (Heiss, 1996), cognitive performance (Burchinal, Campbell, Bryant, & Wasik, 1997), social and emotional adjustment (Taylor & Wang, 1997), and self-esteem (Brody & Flor, 1997; Mandara & Murray, 2000; Murray & Mandara, in press). However, common among most of these studies is that they were limited to examinations of isolated family functioning variables. The variable-centered, or dimensional, approach used by these studies examines the covariation among variables across families (Bergman, Cairns, Nilsson, & Nystedt, 2000). Therefore, the focus of analysis was on variables. This approach, similar to all bottom-up approaches, assumes that the objects of interest can be described and explained by reducing them to their lowest common denominators (e.g., variables). It further assumes that variables are interrelated similarly and linearly across

all families (McQuitty, 1987). Therefore, the variable-centered approach assumes that the meaning and covariation of all variables is the same for all families (see Block, 2000, p. 160). Although the variable-centered studies have revealed a great deal of information regarding the importance of specific family functioning dimensions, families also exhibit patterns of functioning across several interrelated dimensions that are not captured by unidimensional analyses (Baumrind, 1991, Fisher & Ransom, 1995; McQuitty, 1987).

Typological Approach The case-centered, or typological, approach is a less commonly used approach that examines interrelated dimensions of family functioning simultaneously (Bergman et al., 2000). Similar to all top-down or systemic theories, the basic logic of the typological approach stems from the idea that types possess more information than the variables used to create them (Bergman et al., 2000). This is because types are assumed to be “gestalt-like in that their effects are not entirely predictable from a single aggregation (average correlation or dimensional location) of information about constituent trait members; the whole is more than the sum of the parts” (Anderson & Sedikides, 1991, p. 205). The empirical typological approach first creates a classification of families based on several family functioning dimensions, then examines the multidimensional relationships among dimensions (patterns), and between belongingness to a family type and other variables (Roberts & Leonard, 1998). Therefore, the typological approach allows for the assessment of the unique interactive relationships of variables within a family type, as well as comparisons across family types (Bergman et. al., 2000; McQuitty, 1987). The units of analysis then shift from variables to families (Bergman et al., 2000; Fisher & Ransom, 1995), but the importance of the individual variables is retained. Therefore, researchers can provide an empirically valid and

Jelani Mandara and Carolyn B. Murray, Department of Psychology, University of California, Riverside. This research was supported in part by a grant from the National Institutes of Health (HD30614-02). We would like to thank Jon Mettrick and Mary Gauvain for their very helpful comments on earlier drafts of this article. Correspondence concerning this article should be sent to Jelani Mandara, who is now at the Department of Human Development and Social Policy, Northwestern University, Evanston, Illinois 60208-2610, or Carolyn B. Murray, Department of Psychology, University of California, Riverside, California 92521. E-mail: [email protected] or [email protected]

318

AFRICAN AMERICAN FAMILY TYPOLOGY

rich source of quantitative and qualitative information about families or family members (Bergman et. al., 2000).

Family Typologies Several family researchers have also begun to utilize the typological approach and have developed a variety of family typologies. In a series of studies, Baumrind (1968, 1972, 1989, 1991) and others developed the most famous and important family typology thus far. On the basis of observations of 134 middle-class two-parent European-American families, Baumrind theorized that families could be reliably classified into variations of four prototypical family types according to their relative level on two broad dimensions: parental demandingness (control and restrictiveness) and responsiveness (warmth and noncoerciveness). The four family types are: (a) authoritative, (b) authoritarian, (c) permissive, and (d) neglectful. Baumrind’s work and that of others building on her theory has generally found that authoritative parenting is associated with better well-being in children and adolescents and that neglectful parenting is related to a host of problem behaviors (Baumrind, 1991; Pomerantz & Ruble, 1998; Steinberg, Lamborn, Darling, Mounts, & Dornbusch, 1994; Weiss & Schwarz, 1996). Even though the work of this tradition has greatly advanced researchers’ understanding of families, notable limitations for the present purpose persist. For instance, Baumrind’s main concern was to understand how families raise competent children, not necessarily what types of families exist. Therefore, for those researchers interested in understanding family psychology in general (including those factors that affect children), many questions remain to be answered. Furthermore, the parenting prototypes are classified according to whether they were high or low on demandingness and responsiveness. When dealing with mostly continuous variables, this method loses a great deal of the information concerning the variables because it artificially classifies objects on the basis of being either high or low on certain variables (see Block & Ozer, 1982). Another limitation of Baumrind’s typology for describing families (not predicting outcomes) is that because the classification of families was based on their relative standing on only two (albeit very broad and important) dimensions of family functioning, much of the details of families and explanations of why they function as they do may be lost (see Moos & Moos, 1976, p. 368, for a discussion of this issue in family research). When trying to accurately describe an object, it is best to use as much specific information as possible. The parsimonious nature of Baumrind’s typology makes it easier to classify families, but it also makes it less specific in its descriptions of families. The more empirical approach is to use multidimensional statistical methods to first classify families according to their similarities to and differences from each other and then to examine the patterns of functioning within each family cluster. Therefore, the validity of the patterns can be easily assessed and modified by determining whether the patterns hold up over time and across different samples. The more

319

data that accumulate, the more stable the classification will become. Furthermore, the continuous nature of the variables and the multidimensional relationships among several dimensions can be retained and examined. This allows for a more detailed description of the families. Several researchers have recently created empirically derived family typologies on the basis of various family functioning indicators (Crouter & Manke, 1997; Fisher & Ransom, 1995; Lustig, 1997). Unfortunately, similar to Baumrind’s (1968) system, these typologies were more concerned with certain outcome variables (e.g., dealing with bereavement, alcoholism, health, etc.) that may be caused or influenced by family functioning than with family functioning per se. Also, relatively few attempts have been made to develop an empirically derived typology of non–European American families for the purpose of understanding the family functioning patterns in other cultural groups. Although all families share many common needs and practices, cultural and sociopolitical differences require somewhat different socialization concerns and strategies (Baumrind, 1972; Lamborn, Dornbusch, & Steinberg, 1996; Murray & Mandara, 2001). Because most typologies are primarily derived from the family functioning concerns and practices of middle-class European-American families, these characteristics have come to define the standard of normal family functioning. Although researchers have attempted to address this issue, they have mainly done so by using racially representative samples when devising scales and developing typologies or by lumping all non–European American families into one category (Lamborn et. al., 1996; Moos & Moos, 1976). However, the meaning and importance of each family dimension may be unique to different cultural groups (Gorman-Smith et al., 1996). For example, Baumrind (1972) included a sample of 16 African American families she did not use in her original typology and found that authoritative and authoritarian disciplinary practices were qualitatively different for African American and European American families. Recent empirical studies have also shown that the most effective disciplinary practices of African American families are different than those used by European American families (Starrels, 1994; Steinberg, Dornbusch, & Brown, 1992). The typical finding is that the negative impact of parental strictness on European American youths is not the same for other youths (Chao, 1994). For instance, Lamborn, Dornbusch, and Steinberg (1996) found that authoritarian parenting was not as detrimental to African American youths compared with European American youths in either predominately White or mixed neighborhoods. Moreover, explicit racial socialization is a distinctive characteristic of 65% of all African American families (Thornton, 1997) but of only a small percent of European American families (Helms, 1990). In like manner, the meaning and importance of racial socialization is completely different to groups depending on whether they are in the numerical majority or minority in a given population (Murray & Mandara, 2001). Therefore, these approaches obscure the normative functioning of different cultural groups.

320

MANDARA AND MURRAY

Many studies have been done on the relationship of family functioning and well-being for African Americans (Cooper, Holman, & Braithwaite, 1983; Gorman-Smith et al., 1996). Family environment quality significantly predicted African American adolescent psychological adjustment (Dancy & Handal, 1984), academic achievement (Heiss, 1996; Lamborn, Dornbusch, & Steinberg, 1996), violence and delinquency (Gorman-Smith et. al., 1996), and self-esteem (Mandara & Murray, 2000), among many other variables. However, none of the studies empirically assessed the types of family functioning patterns within African American populations. In a series of studies, Steinberg, Dornbusch, and their colleagues (Glasgow, Dornbusch, Troyer, Skinberg, & Ritter, 1997; Lamborn et al., 1996; Steinberg et al., 1992) have successfully classified African American families according to variations of Baumrind’s typology, and have uncovered a great deal about the relationships between parenting styles and well-being for African Americans. However, all the limitations of Baumrind’s typology for assessing patterns of African American family functioning still apply. In another recent study, Gorman-Smith, Tolan, Loeber, and Henry (1998) identified the rates of family problems in a group of urban African American and Latino families using Stouthamer-Loeber and Loeber’s (1988) family problems typology. They found that over a 4-year period, according to the perceptions of a young male adolescent in the family, about 16% of the families were neglectful of their children, 26% were classified as conflictive, 33% exhibited patterns of deviance, and over 80% suffered from some form of disruption. They also found that family problems were significantly related to antisocial behaviors and other major types of problem behaviors in the boys. However, the percentage of problems economically impoverished families face is very different than the family functioning of nonimpoverished African American families.

Purpose of the Present Study Therefore, the purpose of the present study is to develop an empirical typology of African American family functioning that includes a wider range of class and income than in previous studies. The present study sought to identify types of African American families assessed on a variety of critical family functioning dimensions. Because the typology requires predictive and other forms of construct validation (Aldenderfer & Blashfield, 1984), a further purpose was to determine whether the family types differed on important demographic and attitudinal characteristics, as well as adolescent self-esteem, ethnic identity, and personality. A major benefit of a good typology is that it captures the multidimensional relationships of variables. A composite variable accounts for the additive or linear effects of the variables. However, this is not the same as assessing the effects of the multidimensional relationships between the variables. Implicit in the logic of the typological approach is that this multidimensionality accounts for information not accounted for by the linear effects of the variables. A further

purpose then, was to compare the derived typology against a composite family functioning variable for predicting various demographic and outcome variables. If the differences between the types are not meaningful after the linear effects of a composite family functioning variable are controlled, then the typology may have little applied benefit. However, if the typology accounts for an important amount of variance over and above what is accounted for by a composite variable, then further support for the predictive utility of the typology exists (Anderson & Sedikides, 1991). A limitation of multivariate subtyping is that several different clustering solutions may exist for the same sample. This makes the reliability of the resulting typology a major concern (Breckenridge, 2000; Milligan, 1996). Therefore, the final purpose of this study was to examine the reliability of the typology by using a standard cross-validation procedure and by assessing the replicability of the typology over time.

