BEHAVIOR THERAPY 12, 461-472 (1981)
Engagement in Behavioral Family Therapy: A Multisite Investigation ELAINE A. B L E C H M A N Wesleyan University K A R E N S. B U D D
EDWARD R. CHRISTOPHERSEN
University of Nebraska Medical Center
University of Kansas Medical Center
STEVEN S Z Y K U L A
ROBERT W A H L E R
Family Teaching Center, Montana
University of Tennessee, Knoxville
L Y N N E H . EMBRY
KATE KOGAN
University of Kansas, Lawrence
University of Washington, Seattle
K. DANIEL O'LEARY
L I N D A S. R I N E R
State University of New York, Stonybrook
Universi~ of Nebraska Medical Center
For 2 years, eight sites providing behavioral family treatment collected data about demographic characteristics, the type of intervention provided, and the record of treatment attendance, yielding data on 181 families. A stepwise discriminant analysis found four functions which distinguished between locations: family communication training, home token economy, marital intervention, and parent support group, and accounted for 76% of between-locations variance. These functions describe naturally occurring clusters of treatment and family characteristics, suggesting that properties of the family ecology were taken into account when type of behavioral family intervention was selected. A second discriminant analysis contrasted families engaged (74%) and not engaged (26%) in treatment and found that engaged families tended to receive family communication training (often together with other types of intervention) and to have fathers and mothers with high occupational prestige, fathers who worked many hours, and two natural parents. Taken together these results suggest that evaluation of a behavioral family intervention must consider background family characteristics as well as the magnitude of family behavior change.
Preparation of this paper was partly supported by NIMH grant 31403 to the first author. Discussions about the effects on treatment of demographic characteristics took place during the 1976 meeting of the Pre-AABT Family Social-Learning Group and led to the organization of this study during the group's 1977 and 1978 meetings. In consultation with the other authors, the first author designed the survey questionnaire, analyzed the resulting data, and prepared this paper for publication. All authors collected at their home sites the research data which is the foundation of this paper. Requests for reprints should be sent to Elaine A. Blechman, Department of Psychology. Wesleyan University, Middletown, CT 06457. 461 0005-7894/81/0461~)47251.00/0 Copyright 1981by Associationfor Advancementof BehaviorTherapy All rightsof reproductionin any form reserved.
462
BLECHMAN ET AL.
Blechman (1981), devised an algorithm to guide the choice of a behavioral intervention for a client family. Instead of routinely using family socioeconomic status to select the intervention modality, current family behavior and family member priorities would be the deciding factors. In current practice, family characteristics may determine both the type of behavioral intervention provided to families and the family's subsequent level of engagement in treatment. The low income one-parent family has repeatedly been described as hard to keep in parent contingency management training (Eyberg & Johnson, 1974; Fleischman, Note 1; Patterson, 1974). A survey of researchers of parent contingency management found that recipients of this intervention were predominantly middle class, and that many authors saw no reason to document such vital family statistics as per capita income and parent education (Hargis & Blechman, 1979). This study searched for the implicit rules which determine the assignment of a family to a type of behavioral family intervention and for the characteristics which distinguish families who are successfully engaged in the treatment process. Among the background family characteristics of interest was one that has frequently been advanced as a likely determinant of family competence and of child adjustment--number of parents.
METHODS Eight applied research groups, each engaged in the development of methods of behavioral family intervention, cooperated in the execution of this study. Five of the groups participated in both 1978 and 1979. Participants in 1978 were Connecticut, Tennessee, Montana, Nebraska, Kansas I, New York, and Washington. Participants in 1979 were the first five locations and Kansas II. The first author designed a questionnaire 1 to collect the following types of information about families applying for mental health services: demographic characteristics of family members and the family group, descriptions of the types of behavioral intervention offered by the clinic to the family, and ratings of the family's subsequent record of treatment attendance. The group leaders met in December of 1977 and 1978 to agree on the final format of the questionnaires. In February of 1978 and in January of 1979 final versions were distributed by mail. Group leaders agreed to complete a questionnaire for a specific number of consecutive clients (10 in 1978, 20 in 1979) seen for intake in their clinics. The completed questionnaires were returned by mail to the first author in September of 1978 and 1979 for data analysis. The 1979 questionnaire incorporated the entire 1978 questionnaire and added to it questions about the outcome of treatment. The added questions have been repeated in a 1980 version now being administered. With the exception of one item, data resulting from these questions are not discussed in this paper. 1 A complete questionnaire may be obtained from the first author.
ENGAGEMENT IN BEHAVIORAL FAMILY THERAPY
463
Questions completed by treatment personnel requested information about the target child, the reason for the referral, the child's parents and siblings, the types of treatment offered to the family, the family's response to the offer of treatment, and the persistence of the family in treatment. A question which appeared only in the 1979 version elicited a rating of treatment outcome from a staff member. Mothers' and fathers' stated occupations were assigned occupational prestige ratings ranging from 1 to 100 (Duncan, 1961).
RESULTS In 1978, 71 questionnaires were returned from: Nebraska (10), Montana (10), New York (10), Kansas I (12), Washington (10), Tennessee (9), and Connecticut (10). In 1979, 110 questionnaires were returned from: Connecticut (20), Tennessee (19), Montana (20), Nebraska (21), Kansas I (10), Kansas II (20). In 1978, 74% of the target children were boys, 73% were from two-parent families, 96% were white, and 47% had been referred for treatment by a public agency or private practitioner. The mean family income was $12,040. Children's ages ranged from 4.36 to 11.6 years, with a mean age of 7.4 years. The presenting problems most often mentioned by parents were: school (24%), disturbed parent-child relationships (20%), and antisocial behavior (17%). In 1979, 73% of the target children were boys, 61% were from two-parent families, 79% of families were white, and 39% had been referred for treatment by a public agency or private practitioner. The mean family income was $11,580. Children's ages ranged from 3.5 to 8.8 years with a mean age of 6 years. The presenting problems most often mentioned by parents were disturbed parentchild relations (29%), school (21%), and antisocial behavior (20%). Discriminating between locations. Group leaders' self-reports indicated that locations were best distinguished by the types of treatment provided, rather than by types of families served. Validation of this assumption seemed a necessary first step. To determine which variables best distinguished between locations, a stepwise discriminant function analysis was executed with 1978 and 1979 locations as the grouping dependent variable (Cooley & Lohnes, 1971; Klecka, 1975). The 31 independent variables in the analysis were the result of 17 questionnaire items concerning family background characteristics, the family's presenting problem, and the type of treatment provided to the family. Thus one question about type of treatment provided 12 binary responses concerning all the types of treatment a family received. The stepwise criterion employed for the selection of variables was the overall F ratio for the test of differences among the group centroids (i.e., the group means on a discriminant function). The objective of the analysis was to find meaningful discriminant functions which minimized Wilks' Lambda and maximized the multivariate F ratio. The independent variables were entered simultaneously, and no order of entry was specified because no assumptions
464
BLECHMAN ET AL.