Method The participants are from the African American Families and Child Outcomes Project. Data was collected at three time periods. The first assessments were taken between 1994 and 1995. Wave 2 assessments were between 1995 and 1996, and Wave 3 data were collected between 1996 and 1997. The data for this study were part of the first and second wave assessments.

Participants One hundred sixteen 15-year-old African American adolescents (64% female, 36% male) from various high schools in southern California and their parents participated in the study. Fifty percent of the parents were married, 37.5% were divorced, and 12.5% were never-married mothers. No single fathers were in the sample. The mothers’ ages ranged from 30 to 52 years (M ⫽ 36.9, SD ⫽ 5.2). The number of children per family ranged between 1 and 5 (M ⫽ 2.5, SD ⫽ 2.0). Only 9% of the parents in the sample did not graduate from high school. Approximately 28% of the parents graduated from high school only. Thirty eight percent experienced some college, and roughly 12% of the sample graduated from college. Twelve percent did not report any educational information. Income was obtained by asking parents to rate their levels of annual family income on sixteen $5,000 intervals ranging from 0 –$4,900 to over $90,000. The average annual household income for the sample was $27,500 (SD ⫽ $12,000). Approximately 20% of the sample earned less than $20,000, and 35% earned over $35,000 annually. This sample’s demographics are similar to the general trends for African American families in Southern California (U.S. Bureau of the Census, 1996).

Procedure Participants were recruited from four school districts. Parental permission was obtained for each student prior to participation. Each student received $10, and the parents received $25 for a single 2-hour session. Information regarding the perceptions of family functioning and child behavior was obtained from the adolescents; all other demographic and attitudinal information was obtained from their parents.

AFRICAN AMERICAN FAMILY TYPOLOGY

Measures Two instruments were used to assess family functioning: the Family Environment Scale (FES; Moos & Moos, 1986), and the Black Family Process Q-Sort (BFPQ; Peacock, Murray, Ozer, & Stokes, 1996). Together, these instruments assess a total of 18 different family functioning dimensions. The adolescent completed both family functioning measures. The adolescents also completed the Multi-Dimensional Self-Esteem Inventory (MDSEI) to assess their perceived levels of self-esteem. Family Environment Scale. The FES is a 90-item true-or-false test that assesses perceptions of 10 areas of family functioning. The subscales are derived from three family environment dimensions. Table 1 presents a brief description and example for each domain and subscale (taken from Moos & Moos, 1976, p. 360). The Relationship dimension consists of the Cohesion, Conflict, and Expressiveness subscales. This dimension assesses perceptions of the degree to which family members express concern and commitment to the family, express anger and engage in conflictive encounters, and express their feelings. The Personal Growth dimension measures the emphasis placed on Independence– Autonomy, Achievement, Intellectual–Cultural activities, Active Recreational activites, and Moral–Religious areas of family functioning. The Systems Maintenance dimension consists of the last two subscales, Family Organization and Control. This dimension assesses the general organization and structuring of activities

within the family and the extent to which family members exert control over each other. Norms for the FES are derived from 285 families sampled from predominately middle-class and upper-class SES European American families. The 10 subscales show intercorrelations averaging around .20 and exhibit adequate internal consistency (Cronbach alphas range from .64 to .79). Eight-week test–retest reliabilities ranged from .68 to .86 (Moos & Moos, 1986). For the present sample, alpha coefficients ranged between .20 and .75. A 1-year lag test–retest showed an average reliability of .52. The FES has been used with African American samples many times, and has consistently shown excellent predictive validity (Clark & Armstead, 2000; Dancy & Handal, 1984). Black Family Process Q-Sort. The 65-item BFPQ assesses eight family functioning areas across (a) racial socialization, (b) family disciplinary practices, (c) family communication, and (d) religious socialization domains. Table 2 presents a brief description and example of each BFPQ domain and subscale. The BFPQ assesses four racial socialization styles: (a) race defensive, (b) race aware, (c) race empowered, and (d) race naı¨ve (Murray & Mandara, in press). The race defensive factor reflects a reactive racial socialization style that teaches a dislike for other racial groups while simultaneously imitating European American behavior. The race aware factor reflects an active socialization style that teaches children to be proud of their racial group. The

Table 1 Family Functioning Variable Descriptions (Family Environment Scale) Variable

Description Relationship dimensions

9. Cohesion 11. Expressiveness 18. Conflict

The extent to which family members are concerned and committed to the family and the degree to which they are helpful and supportive to each other. The extent to which family members are allowed and encouraged to act openly and to express their feelings directly. The extent to which the open expression of anger and aggression and generally conflictual interactions are characteristic of the family. Personal-growth dimensions

12. Independence 10. Achievement orientation 13. Intellectual orientation 14. Active recreation 15. Religious emphasis (morality)

The extent to which family members are encouraged to be assertive and self-sufficient, to make their own decisions, and to think things out for themselves. The extent to which different types of activities (e.g., school and work) are cast into an achievement-oriented or competitive framework. The extent to which the family is concerned about political, social, intellectual, and cultural activities. The extent to which the family participates actively in various recreational and sporting activities. The extent to which the family actively discuses and emphasizes ethical and religious issues and values. System-maintenance dimensions

16. Organization

17. Control

321

The extent to which order and organization are important in the family in terms of structuring family activities, financial planning, and the explicitness and clarity of rules and responsibilities. The extent to which the family is organized in a hierarchical manner, the rigidity of rules and procedures, and the extent to which family members order each other around.

Note. Numbers correspond to the listing of variables in Table 3.

322

MANDARA AND MURRAY

Table 2 Family Functioning Variable Descriptions (Black Family Process Q-Sort) Variable

Description Racial socialization dimension

1. Empowerment 2. Awareness 3. Defensiveness 4. Naı¨vete

The extent to which parents stress positive racial identity and proactive personal development. The extent to which parents teach children to be proud of their racial group. The extent to which parents stress reactive positions toward other racial groups while not teaching children to be proud of their racial group. The extent to which parents are passive in their teaching of racial issues. Disciplinary/communication dimension

5. Authoritative 6. Authoritarian 7. Neglectful 8. Religious

The extent often. The extent with the The extent The extent

to which parents are supportive and nurturing and give praise to which parents are controlling, critical, and express unhappiness children. to which parents express concern and emotion for their children. to which religious issues are focused on in the home.

Note. Numbers correspond to the listing of variables in Table 3.

race empowered factor reflects a proactive racial socialization style that stresses racial identity and the ability to overcome obstacles in life despite racial barriers. The race naı¨ve factor is the passive racial socialization style, which naively teaches about modern race issues. The BFPQ also assesses three disciplinary/communication styles, and one religious values dimension of family functioning. To reflect the content of the items, the discipline/communication factors were subsequently named Authoritative, Authoritarian, and Neglectful. The Authoritative factor assesses the degree to which parents are supportive, nurturing, and involved with their children as well as giving praise often. The Authoritarian factor assesses the degree to which parents are controlling and critical and express unhappiness with their children’s performance and abilities. The Neglectful factor assesses the degree to which parents express concern and emotion for their children. Development of the BFPQ was based on relevant information concerning African American family functioning derived from scientific literature and high school students’ reports of their family life. The children sorted the 65 items into three piles: (a) “like my primary caregiver,” (b) “not like my primary caregiver,” and (c) “not sure.” The cards were then successively ranked into seven categories ranging from 1 (not like my primary caregiver) to 7 (most like my primary caregiver). This procedure forces a quasinormal distribution on the items. Cronbach alphas for the present sample are presented in Table 3 (see Peacock, Murray, Ozer & Stokes, 1996, for more detailed psychometrics). Demographic characteristics. A standard demographics questionnaire was also administered to the parents to obtain information about the families’ annual income, parental marital status, parents’ age, parents’ education, number of children, religiosity, parents’ educational expectations for their children, and how well parents thought their children were disciplined. The adolescents also rated how well they obey. Religiosity. To assess how parents feel about religion and church, we asked two questions. First, “How important is religion in your home?” (1 ⫽ not at all, 4 ⫽ very important). Second, “How often do you attend church?” (1 ⫽ not at all, 4 ⫽ very often). Educational expectations. To assess parents’ attitudes about their child’s education, we asked two additional questions. First, “How important is your child’s school attendance?” (1 ⫽ not at

all, to 4 ⫽ very important) Second, “How far in school do you expect your child to go?” The choices ranged from finishing grade school to medical or professional school. Discipline and child behavior. Parents’ attitudes about how effective the child is disciplined were assessed by a question that asked, “Think of the person primarily responsible for disciplining your child. To what extent do you feel your child is effectively disciplined?” (1 ⫽ not effectively at all, 4 ⫽ very effectively). To assess the adolescent’s perception of how well he or she obeys, we asked the adolescents to rate the statement, “I always obey,” on a 7-point scale ranging from 1 (least descriptive) to 7 (most descriptive). The Multi-Dimensional Self-Esteem Inventory. The MDSEI is a 116-item instrument that assesses Global and 9 domain-specific aspects of self-esteem. Adolescent participants indicated on a 5-point Likert-type scale how accurately 61 of the items describe them and how often they experience the thoughts and feelings described in 55 of the items. The nine domain-specific subscales are Feelings of Competence, Personal Power, Lovability, Likability, Identity Integration, Self-Control, Moral Self-Approval, Body Functioning, and a Defensiveness factor (i.e., the degree to which a person denies ubiquitous human weaknesses). The alphas for the present sample ranged from .64 to .85. Separate alphas for male and female participants were also examined, and both groups averaged .72 for the 10 subscales. For more specific details of the domain content and reliabilities of the subscales for other samples see O’Brien and Epstein (1988). African Self-Consciousness Measure, Revised (ASCR). The ASCR (Stokes, Murray, Peacock, & Kaiser, 1994) was adapted from Baldwin and Bell’s (1985) African Self-Consciousness Scale, a 42-item personality questionnaire designed to assess the personality construct, African self-consciousness. The ASCR has the following four factors: (a) Personal Identification With the Group (e.g., “I feel like I am also being mistreated in a situation where I see another African-American person being mistreated”), (b) SelfReinforcement Against Racism, (c) Racial and Cultural Awareness, and (d) Value for African American Culture (e.g., “I have difficulty identifying with the culture of African American people”). The Cronbach alphas for these four factors were .66, .70, .46, and .57, respectively. Multigroup Ethnic Identity Measure (MEIM). The MEIM (Phinney, 1992) was developed to measure elements of identity