about causal priority were justified. For example, high income families might have gravitated to certain locations to live and subsequently seek treatment, i.e., income might have temporally preceded location; or, certain groups might have inadvertently recruited high income families, i.e., location might have temporally preceded income. Four functions and twenty-seven constituent variables discriminated between locations yielding the multivariate F(27,142) = 5.26 and accounting for 75.86% of the between-locations variance. Weights of variables loading on these functions indicate that locations were best distinguished by treatment types. Function 1, family communication training (eigenvalue = 6.81, Wilks' Lambda = .002), accounted for 41.39% of the between-locations variance. Function 2, home token economy (eigenvalue = 2.25, Wilks' Lambda = .007), accounted for an additional 13.69% of variance. Function 3, marital intervention (eigenvalue = 1.89, Wilks' Lambda = .019), accounted for an additional 11.48% of variance. Function 4, parent support group (eigenvalue = 1.53, Wilks' Lambda = .05), accounted for an additional 9.3% of variance. Table 1 lists the 27 variables, their standardized coefficients on the 4 functions, and the locations' means on these variables. Table 2 displays the locations' scores (group centroids) on the four discriminant functions. As Table 1 shows, positive scores on family communication training indicate one or more of the following: provision of family communication training, provision of teacher contingency management, families with many children, female target children. Negative scores on family communication training indicate provision of parent contingency management or of school token economy. As Table 2 shows, locations with extreme positive scores on family communication training were: Connecticut 1978, 1979. Locations with extreme negative scores on family communication training were: Nebraska 1978, Kansas II 1979, and Montana 1979. Positive scores on home token economy indicate one or more of the following: provision of home token economy, well-educated mothers, families presenting child physical disability as a complaint, white families, older target children. Negative scores on home token economy indicate: provision of parent-child contingency contracting and families presenting problems other than those listed on the questionnaire. Locations with extreme positive scores on Function 2 were: Montana 1978, 1979, and New York 1978; locations with extreme negative scores were: Kansas I 1978, Tennessee 1978, and Kansas II 1979. Positive scores on marital intervention indicate one or more of the following: fathers with high occupational prestige, a presenting problem of inadequate parenting behavior. Negative scores on Function 3 indicate: provision of marital intervention, provision of parenting education, families with high income. Locations with extreme positive scores on Function 3 were: Washington 1978 and Kansas II 1979; locations with extreme negative scores were: Tennessee 1978, 1979, and Montana 1978, 1979. Positive scores on parent support group indicate: provision of parenttraining group, fathers and mothers who work many hours per week,
-
-.56 .35
Marital intervention Parenting education Father's occupational prestige Year's gross income~ Problem: Inadequate parenting behavior 52 17.4 16%
0
92% 0 1.9 20%
0 0 2.5 40%
61 11.7
8%
100%
0 75%
17%
100%
0 0
12
10
0
32 12.7
0 20%
0 40% 3.5 20%
0
20%
10
MT
0
50 14.3
0 80%
90% 0 2.5 30%
0
0
10
NE
0
48 15.3
0 50%
30% 10% 2.9 20%
20%
0
10
NY
55%
36 12.8
22% 56%
0 0 2.6 50%
0
22%
9
TN
0
57 --
0 0
0 0 2.4 30%
0
0
10
WA
0
65 13.9
0 0
0 0 2.1 35%
95%
100%
20
CT
Home token economy 0 8.3% 90% 20% 20% 0 0 0 Parent-child contracting 0 8% 10% 0 10% 33% 0 100% Mother's years school 12.5 15.0 11.7 1 1 . 6 1 3 . 3 11.8 12.7 1 2 . 3 Problem: Other 0 0 0 0 0 0 0 0 Problem: Physical disability 0 10% 0 0 0 0 0 0 Race: White 100% 92% 1 0 0 % 90% 100% 100% 1 0 0 % 47% Age of target child/months 109 117 112 67 140 73 52 113
KS I
4.19 1.00 .74 -.32 .26 .26 • 15
-3.81
4.67
3.25
-3.35 1.94 .62
-
CT
1978
- 1.60 • 13 • 15
Location
95% 25% 1.9 35%
15%
15%
20
21% 0 2.4 15%
0
0
20
62% 0 2.5 79%
0
10%
21
NE
0 7% 1.9 26%
7%
71%
19
TN
0
66 15.2
10% 60%
5%
66 12.5
5% 80%
0
66 10.3
6% 82%
5%
69 13.4
0 57%
31%
66 10.5
0 73%
10% 25% 94% 19% 14% 20% 25% 0 10% 73% 1 2 . 6 1 4 . 7 11.9 1 2 . 5 11.7 20% 5% 0 0 0 0 0 0 0 0 1 0 0 % 90% 1 0 0 % 95% 90% 53 49 85 64 73
80% 10% 1.8 43%
20%
0
10
MT
1979 KS I KS II
Unstandardized means and percentages
Variable N Family communication training Teacher contingency management Parent contingency management School token economy Number of children Target child: Girl
Standardized coefficient
TABLE 1 LOCATION MEANS AND VARIABLE COEFFICIENTS FOR THE FOUR BETWEEN-LOCATIONS FUNCTIONS
c)
o ,"
= ~ ~ =o o
."=
Function name
t~
"o
;=
F
> E
r-
> <
.}
3-
m
Parent support group Parent training group Hours/week father works Hours/week mother works Mother's age/years Mother's occupational prestige Problem: Social relationships Father's age/years Father's years school
Location
17% 33 15
38
35
0 33 13
0 0 40 22 30
KS I
0 0 16 16" 32
CT
0 35 14
38
0 20% 28 28 33
MT
0 32 15
25
0 40% 29 14 29
NE
1978
10% 37 14
48
0 0 0 4 37
NY
44% 34 13
48
0 0 28 11 29
TN
0 35 15
47
0 70% 32 10 31
WA
0 36 12
63
0 0 26 18 32
CT
30% 30 16
68
0 0 26 12 30
KS I
15% 31 14
59
0 50% 16 11 28
0 33 13
68
0 24% 23 7 30
MT
1979 KS II
Unstandardized means and percentages
5% 34 15
70
0 0 22 8 30
NE
43% 32 12
59
50% 0 10 9 30
TN
Function name
Note. (CT = Connecticut, KS I = Kansas I, MT = Montana, NE = Nebraska, NY = New York, TN = Tennessee, WA = Washington, KS II = Kansas II). a In multiples of $1,000.
-.50 .50
.55
-5.74 2.90 .80 .68 .57 -.66
dardized coe~ ficient
Stan-
TABLE1 CONTINUED
> z
rrn
4:-
ENGAGEMENT IN BEHAVIORAL FAMILY THERAPY TABLE
467
2
LOCATION MEANS ON THE FOUR BETWEEN-LOCATION FUNCTIONS
Function
Location
Family communication training I
Home token economy
Marital intervention
Parent support group
II
III
IV
1978
Connecticut Kansas I Montana Nebraska New York Tennessee Washington
5.29 - 1.77
.36 -2.13
.74
.39
- . 16
1.42
.44
1.72
- 1.00
1.71
- 2.09
- . 05
.52
I. 37
.01
1.84
.03
- 1.56
-2.32
.64
.31
.20
2.11
2.05
5.05
- .36
- . 15
-2.93
1979
Connecticut Kansas I Kansas I I Montana Nebraska Tennessee
- . 13
.71
- 1.09
-.09
.65
-2.55
- 1.47
1.73
- 1.36
2.86
- .54
.15
-.07 - 1.25
.24 - 2.56
.05 - .67
-2.18 - 1.16 .54
-.55
older mothers, families presenting problems of disturbed child social relationships, well educated fathers. Negative scores on Function 4 indicate: provision of parent support group, mothers with high occupational prestige, older fathers. Locations with extreme positive scores on Function 4 were: Montana 1978, Kansas I 1978, Nebraska 1978; locations with extreme negative scores were: New York 1978, Kansas II 1979. Five of the eight sites participated in the study in 1978 and 1979 (Connecticut, Kansas I, Montana, Nebraska, Tennessee). Table 2 shows considerable stability in the placement of these sites in respect to the four discriminant functions. Nebraska and Tennessee scored in the same direction on the four functions during both years. Connecticut and Kansas I scored in the same direction on two of the four functions. Discriminating between engaged and nonengaged families. Each family was rated by treatment personnel in respect to good treatment attendance, and in 1979 the success of treatment was rated by staff. These scores were not used to group families---zbias was likely and sites varied considerably in respect to the time of year when clients entered treatment and questionnaires were completed. Two questionnaire items with multiple-choice answers provided information about treatment type and about treatment disposition. A composite score was constructed which would indicate whether a client family had become engaged in treatment by being offered a specific treatment and accepting it. Compared to ratings by treatment personnel, this score was less likely to be biased by
468
BLECHMAN ET AL.