323

⫺.27* .65 ⫺.20* .10 .6 .03 .14 .06 .59 ⫺.08 ⫺.02 .32* ⫺.20* .40 ⫺.15 ⫺.05 .16 .07 .09 .40 ⫺.02 .09 .35* .27* .20* .05 .63 ⫺.05 .01 .53* .17 .29* .23* .49* .63 ⫺.18 .55* .20* .35* .17 .10 .22* .22* .52 ⫺.09 .10 .54* .17 .00 .06 .39* .26* .21* .54 .03 .05 ⫺.01 .27* ⫺.21* .07 ⫺.07 ⫺.03 .23* .16 .36 .15 ⫺.03 ⫺.44* .20* ⫺.35* ⫺.11 ⫺.22* ⫺.26* ⫺.05 ⫺.35* .21* .65

common to all ethnic groups. The instrument consists of 24 items to which subjects indicate agreement along a continuum ranging from 1 (strongly disagree) to 4 (strongly agree). The scale consists of two factors: one for ethnic identity and the other for orientation toward other groups. Components of ethnic identity include (a) ethnic self-identification, (b) ethnic behavior and practices (social activities and participation in cultural traditions), (c) affirmation and belonging (attachment to the group), and (d) ethnic identity achievement (sense of self as a member of one’s ethnic group). Reliability (assessed by Cronbach’s alpha) was .81 for a high school sample and .90 for a college sample (Phinney, 1992). For the present sample, Cronbach’s alpha for the total was .79. Personality. The five-factor model of personality was derived from an Adjective Q-Sort developed by Aguilar, Kaiser, Murray, and Ozer (1998). The Adjective Q-Sort was adapted from the Self-Descriptive Q-Set (Block & Block, 1980) to function more appropriately in samples of African American adolescents and children. The participants described themselves by placing each of the 43 items in one of seven categories ranging from least descriptive to most descriptive. Each category held six cards, with the exception of the neutral or unsure category, which held seven cards. The five personality factors are Extraversion, Agreeableness, Conscientiousness, Emotional Stability, and Openness to Experiences. The alphas for the present sample are .53, .75, .50, and .40. The Openness to Experiences scale only used one Q-sort item.

.32 ⫺.11 .05 ⫺.16 ⫺.02 .03 ⫺.12 .07 .04 ⫺.06 ⫺.03 ⫺.06 ⫺.17 .01 .08

.74 ⫺.76* .12 .02 .30* .00 .15 ⫺.01 .12 .13 ⫺.01 .24* ⫺.26* ⫺.33*

.72 .59 ⫺.26* ⫺.39* ⫺.05 ⫺.16 .01 ⫺.15 ⫺.25* ⫺.10 ⫺.18 .23* .32*

Results

Note. Alphas are presented along the diagonal. *p ⬍ .05.

.75 .00 ⫺.52* .38* ⫺.18 ⫺.22* ⫺.05 ⫺.13 .05 ⫺.10 ⫺.23* ⫺.09 ⫺.16 ⫺.16 .04 .08

17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2

.55 ⫺.43* ⫺.15 ⫺.02 ⫺.16 ⫺.13 .18 ⫺.04 .05 ⫺.13 .27* .06 .01 .13 .12 .09 .07 .44 .25* ⫺.35* ⫺.26* .16 ⫺.38* ⫺.13 .11 .20* .24* ⫺.04 .09 .32* .32* .20* .14 .01 ⫺.16

1 SD

0.79 0.85 0.66 0.76 0.64 0.91 0.72 1.76 0.23 0.16 0.19 0.15 0.22 0.23 0.19 0.21 0.24 0.79 4.74 5.26 2.51 3.39 4.84 3.07 5.36 4.45 1.67 1.73 1.49 1.56 1.59 1.63 1.71 1.65 1.42 4.74

M Variable

Empowerment Awareness Defensiveness Naı¨vete Authoritative Authoritarian Neglectful Religious Cohesion Achievement Expressiveness Independence Intellectual Recreation Morality Organization Control Conflict

Table 3 Means, Standard Deviations, Alphas, and Intercorrelations for All Family Functioning Variables Based on Adolescent Perceptions

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18.

18

AFRICAN AMERICAN FAMILY TYPOLOGY

We used standard cluster-analytic procedures (Milligan, 1996) to classify families on the basis of their pattern of functioning across the 18 family functioning variables. To form clusters with the smallest within-group differences and largest between-group differences, we performed two related procedures. First, agglomerative hierarchical cluster analysis was initially conducted to establish the number of clusters in the sample. Then, k-means procedures were used to group families. The biggest limitation of cluster analysis, as with factor analysis, is that there are no standard criteria for choosing the best solution from among the options (Nunnally & Bernstein, 1994). Therefore, the selection needs to be based on the examination of several different cluster solutions, in as many ways as the technology allows, as well as some theoretical notions. Thus, we examined the icicle plots, agglomerative schedule, dendrogram, and cluster memberships at different stages of the agglomerative hierarchical cluster analysis to determine the best cluster solution. A four-cluster solution was determined to best represent the data, because the cluster profiles were most interpretable, and they seemed to maximize the differences between and minimize the differences within each cluster the best. We also examined the three- and five-cluster solutions, but they resulted in weaker distributions of cases per cluster and ambiguous interpretations of cluster characteristics. For instance, the clusters were significantly different on only half of the variables used to create the typology for the threeand five-cluster solution, but most were significantly different across the four-cluster solution. We then computed a k-means cluster analysis using a four-cluster solution and the initial cluster centers obtained

324

MANDARA AND MURRAY

Table 4 Family Functioning Profile Means and Univariate Tests for Family Types Family type means Family variable FES Cohesion Achievement Expressiveness Independence Intellectual Recreation Morality Organization Control Conflict BFPQ Empowerment Awareness Defensiveness Naı¨vete Authoritative Authoritarian Neglectful Religious

Cohesive (n ⫽ 44)

Conflictive (n ⫽ 40)

Defensive (n ⫽ 25)

F(2, 103)

p

␩2

1.83a 1.72b 1.59a 1.57a 1.70a 1.76a 1.77a 1.75a 1.61b 1.21c

1.60b 1.84a 1.38c 1.57a 1.57b 1.61b 1.76a 1.64b 1.73a 1.58a

1.52b 1.55c 1.48b 1.50b 1.50c 1.44c 1.51b 1.49c 1.58b 1.50b

25.61 45.63 16.81 2.28 19.09 21.32 22.74 17.07 9.14 47.32

.00 .00 .00 .10 .00 .00 .00 .00 .00 .00

.33 .47 .25 .04 .27 .29 .30 .25 .15 .48

4.72a 4.98 2.32b 3.59 5.18a 2.62b 4.13b 4.71a

4.55a 5.04 2.33b 3.52 4.68b 3.45a 4.35a 4.80a

4.08b 4.78 2.85b 3.59 4.65b 3.61a 4.40a 3.50b

14.68 1.60 15.40 0.18 12.65 21.52 4.79 5.63

.00 .20 .00 .83 .00 .00 .01 .00

.22 .03 .23 .00 .20 .30 .09 .10

Note. FES ⫽ Family Environment Scale; BFPQ ⫽ Black Family Process Q-Sort. Different subscripts within a row indicate significantly different means.

from the hierarchical analysis. This analysis was then compared with a four-cluster k-means analysis in which the initial cluster centers were not provided, but rather iteratively estimated by the procedure. We subsequently used the solution in which the cluster centers were not specified because it yielded the best distribution of cases, and the most interpretable cluster differences, using the same criteria as before. The one similarity of both k-means cluster solutions was a cluster containing only two cases. Because the cluster was too small to use in any analysis, and because empirical methods or sound theoretical reasoning could not group the two families into one of the other clusters, we dropped the cases from all subsequent analyses. This is a common procedure in cluster analysis and typology development (Bergman, 1988). To examine the characteristics of the three clusters, we computed a multivariate analysis of variance (MANOVA) on the 18 family functioning subscales, with the three clusters serving as the factor. Table 4 presents the means, standard deviations, univariate tests of significance, pairwise comparisons, and effect sizes for the analyses. As would be expected, we obtained a significant multivariate effect, Wilks’s lambda ⫽ .09, F(36, 178) ⫽ 1.60, p ⬎ .00, ␩2 ⫽ .70, indicating that about 70% of the variability in family functioning was accounted for by group differences among the three clusters. Univariate analyses of variance (ANOVAs) of each clustering variable for the three clusters were then performed. As can be seen in Table 4, all ANOVAs were significant, except for the race socialization styles naı¨vete and awareness, and family independence orientation.1 The pseudo-F tests should be used only for descrip-

tive purposes, and not as tests of hypotheses, as the clusters have been chosen to maximize the differences between clusters (Norusis, 1994). Figures 1 and 2 present the standardized profile means of the three clusters. On the basis of the most salient characteristics of the families, the clusters were subsequently named (a) Cohesive–Authoritative (n ⫽ 44, 40%), (b) Conflictive–Authoritarian (n ⫽ 40, 37%), and (c) Defensive– Neglectful (n ⫽ 25, 23%). Pairwise comparisons revealed that the three clusters are relatively distinct types of families. The Cohesive– Authoritative family type exhibits the highest overall level of family functioning, as based on the observation that they are significantly higher than the other two family types on many of the family functioning dimensions. Their main characteristics are high levels of family cohesion and authoritative disciplinary styles. Chaotic family relationships and a focus on achievement and parental control distinguish the Conflictive–Authoritarian family type. Significantly higher levels of defensive racial socialization and neglectful and authoritarian disciplinary practices distinguish the Defensive–Neglectful family type. This family type is also significantly lower than the other two types on most of the dimensions of personal growth and development. 1 We computed separate cluster analyses with the race naı¨ve and intellectual independence variables removed. The solutions were not as good as the one with these variables included. They did not separate the clusters as well, nor did they predict outcome variables as well as the original solution. Therefore, we used the solution including those two variables.

AFRICAN AMERICAN FAMILY TYPOLOGY

325

Figure 1. Family Environment Scale profile means by family type.

The next purpose of this study was to examine demographic and attitudinal differences among the family types. This form of external validation is critical to any typology because the cluster analysis itself cannot validate the usefulness of the typology (Aldenderfer & Blashfield, 1984).

Therefore, we computed a separate MANOVA, using family type as the factor, (a) demographic variables (family income, marital status, household size, parental age, and parental education), (b) religiosity, (c) parents’ educational expectations for their child, and (d) subjective evaluations

Figure 2. Black Family Process Q-Sort profile means by family type.

326

MANDARA AND MURRAY

of parental discipline and child behavior (see Table 4). Separate univariate analyses of variance (ANOVAs) and Fisher’s protected least significant difference (LSD) test were used to examine specific family type differences. Studies have shown the Fisher’s protected LSD procedure to maintain acceptable family-wise alphas when the number of comparisons is three (Seaman, Levin, & Serlin, 1991). Also, the assumption that the typology captures more of the complexity in African American family functioning than a simple family functioning composite variable needed to be tested; thus, each analysis was repeated while controlling for the effects of the composite variable.2 This approach, multivariate analyses of covariance (MANCOVA), yielded the multivariate and univariate effects of the composite variable, as well as the effects of the differences in family type after the linear effects of the composite variable were statistically partialed. The mean scores of all the standardized family functioning variables created the composite family functioning variable. The negative variables (e.g., authoritarian and neglectful parenting, defensive and conflictive parenting) were reversed, so that the higher the score, the better the overall quality of family functioning. A scree plot revealed one very strong factor and maybe two weaker factors among the 18 variables. Therefore, we were confident that the composite score represented enough of the variability in the measures for the purpose of this study. The alpha reliability of the composite variable was good rr ⫽ .75. However, because the composite variable is not a single measure but a linear combination of several measures, the reliability estimate of a linear combination is more appropriate than Cronbach’s alpha (Nunnally & Bernstein, 1994). Therefore, the more appropriate estimate of the composite’s reliability was computed and found to be very good, ryy ⫽ .89.

variable were found. However, even after controlling for the composite variable, family type still had a significant multivariate effect on the demographic variables. All the univariate effects of family type, except for parental education, remained significant as well. These results offer some support for the argument that group differences in family type yield more important information than a simple family functioning composite variable.

The results of the MANOVA on religiosity are presented in Table 5. Family type had a significant multivariate effect on religiosity, Wilks’s lambda ⫽ .89, F(4, 186) ⫽ 2.74, p ⬍ .05, ␩2 ⫽ .05. Univariate analysis of variance tests showed that family type only had a significant effect on church attendance, but not for the importance placed on religion. Post hoc analysis showed that the Defensive–Neglectful family type was significantly less likely than the other two family types to regularly attend church or other religious institutions. There was no difference in attendance between the Cohesive and Conflictive family types. To examine the effects of the composite variable, we used the same analysis while controlling for the family functioning composite variable. This analysis resulted in a very significant multivariate effect of the composite variable on religiosity, Wilks’s lambda ⫽ .72, F(2, 92) ⫽ 17.81, p ⬍ .00, ␩2 ⫽ .28. However, the effects of family type were no longer significant when the effects of the composite were controlled, Wilks’s lambda ⫽ .93, F(4, 184) ⫽ 1.64, p ⫽ .17, ␩2 ⫽ .03. The univariate tests for the composite were both significant as well. This result implies that the composite variable is a slightly more powerful predictor of religiosity.

Demographic Characteristics of the Family Types

Parents’ Educational Expectations

The MANOVA for the demographic variables resulted in a significant multivariate effect of family type, Wilks’s lambda ⫽ .685, F(10, 154) ⫽ 3.20, p ⬍ .001, ␩2 ⫽ .179 (see Table 5). The univariate tests illustrated that there are family type differences in annual income, parental education, number of children, and marital status. There was no difference in parental age among the three family types. Post hoc analyses showed that compared with the other two family types, the Defensive–Neglectful family type had significantly less annual income, parental education, and were more likely to not be married. The Defensive– Neglectful family type also had significantly more children than the Conflictive–Authoritarian type, who had significantly more children than the Cohesive–Authoritative family type. To examine the effects of the total family functioning composite variable on the demographic variables and the effects of family type after the composite was controlled, we computed a MANCOVA using family type as the factor and the family functioning composite as the covariate. No significant multivariate, Wilks’s lambda ⫽ .953, F(5, 76) ⫽ .74, ns, ␩2 ⫽ .04, or univariate effects of the composite

A MANOVA with the two parental education expectation variables serving as the dependent variables and family type as the factor revealed a significant multivariate effect of family type, Wilks’s lambda ⫽ .90, F(4, 186) ⫽ 2.62, p ⬍ .05, ␩2 ⫽ .05 (see Table 5). Univariate tests were similar to those of the religiosity tests, because there were no family type differences in expectations of school attendance, but a significant family type difference in how far parents expect their children to go in school. Post hoc tests found that the parents in the Defensive–Neglectful family type had significantly lower expectations of how far their children would go in school compared with the other two family types. To examine the effects of the composite variable, we computed a MANCOVA on educational expectations. Results of this analysis showed that the composite family

Religiosity Characteristics

2 We also computed a principle-components analysis on the 18 family functioning variables and used the first component as the covariate instead of the nonweighted linear composite. The results with the weighted composite were very similar to the ones with the nonweighted composite. Therefore, we only discuss the results of the nonweighted composite in the manuscript.

AFRICAN AMERICAN FAMILY TYPOLOGY

327

Table 5 Family Demographic and Attitudinal Means, Percentages and Univariate Tests for Family Types Family type means Variable Demographics Parents’ age Parental education Family income No. children % married Religiosity Religious importance Church attendance Educational expectations Educational importance Educational expectations Discipline and child behavior Parental discipline Child disobedience

Cohesive (n ⫽ 44)

Conflictive (n ⫽ 40)

Defensive (n ⫽ 25)

F(2, 103)

p

␩2

40.94 8.36a 7.16a 2.11c 57.90a

39.68 7.76a 6.45a 2.76b 63.20a

39.86 6.18b 4.19b 3.15a 27.30a

0.34 4.36 7.39 5.89 4.10

.71 .01 .00 .02 .00

.00 .09 .14 .08 .12

3.68 3.42a

3.71 3.45a

3.57 2.76b

0.34 4.70

.71 .01

.00 .09

4.00 4.82a

4.00 4.87a

3.95 4.50b

1.84 4.19

.16 .01

.04 .08

3.37a 3.59c

3.21b 4.50b

2.77c 5.32a

7.70 4.53

.00 .01

.14 .09

Note. Different subscripts within a row indicate significantly different means.

functioning variable had only a marginally significant multivariate effect on educational expectations, Wilks’s lambda ⫽ .95, F(2, 92) ⫽ 2.55, p ⫽ .08, ␩2 ⫽ .05. In like manner, the effects of family type were only marginally significant after controlling for the effects of the composite variable, Wilks’s lambda ⫽ .92, F(4, 184) ⫽ 2.06, p ⫽ .08, ␩2 ⫽ .04. Neither of the univariate tests for the composite family functioning variable or family type was significant. This set of results implies that family type is more powerful than the composite variable in predicting educational expectations, as its original multivariate effects were significant. Moreover, even when controlling for the composite variable, family type still accounted for a marginally significant amount of variability in educational expectations.

We used another MANCOVA to assess the effects of the composite family functioning variable on the discipline and behavior variables. As with the demographic results, the results of this analysis showed that there were no significant multivariate effects, Wilks’s lambda ⫽ .99, F(2, 93) ⫽ .36, ns, and no significant univariate effects of the composite variable. However, the multivariate differences in family type remained significant, as did both of the univariate tests. This set of analyses also illustrates the relatively superior power of the family type variable compared with the composite variable in predicting subjective evaluations of parental discipline practices and adolescent behavior, as well as the importance of group differences over and above what was accounted for by the composite family functioning variable.

Discipline and Adolescent Behavior We also computed a MANOVA on the two variables regarding how well parents think they discipline their adolescent and how well the adolescent thinks he or she obeys. The results of this analysis are presented in Table 5, and they revealed a significant multivariate effect of family type, Wilks’s lambda ⫽ .82, F(4, 188) ⫽ 5.02, p ⬍ .001, ␩2 ⫽ .10. The two subsequent univariate tests also revealed significant family type differences in how well parents discipline their children, as well as adolescents’ perceptions of how well they obey. Post hoc tests revealed the same consistent trend found in the previous results. The Defensive–Neglectful family type was significantly lower than the other two family types in perceptions of adolescent discipline. Also, the adolescents in the Cohesive–Authoritative family type perceive themselves as obeying significantly more than those in the Conflictive–Authoritarian and the Defensive–Neglectful family types. There was no significant difference found between the Defensive and Conflictive family types.