TABLE 3 ENGAGED VS NONENGAGED GROUP MEANS AND VARIABLE COEFFICIENTS ON THE BETWEEN-GRouPs FUNCTION
Standardized coefficients 1.77 - 1.41 -.49 -.48 -.48
.47 .31 .25 .23 .21 -. 17
Unstandardized meansb treatment and percentages
Variablesa Family communication training + Assessment + Problem: Antisocial child behavior Other intervention + Parent-child contracting + Problem: Symptomatic child behavior Father's occupational prestige rating Mother's occupational prestige rating Hours father works Number natural parents Number children
Engaged in treatment (133)
Not engaged in treatment (48)
36% 8% 14% 10% 33% 2% 60 (22) 52 (19) 43 (10) 1.7 2.23 (1.02)
7% 10% 4% 10% 5% 6% 50 (23) 51 (20) 43 (6) 1.4 2.51 (1.43)
a Variables followed by a + were used as indicators of engagement. The complete list of variables used to score engagement is given in Table 4. b Standard deviations are shown in parentheses beside means.
enthusiasm about the local brand of intervention, and it was less likely to be influenced by seasonal variations from site to site. Families were classified as engaged (n = 133) if they had received one or more of the interventions n a m e d in Table 4, and if they had not declined treatment, been rated as having poor attendance in treatment, or been referred elsewhere. Families were classified as not engaged (n = 48) if they received none of the interventions listed in Table 4 and if they were not offered treatment, declined the offer of treatment, showed poor attendance in treatment, or were referred elsewhere. A stepwise discriminant function analysis was executed with engagement in treatment as the grouping dependent variable. The 41 independent variables in the analysis were the product of 16 multiple-choice questionnaire items about type of intervention, family background, and type of presenting problem. T y p e of intervention was included because the majority o f families who received treatment received more than one type o f intervention, and because the analysis aimed to determine which types of intervention predicted engagement. Treatment-related variables not included in the analysis are listed in Table 4. One discriminant function, engagement (eigenvalue = .41, Wilks' L a m b d a = .71) produced the multivariate F(10,170) = 6.28. The 11 constituent variables and their standardized coefficients are shown in Table 3. Engaged families had a positive score (.38) on engagement, nonengaged had a negative score ( - 1 . 0 6 ) . Differences between engaged and nonengaged families on con-
E N G A G E M E N T IN B E H A V I O R A L F A M I L Y T H E R A P Y
469
TABLE 4 MEANSb OF ENGAGED AND NONENGAGED CLIENTS ON TREATMENT VARIABLES
Variable~ Weeks in treatment~ 1979 staff rating of success Treatment not offered - Declined offer of treatment Poor attendance in treatment Good attendance in treatment Completed treatment to date - Referred elsewhere Other disposition of case + Parent contingency management + Parenting education + Teacher contingency management + Home token economy + School token economy + Parent support group + Parent training group + Marital intervention + Family communication training + Assessment + Other intervention + Parent-child contracting -
-
Engaged (133) 9.39 (8.35) 27.53 (36.9) 0 0 0 80% 52% 0 13% 47% 53% 28% 24% 8% 6% 15% 4% 36.4% 7.6% 9.8% 33.1%
Not engaged (48) 6.05(7.87) F(1, 165) = 4.8, p < .05 10.58(21.9) F(1, 139) = 5.07, p < .05 4% 28% 51% 6% 13% 11% 28% 42% 42% 2% 24% 5% 5% 17% 0 7.3% 9.8% 10% 4.9%
a Variables preceded by a minus sign were used as indicators of nonengagement in treatment; variables preceded by a + sign were used as indicators of engagement. Engaged families had one or more plus and no minus. Nonengaged families had one or more minus and no plus. b Standard deviations are listed in parentheses following means. c Variables not included in the Engagement analysis are preceded by no - or + sign.