Self-Esteem To assess the effects of family type on adolescent selfesteem, we computed another MANOVA using the 10 selfesteem variables as the dependent variables and family type as the factor. Family type had a strong multivariate effect on adolescent self-esteem, Wilks’s lambda ⫽ .47, F(20, 194) ⫽ 4.48, p ⬍ .001, ␩2 ⫽ .32. The univariate tests showed that family type was a significant predictor of all 10 self-esteem measures (see Table 6). The post hoc tests revealed a very consistent difference in self-esteem among the three family types. The adolescents in the Cohesive– Authoritative family type perceived their self-esteem to be consistently and significantly higher than the adolescents in the other two family types (see Figure 3). Adolescents in the Conflictive–Authoritarian family type were also significantly higher than those in the Defensive family type on most of the self-esteem measures. To assess the effects of the composite variable on selfesteem, we computed another MANOVA on the 10 self-

328

MANDARA AND MURRAY

Table 6 Self-Esteem Means and Univariate Tests for Family Types Family type means Self-esteem measure

Cohesive (n ⫽ 44)

Conflictive (n ⫽ 40)

Defensive (n ⫽ 25)

F(2, 106)

p

␩2

Total Global Self-approval Lovability Likability Self-control Self-identity Power Competence Defensive Body function

3.74a 4.05a 3.88a 3.97a 3.74a 3.80a 3.76a 3.87a 3.83a 2.83a 3.74a

3.49b 3.67b 3.50b 3.36b 3.62a 3.58a 3.55a 3.80a 3.71a 2.54b 3.59a

3.19c 3.36b 3.21c 3.23c 3.23b 3.18b 3.20b 3.47b 3.28b 2.67a 3.26b

19.76 7.05 15.97 26.50 9.41 9.32 7.44 4.39 11.76 4.47 5.00

.00 .00 .00 .00 .00 .00 .00 .01 .00 .01 .01

.27 .12 .23 .33 .15 .15 .12 .07 .18 .08 .09

Note. Measures of self-esteem are from the Multi-Dimensional Self-Esteem Inventory. Different subscripts within a row indicate significantly different means.

esteem measures, with the composite serving as the factor. The composite family functioning variable also had a highly significant multivariate effect on self-esteem, Wilks’s lambda ⫽ .61, F(10, 101) ⫽ 6.47, p ⬍ .001, ␩2 ⫽ .39. The univariate tests showed that the composite significantly predicted all the self-esteem measures except defensiveness. However, the more important issue is how much the family type variable accounts for after the composite is controlled. Therefore, we computed a MANCOVA on the 10 self-esteem measures with family type as the factor and the composite as the covariate. The findings supported the hypothesis that family type accounts for information in well-being not accounted for by the composite family func-

tioning variable. The multivariate effects of the composite were no longer significant, Wilks’s lambda ⫽ .91, F(10, 96) ⫽ 0.95, p ⫽ .49, but the multivariate effects of family type were still significant, Wilks’s lambda ⫽ .73, F(20, 192) ⫽ 1.47, p ⬍ .05, ␩2 ⫽ .14. Furthermore, ANCOVAs showed that the typology significantly predicted all 10 self-esteem variables with the composite variable controlled.

Ethnic Identity To assess the effects of family type on adolescent ethnic identity, we computed another MANOVA on the two ethnic

Figure 3. Adolescent self-esteem profiles (from the Multi-Dimensional Self-Esteem Inventory) by family type.

AFRICAN AMERICAN FAMILY TYPOLOGY

identity variables. Family type had a significant multivariate effect, Wilks’s lambda ⫽ .69, F(12, 198) ⫽ 3.36, p ⬍ .001, ␩2 ⫽ .17. The univariate tests showed that family type was a significant predictor of 3 of the 6 ethnic identity variables (see Table 7). Post hoc tests revealed trends that were very consistent with the differences in self-esteem among the three family types. The adolescents in the Cohesive–Authoritative family type had significantly higher value for African American culture, and more ethnicspecific behaviors compared with the adolescents in the other two family types (see Figure 4). Furthermore, the adolescents in the Cohesive family type had a more affirmed identity than those in the Defensive family type. Adolescents in the Conflictive family type were also significantly higher than those in the Defensive family type on these same variables. As with the previous analyses, we then computed another MANOVA using the composite family functioning variable as the factor. The composite had a strong multivariate effect on ethnic identity, Wilks’s lambda ⫽ .77, F(6, 102) ⫽ 4.96, p ⬍ .001, ␩2 ⫽ .22. The composite significantly predicted the same three individual variables as did the typology. We then computed a MANCOVA on the ethnic identity variables, with the typology variable serving as the factor and the composite serving as the covariate. The Wilks’s lambda for the composite was no longer significant, Wilks’s lambda ⫽ .94, F(6, 97) ⫽ 0.87, p ⫽ .52, ␩2 ⫽ .05. The typology still had a moderately significant multivariate effect on ethnic identity, Wilks’s lambda ⫽ .82, F(12, 194) ⫽ 1.60, p ⫽ .09, ␩2 ⫽ .09. However, the more important issue is that the effect of the typology with the composite controlled did not significantly decrease when the composite was controlled.

Personality To assess the effects of family type on adolescent personality, we computed another MANOVA on the five personality variables. Family type again had a strong multivar-

329

iate effect, Wilks’s lambda ⫽ .77, F(10, 204) ⫽ 2.88, p ⬍ .002, ␩2 ⫽ .124. The univariate tests showed that family type was a significant predictor of three of the five variables (See Table 8). Post hoc tests revealed that the adolescents in the Cohesive–Authoritative family type had significantly higher levels of agreeableness and emotional stability compared with the adolescents in the other two family types (see Figure 5). Furthermore, the adolescents in the Cohesive family type were assessed as more conscientious than those in the Defensive family type. Adolescents in the Conflictive family type were also significantly higher than those in the Defensive family type on these same three variables. We then computed another MANOVA using the composite family functioning variable as the factor. The composite had a strong multivariate effect on personality as well, Wilks’s lambda ⫽ .73, F(5, 106) ⫽ 7.98, p ⬍ .001, ␩2 ⫽ .27. Univariate tests showed that the composite significantly predicted the same 3 personality variables as did the typology. We then computed a MANCOVA on the personality variables, with the typology variable serving as the factor and the composite serving as the covariate. The Wilks’s lambda for the composite remained significant, Wilks’s lambda ⫽ .87, F(5, 101) ⫽ 2.88, p ⫽ .018, ␩2 ⫽ .125. The multivariate effects of the typology were no longer near significance, Wilks’s lambda ⫽ .94, F(10, 202) ⫽ 0.66, p ⫽ .76, ␩2 ⫽ .03. Furthermore, none of the univariate effects of the typology remained significant.

Reliability of the Typology The final purpose of this study was to assess the reliability of the derived typology. To accomplish this, we used two separate procedures recommended by previous researchers (Asendorp, 2000; Breckenridge, 2000; Milligan, 1996). The first was a cross-validation procedure used to assess the cluster replication or stability of the cluster solution across samples (Breckenridge, 2000). There are five steps to the

Table 7 Ethnic Identity Means and Univariate Tests for Family Types Family type means Ethnic identity measure ASCR Total ASCR Cultural Awareness Anti-ASC Value for culture Group Identification MEIM Total MEIM Affirmed identity Achieved identity Ethnic behaviors

Cohesive (n ⫽ 43)

Conflictive (n ⫽ 39)

Defensive (n ⫽ 24)

F(2, 103)

p

␩2

3.45a 3.84 2.69 4.18a 3.22

3.39a 3.91 2.71 3.84b 3.23

3.23b 3.69 2.92 3.65c 3.03

4.12 .88 1.18 6.69 .89

.02 .42 .31 .00 .41

.07 .01 .02 .11 .01

3.36a 3.73a 3.15 3.17a

3.21a,b 3.61a 3.01 2.94a

3.01b 3.31b 3.01 2.27b

5.64 5.79 1.06 12.17

.01 .00 .35 .00

.09 .10 .02 .19

Note. Ethnic identity variables are from the African Self-Consciousness Measure, Revised (ASCR) and the Multigroup Ethnic Identity Measure (MEIM). Different subscripts within a row indicate significantly different means. ASC ⫽ African Self-Consciousness.

330

MANDARA AND MURRAY

Figure 4. Adolescent ethnic identity profiles (from the African Self-Consciousness Measure, Revised [ASCR] and the Multigroup Ethnic Identity Measure [MEIM]) by family type. ASC ⫽ African Self-Consciousness.

procedure. First, the data are randomly divided into two samples. Second, a full cluster analysis is performed on sample A. Next, a full cluster analysis is performed on sample B. Fourth, sample B is classified into clusters according to the cluster centers derived from sample A. Finally, the agreement is computed between the two sample B solutions. The more agreement between the two solutions, the more reliable or stable is the cluster solution. We used Cohen’s kappa as a measure of agreement. Following these procedures, we randomly divided the sample into two subsamples (ns ⫽ 55 and 56). We then performed a hierarchical cluster analysis using betweengroups linkage and squared Euclidian distances on Sample A. The dendrogam clearly identified three main clusters with one or two very small clusters. As before, we then computed a k-means cluster analysis, selecting four groups.

One group had only 1 subject, so we dropped this “group” from all other replication analyses, as before. We then repeated this procedure with Sample B. Again, we ended up with three major groups. We then computed a noniterated k-means cluster analysis on Sample B using the cluster centers from sample A. As suggested, we then compared the two solutions from Sample B on agreement. Table 9 presents a 3 ⫻ 3 contingency table. The values along the diagonal represent cases assigned to the same clusters in both solutions. The agreement between the two solutions was very high (␬ ⫽ .82). As a further test of the replicability of the solution, we then computed a noniterated k-means with Sample A while using the cluster centers from sample B. Although not as strong as the sample B solutions, the agreement between the two Sample A solutions was also good (␬ ⫽ .65). We also

Table 8 Personality Means and Univariate Tests for Family Types Family type means Five-factor personality dimension

Cohesive (n ⫽ 44)

Conflictive (n ⫽ 40)

Defensive (n ⫽ 25)

F(2, 106)

p

␩2

Extraversion Agreeableness Conscientiousness Emotional Stability Openness to Experiences

4.88 5.28a 5.07a 5.34a 4.75

4.87 4.72b 4.92a 4.80b 4.82

4.76 4.50c 4.58b 4.16c 4.78

0.13 6.15 4.40 8.41 0.04

.88 .00 .02 .00 .96

.00 .10 .07 .13 .00

Note. Different subscripts within a row indicate significantly different means.

AFRICAN AMERICAN FAMILY TYPOLOGY

331

Figure 5. Adolescent personality profiles (five-factor model) by family type.

compared the solutions derived from Samples A and B with the original full sample solution. Both samples had high agreement with the full-sample solution (␬ ⫽ .69 and .82, respectively). As a further test of the reliability of the cluster solution we compared the full sample solution at Time 1 with data collected after a 1-year lag. A sample of 68 families remained from the initial assessment. To determine whether the typology exhibits consistency over time, we compared the subtype that families were assigned to at the first assessment with the second assessment classification. The initial cluster centers (profile means) from the original cluster solution were subjected to iterated k-means cluster analysis to derive the clusters at Time 2. A 3 ⫻ 3 (Time 1 clusters ⫻ Time 2 clusters) contingency table is presented in Table 10. An examination of the diagonal cells reveals the number of correctly classified families for each type at Time 2. Overall, the typology was very consistent over time, ␹2(4, N ⫽ 68) ⫽ 65.13, p ⬍ .001.

Table 9 Cross-Validation of African American Family Typology Cluster from Subsample B 1 2 3

27 3

2 12 2

Discussion The purpose of this study was to identify types of African American families, and to determine whether the types differed on important demographic and attitudinal characteristics and adolescent self-esteem. A further purpose was to assess the consistency of the typology over time. Using

Table 10 One-Year Lag Reliability of the African American Family Typology

Cluster from Subsample A ⫻ B 1

Cohen’s kappa coefficient was high at .77. Almost 90% of the Cohesive–Authoritative group, 68% of the Conflictive– Authoritarian group, and 80% of the Defensive–Neglectful group were correctly classified. This finding illustrates the stability of the typology over time. However, the average test–retest reliability of the individual variables was relatively low (␣ ⫽ .52). This illustrates the relative stability of the typology compared with the individual variables used to create it. Taken together, these results imply that the cluster solution is very reliable.

Time 2

3 1 9

Note. A ⫻ B refers to clusters derived from Subsample B using Subsample A centroids. ␹2(4, N ⫽ 54) ⫽ 71.83, p ⬍ .001, r ⫽ .92.

Time 1

Cohesive

Conflictive

Defensive

Cohesive Conflictive Defensive

25 3 3

2 17 0

1 5 12

Note. ␹2(4, N ⫽ 68) ⫽ 65.13, p ⬍ .001, r ⫽ .70, ␬ ⫽ 77.

332

MANDARA AND MURRAY

cluster-analytic procedures, we obtained an empirically derived family functioning typology of three types of African American families. According to their most salient family functioning dimensions, the family types were named Cohesive–Authoritative, Conflictive–Authoritarian, and Defensive–Neglectful. The characteristics of and differences between the family types make sound theoretical and empirical sense, and they are conceptually similar to the findings of Baumrind (1991). We also found that the families differed on important variables such as income, parental education, marital status, religiosity, educational expectations, adolescent discipline, behavior, self-esteem, ethnic identity, and personality development, even after the linear effects of overall family functioning were controlled for most variables. However, for some variables such as personality, the effects of the typology were totally accounted for by the composite variable. This implies that the typology adds important information not accounted for by a linear composite of the family functioning variables, but this is not the case for all variables. Another important finding was that the typology was very stable over time, even though the variables used to create the typology were relatively low in reliability. On average, about 87% of the families remained in the same type after a 1-year lag. Furthermore, a cross-validation test of the typology revealed strong support for the replicability and stability of the typology. This is strong evidence that the classification of the families may reflect a real typology in the population (Breckenridge, 2000). A detailed discussion of family characteristics is presented below.

Cohesive–Authoritative Family Type The most salient features of the cohesive–authoritative are family cohesion and expressiveness, authoritative disciplinary practices, and a very low level of family conflict. Similar to Baumrind’s (1991) type with the same name, the families of this type also emphasize personal growth and development of family members, as evidenced by their high family recreational activities, intellectual– cultural discussions, and encouragement of family members to be assertive and think things out for themselves. They are high on proactive racial socialization and low in defensive racial socialization. The Cohesive–Authoritative family type is significantly lower than the other two family types in authoritarian and neglectful disciplinary styles. This suggests that families of this type are unlikely to be overly controlling and critical or to express unhappiness with their adolescent’s performance and abilities. They also express a great deal of concern and emotion for their children’s wellbeing. Families of this type place a moderate level of emphasis on moral and religious socialization. They are also distinguished from the other two groups by their balanced emphasis on the extent to which activities are focused on achievement and the extent to which the family system is based on rigid parental control. However, even though they do not exhibit rigidity in rules and procedures, organization

within the family is an important characteristic of this family type. Married parents head about 58% of families of this type. The parents in this family type are also higher in formal education compared with the other groups. These families average about two children per household. These three factors may speak to the families’ comparatively higher annual income (about $32,000). Regular church attendance and high educational expectations for their children also mark this family type. The parents in the Cohesive– Authoritative family type feel that they do a good job of disciplining their adolescents, and their children report that they obey much more than adolescents in the other two family types. Overall, this family type exhibits a very high level of family functioning compared with the other two types.

Conflictive–Authoritarian Family Type Extremely high levels of internal conflict, lack of concern and commitment toward members, and overbearing authoritarian discipline styles distinguish this type of family. The Conflictive–Authoritarian family type is marked by distressed family relationships, in which children feel that they are not allowed or encouraged to express their feelings. The Conflictive–Authoritarian family type places a great deal of emphasis on achievement but not necessarily on intellectual stimulation. On the basis of their other characteristics, the focus of such families on achievement is probably different than that of the cohesive family type, and activities are most likely competitive in nature. Similar to Baumrind’s conception of authoritarian families, this group is very controlling and allows little expressiveness or questioning of family rules. They do not focus on family recreational activities and are not too concerned with political, social, or intellectual activities. Religious-oriented socialization is also a focus of this family type. Although families of this type are demanding and controlling and place great emphasis on achievement, they are significantly lower than the cohesive– authoritative family type in family organization and higher in neglectful parenting. This means that they exhibit conflicting characteristics of controlling and overbearing discipline and achievement-oriented activities while being unorganized and neglectful. This family type is also relatively moderate in their focus on racial socialization. The demographics of the Conflictive–Authoritarian family type are very similar to that of the cohesive family type. About 63% of the homes are headed by married parents, and their average annual income is about $27,000. The parents have relatively moderate levels of education and average about 2.7 children per household. They have high regard for religion and attend church often. School is also important to this family type, and they have high expectations for their children’s educational attainment. The parents of families of this type feel that they discipline their children in an average manner, but the children in these families feel that they obey significantly less than the children in the cohesive family type. This may be because the 15-year-old adolescents in

AFRICAN AMERICAN FAMILY TYPOLOGY

333

this family type are beginning to rebel against their authoritarian upbringing.

has minimal effects on family functioning (Mandara & Murray, 2000).

Defensive–Neglectful Family Type

Family Type and Self-Esteem

This final type of family is the smallest in the sample and possibly at the greatest risk for becoming dysfunctional. The most salient characteristics of this family type are their extremely high levels of defensive racial socialization and extremely low levels of empowering racial socialization. The parents socialize their children to dislike other racial groups while simultaneously not teaching them to be proud to be African Americans. In comparison to the cohesive family type, they exhibit high levels of neglectful parenting. Therefore, children in these homes probably receive fewer hugs and kisses from parents and are more likely to stay out as late as they desire, and the parents do not express as much concern for their children as the parents in the other two family types. They are similar to the Conflictive– Authoritarian family type because they are high on authoritarian discipline and low on authoritative disciplinary styles. Therefore, parents are not nurturing or supportive, yet they are very critical. Another distinguishing characteristic of this family type is the low importance they place on personal growth and development. They exhibit extremely low emphasis on achievement and do not stress political, intellectual, or cultural activities, nor do the families actively participate in recreational or sporting activities. The Defensive– Neglectful families are also significantly lower in religiousoriented and moral socialization compared with the other two family types. This is consistent with the result that they attend church significantly less than the other two types. Even though they exhibit high levels of neglect and a general lack of family organization and structure, the children of these families experience significantly lower levels of intellectual independence. Therefore, the children are not encouraged to be assertive or to think things out for themselves. Although not to the same extent as the Conflictive family type, interfamily relationships within these families are relatively chaotic. Such families are low on family cohesion and expressiveness and high on internal conflict and the extent to which their homes are organized in a rigid, hierarchical manner. Single mothers head 73% of the Defensive–Neglectful families. The families average about 3 children per household, yet their average annual income is below $20,000. In like manner, the parents in this family type have significantly lower education than the other two types of families. Although they believe education is important, the parents have very low educational expectations for their children. The parents in these homes feel that they discipline their children much less than the other family types, and their children feel that they disobey much more as well. Overall, chaotic and dysfunctional family processes mar the Defensive–Neglectful family type. It should also be clarified that the high percentage of nonmarried parents in this family type does not fully explain their family functioning patterns. Previous research has shown that family structure

Most researchers conclude that parental communication and closeness, supportive and nurturing parenting styles, and parents’ acceptance of the views and values of their children have positive effects on European American youths’ self-esteem (Cooper et al., 1983; Gecas, 1972; Isberg, Hauser, Jacobson, & Powers, 1989; Rogers, 1951). The present study found that this might also be true for African American youth. This study found that the adolescents in the Cohesive–Authoritative family type had significantly higher self-esteem than the adolescents in the other two types. This is most likely due to the finding that the parents in the Cohesive family type are more likely to express to their children that they are appreciated and valued the way they are and because the other two family types are more likely to express unhappiness with their children’s abilities and performance. Furthermore, the Cohesive family type seems to have found a balance of control and nurturance in the way that Baumrind theorized would be most beneficial to children and adolescents. The higher levels of racial socialization of the Cohesive–Authoritative family type may also contribute to the relatively higher self-esteem found among adolescents in these families (Murray & Mandara, 2001). The adolescents in the Conflictive–Authoritarian family type also had significantly higher self-esteem compared with the adolescents in the Defensive family type. This finding is probably due to the very low self-esteem among adolescents in the Defensive family type, rather than to high self-esteem among adolescents in the Conflictive type. This would not be surprising considering that the adolescents in the Defensive family are not monitored as well as they should be, their parents do not expect much from them or nurture them, and their family life in general is very chaotic compared with the other two family types.