s t i t u e n t v a r i a b l e s are s h o w n in T a b l e 3. E n g a g e d families t e n d e d to rec e i v e f a m i l y c o m m u n i c a t i o n t r a i n i n g , c o m p l a i n o f s y m p t o m a t i c child beh a v i o r , h a v e f a t h e r s a n d m o t h e r s w i t h high o c c u p a t i o n a l p r e s t i g e , h a v e f a t h e r s w h o w o r k m a n y h o u r s , a n d h a v e a high n u m b e r o f n a t u r a l p a r e n t s . N o n e n g a g e d families t e n d e d to r e c e i v e a s s e s s m e n t , c o m p l a i n o f a n t i s o c i a l child b e h a v i o r , r e c e i v e a n i n t e r v e n t i o n o t h e r t h a n t h o s e specified in T a b l e 3, r e c e i v e p a r e n t - c h i l d c o n t r a c t i n g a l o n e ( r a t h e r t h a n t o g e t h e r with family c o m m u n i c a t i o n t r a i n i n g as e n g a g e d f a m i l i e s did), a n d h a v e m a n y c h i l d r e n . O f the 60 c h i l d r e n with o n e p a r e n t , 100% had o n e n a t u r a l p a r e n t . O f the 116 c h i l d r e n with t w o p a r e n t s , 6% (7) h a d n o n a t u r a l p a r e n t , 16.4% (19) h a d o n e n a t u r a l p a r e n t , a n d 77.6% (90) h a d t w o n a t u r a l p a r e n t s . T h e p e r c e n t o f c l i e n t s e n g a g e d in t r e a t m e n t r a n g e d a c r o s s l o c a t i o n s f r o m 48% to 100% w i t h a m e a n o f 74% e n g a g e d in t r e a t m e n t . B e c a u s e l o c a t i o n s v a r i e d g r e a t l y in r e s p e c t to t y p e o f t r e a t m e n t p r o v i d e d , a n d to
470
BLECHMAN
ET AL.
a lesser extent in respect to average client family characteristics, it would be inappropriate to conclude that one site (or treatment type) was generally more effective than another.
DISCUSSION This study suggests that behavioral family therapy is currently being provided to families from diverse social and economic strata. Parents' income and education varied with clinic location from poverty level to middle class; over all locations the average family was lower middle class. The between-location functions represent natural covariations of treatment and family types. They may be the products of self-selection by families into treatment sites, purposeful or unplanned selection of clients by treatment personnel, or the demographic characteristics of the community surrounding the treatment sites. The matching of families and treatments cannot be the sole result of the demographics of the surrounding communities since all sites provided more than one type of treatment to more than one type of family. While it is unclear how the matching occurred, the properties of the family ecology were taken into account when behavioral family intervention was provided. In general, family communication and teacher contingency management were provided to slightly larger families in which the target child was a girl; home token economy to relatively well educated mothers; marital intervention and parenting education to relatively high income families; parent-training groups to families in which both parents worked long hours and mothers were relatively older; parent support groups to mothers with high occupational prestige and relatively older fathers. The natural matches between treatments and families found in this study appear to have rational grounds. Mothers with high occupational prestige and long working hours may have little chance to share information about their adequacy as parents: They received parent training and support groups. Well educated mothers probably follow through easily on homework assignments: They were involved in home token economies. The recipients of marital intervention and parenting education were often high income f/tmilies. It remains to be determined why this natural treatment (marital intervention) by family (high income) match occurred. High income couples may be more willing to discuss their discontent, they may experience more conflict if their high income is the product of two working parents, or their resources may permit the lengthier treatment dictated by a combined marital-parenting focus. Engaged families were best distinguished from nonengaged families by a small cluster of treatment and family type variables. It proved easier to distinguish reliably between locations than between engaged and nonengaged families. The between-locations analysis reduced Wilks' Lambda far more than did the between-families analysis. At the same time, the composite engagement score appears meaningful--it has the expected relationships with treatment length, treatment completion, and staff rating of treatment success. The fact that both parents' occupational prestige