Family Type and Ethnic Identity Several recent investigations have found very consistent trends regarding the effects of a positive ethnic identity, particularly when race is a salient feature of one’s selfconcept (Phinney, 1992). For instance, studies have shown that African Americans with more developed ethnic identities scored significantly higher on measures of self-concept (Wilson & Constantine, 1999), personal self-esteem (Goodstein & Ponterotto, 1997; Murray & Mandara, in press), collective self-esteem and general communal orientations (Azibo, 1991; Baldwin, Brown, & Rackley, 1990), problem-focused coping (Neville, Heppner, & Wang, 1997), anti-drug attitudes in high school students (Resnicow, Soler, Braithwaite, Selassie, & Smith, 1999), and academic achievement (Phinney, 1992). Given these findings, the results of the present study are not surprising. Adolescents in the Cohesive family type tended to behave

334

MANDARA AND MURRAY

in ethnic-specific ways and to value those behaviors more than those in the other two groups. This is not surprising considering that the Cohesive family type uses more proactive and less defensive and reactive racial socialization than the other two family types. Considering that the adolescents in this group have higher self-esteem, are better behaved, and are more emotionally stable than adolescents in the other two groups and because ethnic identity is positively related to well-being (Phinney, 1992), one can understand why the parents in families of this type use these socialization strategies. Because of the correlational nature of this study, it is difficult to assess causation, but it may be that the high levels of defensive racial socialization used and the lack of exposure to positive African American images and activities in the defensive type is one major reason why the adolescents in this family type suffer from so many psychosocial problems. We have presented previous arguments for this rationale (Murray & Mandara, 2001).

Family Type and Adolescent Personality The results of the present study illustrate that adolescents in the different family types have rather different personalities. The adolescents in the Cohesive family type are similar to those of the other two groups on extraversion and openness to experience. They are highly agreeable, which indicates that they are good-natured and trusting. Because of their high levels of conscientiousness, they are probably more organized, reliable, and self-disciplined than the other two groups. Furthermore, they are higher on emotional stability, which suggests that they do not worry as much and that they are calm and more secure with themselves than adolescents in the other two groups. This pattern is consistent with their self-esteem and ethnic identity. According to any standard interpretation of a personality profile, the adolescents in this family type would be considered well-adjusted. Like their self-esteem profile, the Conflictive adolescent’s personality profile reflects a group not high or low on most factors. They are not having problems as a group, but they are not doing as well as they could be. They are a bit low on agreeableness, which means they may be somewhat cynical and uncooperative with others at times, but they are not as low on agreeableness as the Defensive type. An interesting finding is that even with all of their parents’ control and focus on achievement, the adolescent’s in this group are not very high on conscientiousness. It would be interesting to see whether they are actually higher in conscientiousness than they perceive themselves, and whether it is the case that because their parents rarely show their satisfaction with them, these adolescents do not perceive themselves to be as organized and hard-working as they actually are. The personality profile of the adolescents in the Defensive family type is fairly consistent with the rest of their profiles. They are doing very poorly. Because of their very low levels of agreeableness, they perceive themselves to be highly cynical, rude, and uncooperative.

Because of their very low self-esteem, they are very defensive to criticism and suggestions. Their very low levels of conscientiousness reflect that of individuals who are unreliable, lazy, and unorganized. However, probably the most serious characteristic of these adolescents is their extremely low levels of emotional stability, suggesting that they worry a great deal and are nervous and insecure. The adolescents in this family type are undoubtedly at the greatest risk for any number of negative outcomes compared with the other adolescents.

Implications for Research, Application, and Public Policy Although this is only a first attempt at creating a reliable and empirically valid typology of African American family functioning, it nevertheless has profound implications for family researchers, therapists, and policymakers. For instance, the present typology can be the beginning of a detailed and scientific mapping of African American family functioning patterns. After the typology is further modified and validated by replicating these findings on larger and more diverse samples and by assessing its relationship with other variables, researchers can begin to develop secondand third-level classifications, similar to those in the physical and biological sciences. Therefore, once the first-level typology is more stable, researchers can then examine the subtypes within each major type (Bergman, 2000; McQuitty, 1987). Although the typology created here is in its initial stages of development and is far from having clinical significance, it holds promise for future research. Moreover, unlike many others, the present family typology was concerned with understanding the family functioning patterns of African Americans and not with the predictors of various outcome variables. Therefore, family functioning variables were not selected because they correlate with outcomes but because they help describe family functioning patterns. This allows for a more accurate description of African American families. The fact that the typology predicts several important psychosocial factors, even though the family functioning variables were not specifically chosen to correlate with them, supports the argument that these are real types in the population and that these are the real outcomes expected. However, more research needs to be done before that claim can be made definitively. Another advantage is that the present study used 18 different family functioning variables to create the typology. Because of this, very specific details of the families can be systematically described. The more information researchers have about each type of African American family, the more precisely they will be able to describe normative African American behavior and group processes and the easier it will be for social service providers and therapists to identify at-risk families and create interventions. Accurately predicting at-risk children before problems occur is undoubtedly the best way to prevent future negative outcomes. The children in the Defensive–Neglectful family type may

AFRICAN AMERICAN FAMILY TYPOLOGY

need to be identified so that preventative treatment with them and their families can be offered before any major problems occur. This study also indicates that conceptualizations of family functioning based on predominately European American families do not adequately describe African American family processes but that there is a great deal of similarity between the groups. For instance, Baumrind’s (1991) three prototypical types (authoritative, authoritarian, and neglectful) are very similar to those in the present typology. However, because the present study involved more family functioning dimensions, included variables of specific importance to African American families, and was empirically derived, the description of the families is more detailed than what can be derived from Baumrind’s typology. Furthermore, it is still not clear whether each variable means similar things to African American and European American families. Because of the cultural specificity of the present typology, many of those issues are avoided.

Limitations This study has several limitations. First, a few of the measures used to create the typology were not highly reliable. Therefore, this should be taken into consideration when interpreting the data. Although the typology itself was very stable over time, better measures need to be developed before the specific details of the typology can be reliably described, which will increase the stability and usefulness of the typology. Another primary limitation is that there were only 111 families, and they were all from southern California. Future studies need to attempt to replicate this typology with a larger and more geographically diverse sample of African American families. Larger samples are also needed to assess the subtypes within each major type. Furthermore, with a much larger sample, higher order interactions with income, marital status, and other important demographic variables can also be examined to assess their interactive effects on various outcomes.

Conclusion Now that years of variable-centered family research has uncovered several dimensions of family functioning, the typological or case-centered approach offers a method of synthesizing this information (Bergman et. al., 2000). Even with its limitations, the typology presented in this article offers an example of the value of the typological approach for describing people, families, or other groups, and for answering scientific questions. The next step is to validate the present typology by assessing differences among the families on several outcome measures and to replicate this typology on larger and more geographically diverse samples of African Americans. Once the typology is replicated and thoroughly validated, it may help researchers begin to explicate the causal connections between family functioning and behavior.

335

References Aguilar, M. L., Kaiser, R. T., Murray, C. B., & Ozer, D. J. (1998). Validation of an adjective Q-sort as a measure of the big five personality structure. Journal of Black Psychology, 24, 145–163. Aldenderfer, M. S., & Blashfield, R. K. (1984). Cluster analysis. Beverly Hills, CA: Sage. Anderson, C. A., & Sedikides, C. (1991). Thinking about people: Contributions of a typological alternative to associationistic and dimensional models of person perception. Journal of Personality and Social Psychology, 60, 203–217. Asendorpf, J. B. (2000). A person-centered approach to personality and social relationships: Findings from the Berlin Relationship Study. In L. R. Bergman, R. B. Cairns, L. Nilsson, & L. Nystedt (Eds.), Developmental science and the holistic approach (pp. 281–298). Mahwah, NJ: Erlbaum. Azibo, D. A. (1991). An empirical test of the fundamental postulates of an African personality metatheory. Western Journal of Black Studies, 15, 183–195. Baldwin, J. A., & Bell, Y. R. (1985). The African Self-Consciousness Scale: An Afrocentric personality questionnaire. The Western Journal of Black Studies, 9, 61– 68. Baldwin, J. A., Brown, R., & Rackley, R. (1990). Some sociobehavioral correlates of African self-consciousness in AfricanAmerican college students. Journal of Black Psychology, 17, 1–17. Baumrind, D. (1968). Child care practices anteceding three patterns of preschool behavior. Genetic Psychology Monographs, 75, 43– 88. Baumrind, D. (1972). An exploratory study of socialization effects on Black children: Some Black–White comparisons. Child Development, 43, 261–267. Baumrind, D. (1989). Rearing competent children. In W. Damon (Ed.), Child development today and tomorrow (pp. 349 –378). San Francisco, CA: Jossey-Bass. Baumrind, D. (1991). The influence of parenting style on adolescent competence and substance use. Journal of Early Adolescence, 11, 56 –95. Bergman, L. R. (1988). You can’t classify all of the people all of the time. Multivariate Behavioral Research, 23, 425– 441. Bergman, L. R. (2000). The application of a person-oriented approach: Types and clusters. In L. R. Bergman, R. B. Cairns, L. G. Nilsson, & L. Nystedt (Eds.), Developmental science and the holistic approach (pp. 137–154). Mahwah, NJ: Erlbaum. Bergman, L. R., Cairns, R. B., Nilsson, L. G., & Nystedt, L. (2000). Developmental science and the holistic approach. Mahwah, NJ: Erlbaum. Block, J., & Block, J. H. (1980). The California Child Q-Set. Palo Alto, CA: Consulting Psychologists Press. Block, J. (2000). Three task for personality psychology. In L. R. Bergman, R. B. Cairns, L. G. Nilsson, & L. Nystedt (Eds.), Developmental science and the holistic approach (pp. 155–164). Mahwah, NJ: Erlbaum. Block, J., & Ozer, D. J. (1982). Two types of psychologists: Remarks on the Mendelsohn, Weiss, and Feimer contribution. Journal of Personality & Social Psychology, 42, 1171–1181. Breckenridge, J. N. (2000). Validating cluster analysis: Consistent replication and symmetry. Multivariate Behavioral Research, 35, 261–285. Brody, G., & Flor, D (1997). Maternal psychological functioning, family processes, and child adjustment in rural, single-parent, African American families. Developmental Psychology, 33, 1000 –1011. Brooks, A., Stuewig, J., & LeCroy, C. (1998). A family based

336

MANDARA AND MURRAY

model of Hispanic adolescent substance use. Journal of Drug Education, 28, 65– 86. Burchinal, M., Campbell, F., Bryant, D., & Wasik, B. (1997). Early intervention and mediating processes in cognitive performance of children of low-income African American families. Child Development, 69, 935–954. Chao, R. K. (1994). Beyond parental control and authoritarian parenting style: Understanding Chinese parenting through the cultural notion of training. Child Development, 65, 1111–1119. Clark, R., & Armstead, C. (2000). Preliminary study examining relationship between family environment and resting mean arterial pressure in African-American youth. Journal of Adolescent Health, 27, 3–5. Cooper, J. E., Holman, J., & Braithwaite, V. A. (1983). Selfesteem and family cohesion: The child’s perspective and adjustment. Journal of Marriage and the Family, 45, 153–159. Crouter, A. C., & Manke, B. (1997). Development of a typology of dual-earner families: A window into differences between and within families in relationships, roles, and activities. Journal of Family Psychology, 11, 62–75. Dancy, B. L., & Handal, P. J. (1984). Perceived family climate, psychological adjustment, and peer relationship of Black adolescents: A function of parental marital status or perceived family conflict? Journal of Community Psychology, 12, 222–229. Fisher, L., & Ransom, D. C. (1995). An empirically derived typology of families: I. Relationships with adult health. Family Process, 34, 161–182. Gecas, V. (1972). Parental behavior and contextual variations in adolescent self-esteem. Sociometry, 35, 332–345. Glasgow, K. L., Dornbusch, S. M., Troyer, L., Steinberg, L. & Ritter, P. L. (1997). Parenting styles, adolescents’ attributions, and educational outcomes in nine heterogeneous high schools. Child Development, 68, 507–529. Goodstein, R., & Ponterotto, J. G. (1997). Racial and ethnic identity: Their relationship and their contribution to self-esteem. Journal of Black Psychology, 23, 275–292. Gorman-Smith, D., Tolan, P. H., Loeber, R., & Henry, D. B. (1998). Relation of family problems to patterns of delinquent involvement among urban youth. Journal of Abnormal Child Psychology, 26, 319 –333. Gorman-Smith, D., Tolan, P. H., Zelli, A., & Huesmann, R. L. (1996). The relations of family functioning to violence among inner-city minority youths. Journal of Family Psychology, 10, 115–129. Heiss, J. (1996). Effects of African American family structure on school attitudes and performance. Social Problems, 43, 246 –265. Helms, J. E. (1990). Black and White racial identity: Theory, research, and practice. New York: Greenwood Press. Isberg, R. S., Hauser, S. T., Jacobson, A. M., Powers, S. I. (1989). Parental contexts of adolescent self-esteem: A developmental perspective. Journal of Youth and Adolescence, 18, 1–23. Johnson, P., & McNeil, K. (1998). Predictors of developmental task attainment for young adults from divorced families. Contemporary Family Therapy: An International Journal, 20, 237–250. Lamborn, S. D., Dornbusch, S. M., & Steinberg, L. (1996). Ethnicity and community context as moderators of the relations between family decision making and adolescent adjustment. Child Development, 67, 283–301. Lustig, D. C. (1997). Families with an adult with mental retardation: Empirical family typologies. Rehabilitation Counseling Bulletin, 41, 138 –157. Mandara, J., & Murray, C. B. (2000). The effects of parental marital status, family income, and family functioning on African American adolescent self-esteem. Journal of Family Psychology, 14, 475– 490.

McQuitty, L. L. (1987). Pattern-analytic clustering: Theory, method, research and configural findings. Lanham, MD: University Press of America. Milligan, G. W. (1996). Clustering validation: Results and implications for applied analyses. In P. Arabie, L. J. Hubert, & G. De Soto (Eds.), Clustering and classification (pp. 341–375). River Edge, NJ: World Scientific. Moos, R. H., & Moos, B. S. (1976). A typology of family social environments. Family Process, 15, 357–371. Moos, R. H., & Moos, B. S. (1986). Family Environment Scale manual. Palo Alto, CA: Consulting Psychologists Press. Murray, C. B., & Mandara, J. (2001). Racial identity development in African American children: Cognitive and experiential antecedents. In H. P. McAdoo (Ed.), Black children (pp. 73–96). Thousand Oaks: Sage. Murray, C. B., & Mandara, J. (in press). An assessment of racial socialization and ethnic identity as predictors of self-esteem. Annals of Afrocentric Psychology. Neville, H. A., Heppner, P. P., & Wang, L. (1997). Relations among racial identity attitudes, perceived stressors, and coping styles in African American college students. Journal of Counseling & Development, 75, 303–311. Norusis, M. J. (1994). SPSS Professional Statistics 6.1. Chicago: SPSS. Nunnally, J. C., & Bernstein, I. H. (1994). Psychometric theory. New York: McGraw-Hill. O’Brien, E. J., & Epstein, S. (1988). The Multidimensional SelfEsteem Inventory: Professional manual. Odessa, FL: Psychological Assessment Resources. Peacock, M. J., Murray, C. B., Ozer, D., & Stokes, J. (1996). The development of the Black Family Process Q-Sort. In R. Jones (Ed.), Handbook of tests and measurements for Black populations (Vol. 4, pp. 475– 493). Hampton, VA: Cobb & Henry. Phinney, J. S. (1992). The multigroup ethnic identity measure: A new scale for use with diverse groups. Journal of Adolescent Research, 7, 156 –176. Pomerantz, E. M., & Ruble, D. N. (1998). The multidimensional nature of control: Implications for the development of sex differences in self-evaluation. In J. Heckhausen & C. S. Dweck (Eds.), Motivation and self-regulation across the life span (pp. 159 –184). New York: Cambridge University Press. Resnicow, K., Soler, R. E., Braithwaite, R. L., Selassie, M. B., & Smith, M. (1999). Development of a racial and ethnic identity scale for African American adolescents: The Survey of Black Life. Journal of Black Psychology, 25, 171–188. Roberts, L. J., & Leonard, K. E. (1998). An empirical typology of drinking partnerships and their relationships to marital functioning and drinking consequences. Journal of Marriage and the Family, 60, 515–526. Rogers, C. R. (1951). Client-centered therapy: Its current practice, implications, and theory. Boston: Houghton Mifflin. Seaman, M. A., Levin, J. R., & Serlin, R. C. (1991). New developments in pairwise multiple comparisons: Some powerful and practicable procedures. Psychological Bulletin, 110, 577–586. Starrels, M. (1994). Gender differences in parent– child relations. Journal of Family Issues, 15, 148 –165. Steinberg, L., Dornbusch, S. M., & Brown, B. B. (1992). Ethnic differences in adolescent achievement: An ecological perspective. American Psychologist, 47, 723–729. Steinberg, L., Lamborn, S. D., Darling, N., Mounts, N. S., & Dornbusch, S. M. (1994). Over-time changes in adjustment and competence among adolescents from authoritative, authoritarian, indulgent, and neglectful families. Child Development, 65, 754 – 770.

AFRICAN AMERICAN FAMILY TYPOLOGY Stokes, J. E., Murray, C. B., Peacock, M. J., & Kaiser, R. T. (1994). Assessing the reliability, factor structure, and validity of the African Self-Consciousness Scale in a general population of African Americans. Journal of Black Psychology, 20, 62–74. Stouthamer-Loeber, M., & Loeber, R. (1988). Parents as intervention agents for children with conduct problems and juvenile offenders. Child and Youth Services, 11, 127–148. Taylor, R. D., & Wang, M. C. (1997). Social and emotional adjustment and family relations in ethnic minority families. Mahwah, NJ: Erlbaum. Thornton, M. C. (1997). Strategies of racial socialization among Black parents: Mainstream, minority, and cultural messages. In R. T. Taylor, J. S. Jackson, & L. M. Chatters (Eds.), Family life in Black America (pp. 201–215). Thousand Oaks, CA: Sage.

337

U.S. Bureau of the Census. (1996). Marital status and living arrangements: March 1994 (Current Population Reports, Series P-20, No. 484). Washington, DC: U.S. Government Printing Office. Weiss, L. H., & Schwarz, J. C. (1996). The relationship between parenting and older adolescents’ personality, academic achievement, adjustment, and substance use. Child Development, 67, 2101–2114. Wilson, J. W., & Constantine, M. G. (1999). Racial identity attitudes, self-concept and perceived family cohesion in Black college students. Journal of Black Studies, 29, 354 –366.

Received April 14, 2000 Revision received October 22, 2001 Accepted October 30, 2001

Correction to Kenny and Acitelli (1994) The article “Measuring Similarity in Couples,” by David A. Kenny and Linda K. Acitelli (Journal of Family Psychology, Vol. 8, No. 4, pp. 417– 431), contained an error. In the first paragraph on page 424, the formula to test whether the set of correlations significantly differs from one another is described incorrectly. The correct formula (see Snedecor & Cochran, 1989, pp. 186 –187) is to first transform the correlations with a Fisher’s z transformation and to then compute the mean value of z. For each correlation, one then computes the deviation of z from the mean z, squares that deviation, and then sums the squares. This quantity is multiplied by k ⫺ 5, where k is the number of items used to compute the correlation. Under the null hypothesis that the correlations do not differ, the resulting quantity has a chi-square distribution with N ⫺ 3 degrees of freedom, where N is the number of couples. The test results reported in the article are unaffected as they appear to be based on the correct formula, rather than the incorrect one presented in the article.

Reference Snedecor, G. W., & Cochran, W. G. (1989). Statistical methods (6th ed.). Ames: Iowa State University Press.

䡲