ENGAGEMENT IN BEHAVIORAL FAMILY THERAPY
471
ratings predict engagement is consistent with evidence that in the normal population, children's life success is best predicted by parent income, occupation, and education (Mosteller & Moynihan, 1972). It is not clear how behavior of families headed by high and low occupational status parents systematically differs, affecting receptivity to behavioral intervention. Wahler (1980) may have bridged the gap between static family characteristics and intervention-related parent behavior. He has described low income socially isolated mothers with predominantly coercive contacts with people outside the family and aversive reactions to their oppositional children. These mothers and their husbands would undoubtedly score low on measures of occupational prestige, education, and income. Their dilficulties in child rearing, their low receptivity to intervention, and their lack of friendly and supportive adult relationships are intertwined conditions. Therefore, successful intervention with these families may demand more than the provision of child-rearing skills. The cluster of variables distinguishing engaged from nonengaged families supports this notion--engaged families received family communication training along with other more focused interventions. This study is the first to use multivariate analysis to assess the impact of family composition on engagement in behavioral family treatment. Locations did not differ in their provision of intervention to one- or twoparent families. Number of natural parents predicted engagement in treatment, but it was the least important of six predictors of engagement, while parents' occupational prestige was the most important. Thus a single working mother with high occupational prestige was more likely to be engaged in treatment than two parents both with low occupational prestige. This is consistent with other multivariate research (Blechman, in press).
REFERENCE NOTE 1. Fleischman, M. J. Using parenting salaries to control attrition and cooperation in therapy. Unpublished manuscript, Oregon Social Learning Center, undated.
REFERENCES Blechman, E . A . Are children with one parent at psychological risk?: A methodological review. Journal of Marriage and the Family, in press. Blechman, E . A . Toward comprehensive behavioral family interventions: An algorithm for matching families and interventions. Behavior Modification, 1981, 5, 221-236. Cooley, W. W., & Lohnes, P . R . Multivariate data analysis. New York: Wiley, 1971. Duncan, O. D. A socioeconomic index for all occupations. In A. J. Reiss, Jr. (Ed.), Occupations and social status. New York: Free Press of Glencoe, 1961. Eyberg, S. M., & Johnson, S. M. Multiple assessment of behavior modification with families: Effects of contingency contracting and order of treated problems. Journal of Consulting and Clinical Psychology, 1974, 42, 594-606. Hargis, K., & Blechman, E . A . Social class and training of parents as behavior change agents. Child Behavior Therapy, 1979, 1, 69-74.
472
BLECHMAN ET AL.
Klecka, W . R . Discriminant analysis. In N. H. Nie, C. H. Hull, J. G. Jenkins, K. Steinbrenner, & D. H. Bent (Eds.), SPSS: Statistical package for the social sciences. New York: McGraw-Hill, 1975. Mosteller, R., & Moynihan, D. P. (Eds.), On equality of educational opportunity: Papers deriving from the Harvard University faculty seminar on the Coleman Report. New York: Vintage, 1972. Patterson, G . R . Intervention for boys with conduct problems: Multiple settings, treatments, and criteria. Journal of Consulting and Clinical Psychology. 1974, 42,471-481. Wahler, B . G . The insular mother: Her problems in parent-child treatment. Journal of Applied Behavior Analysis, 1980, 13, 207-219. RECEIVED: August 9, 1980 FINAL ACCEPTANCE:February 19, 1981