Adolescents in therapeutic communities.

The literature examining success rates for milieu treatments of antisocial and conduct-disordered adolescents has been dominated by pessimism (Macallair, 1993). The "nothing works" position has provided little useful information for clinical practice and has cast an aura of "diagnostic gloom" (Wolff, 1984) over an increasingly large number of troubled youth (Wells, 1991a; Zimmerman, 1990).

It has been suggested elsewhere that the wholesale condemnation of residential treatment models is unwarranted (Curry, 1991; Small, Kennedy, & Bender, 1991). Upon closer examination, the literature is rife with methodological problems so that existing studies are generally inconclusive (Pfeiffer & Strzelecki, 1990; Wells, 1991b). Criteria for judging the success of client outcomes are poorly defined and lack standardization (Beschner & Friedman, 1985; Curry, 1991; Durkin & Durkin, 1975; Lewis & Summerville, 1991). Many studies rely on self-report data (Bale, 1979; Reed, 1978) or use single global indices of improvement that yield little meaningful information. The most common design is retrospective, with no reporting of baseline rates (Velasquez & Lyle, 1985). Overall, there is a sparsity of empirical work so the whole literature is dominated by theorizing rather than research activity, and research that has been performed has lagged behind the advances in other related fields (Curry, 1991). The complexity of treatments, the ethical drawbacks to random assignment, the tendency of client groups to be small and heterogenous, and the considerable latitude given to front-line staff in applying the treatment have all served to discourage planned research. One of the greatest obstacles to research has been that it is only recently that front-line staff have begun to see themselves as skilled professionals providing treatment that can be evaluated (DeLeon & Rosenthal, 1979).

This study was designed with the primary goal of examining the response to treatment in a Therapeutic Community for conduct-disordered adolescents. This model was originally designed for a milieu of drug addiction (Bratter, Collabolletta, Fossbender, Pennchia, & Rubel, 1985; DeLeon & Rosenthal, 1979) and stresses confrontation and self-governance by the clients themselves (Almond, 1974; Lennard & Allen, 1974). While Therapeutic Community treatment has been applied to adolescents, and initial results have been positive (Bescher & Friedman, 1985) research on this model is especially sparse. This study also represents an attempt to overcome a number of methodological flaws traditionally associated with research in residential treatment. Treatment outcome was compared for two theoretically distinct programs, the comparison of which was facilitated through the use of environmental measures. Multiple measures of client change were incorporated and administered at three separate intervals in the treatment process in a repeated measures design. Baseline rates were utilized to judge the comparability of the two treatment groups at admission.

METHOD

Study Setting

This study was conducted in two different publicly funded residential treatment environments. One program is a Therapeutic Community located within a large general psychiatric hospital (hereafter referred to as TC). It stresses self-regulation, peer confrontation, and the development of insight as necessary for behavior change. The second program is part of a public psychosocial (nonmedical) agency that provides out-of-home care and treatment to nondelinquent adolescents. It is run on behavior modification principles and uses a point system as the basis for a Token Economy. The staff at this unit (hereafter referred to as BU) expect behavior to change as a function of reinforcements and consequences.

In order to obtain an objective index of the differences between the two programs, the Correctional Institutions Environment Scale (Moos, 1974) was administered to all staff in both programs. This scale has been used specifically to assess noncommunity-based programs for juveniles and yields information on three major dimensions of a social environment: (1) the nature and intensity of relationships, (2) personal growth and self-enhancement influence, and (3) system maintenance and change dimensions. These dimensions comprise social climate (Moos, 1974) which is considered to be a critical focus for comparative research (Ainsworth & Fulcher, 1981). The most prominent difference between the two program profiles was the relative importance of Autonomy and Staff Control. TC staff rated Autonomy as important and placed less emphasis on Staff Control. Bu did just the opposite. According to a typology of treatment environments developed by Moos (1977), the profile suggest that TC is, in fact, a Therapeutic Community type program, while BU qualifies as a Control-Oriented environment.

Subjects

The total sample consisted of 28 conduct-disordered adolescents (14 females, 14 males) between the ages of 14 and 18. The mean age in the two programs was 15.2 (BU) and 15.9 (TC), respectively. The majority of clients in both programs were Caucasian, however BU had a larger percentage of black clients (25% as compared with 6.3%). The group studied represents all consecutive admissions to the two treatment programs over a 16-month period, excluding four adolescents who refused to participate in the study. Seven adolescents dropped out of treatment during the course of the study but were retained for purposes of comparison.

Instrumentation

In keeping with the literature (Achenbach, 1982; Durkin & Durkin, 1975; Galliland-Mall & Judd, 1986; Lewis & Summerville, 1991; Manning, 1972; Pfeiffer & Strzelecki, 1990) on the relevance of client input factors and the need for multiple and diverse outcome criteria, the data were collected in a number of ways. The following six instruments were used: (1) Wechsler Intelligence Scale for Children-Revised (Wechsler, 1974); (2) Jesness Inventory (Jesness, 1972), a 155-item personality inventory which yields a profile of characteristics predictive of social, emotional, and behavioral problems of disturbed and delinquent youth. Reliability coefficients for the subscales range from .55 to .79 for test-retest reliability, and from .64 to .88 for split-half reliability (Jesness, 1972); (3) Conceptual Level Paragraph Completion Test (Hunt, Butler, Ney, & Rosser, 1979),. This semi-projective instrument places individuals along a continuum of conceptual complexity and capacity for self-responsibility and independence (Turner, 1985). Reitsma-Street (1982) cites test-retest reliability coefficients of .67 over 3 months and inter-rater reliability at over .80; (4) Internal Locus of Control Scale (Levenson, 1981), a refinement of Rotter's I-E scale which yields scores on three aspects of control: belief in internal control (internality), powerful others, and chance. Reliability coefficients for test-retest reliability are quoted as .66, .62, and. 73, respectively over 7 weeks (Levenson, 1981); (5) Tennessee Self-Concept Scale (Fitts, 1965), a commonly used measure of self-esteem which yields 11 scores on aspects of self-concept. Test-retest reliability coefficients range from .74 to .82 over 2 weeks; (6) the Devereux Adolescent Behavior Rating Scale (Spivack, Spotts & Haines, 1967), which profiles 15 aspects of symptomatic behavior based on staff ratings. Reliability coefficients of .82 and .42 have been cited for test-retest reliability and inter-rater reliability, respectively (Buros, 1972).

In addition, demographic data and background information were obtained from each subject's dossier because these factors are also thought to influence treatment response (Durkin & Durkin, 1975; Lewis, 1985).

Procedure

Data were collected in three steps: (1) just prior to admission; (2) three months after admission; (3) six months after admission. While not all youths under study were discharged at the termination of the research, it was seen as preferable to compare clients at standard time intervals. All instruments were used during the pretest phase to obtain baseline scores.

The behavior rating scale, the Locus of Control scale, and the self-concept measure were repeated as posttest measures because all were designed to reflect change in treatment and appear to be sensitive to change (Galliland-Mall & Judd, 1986; Paulhaus & Christie, 1981; Levenson, 1981).

Statistical Analysis

Statistical analysis was performed using SPSSX (1986). For descriptive purposes, cross tabulations were performed to test for demographic differences between treatment groups (TC and BU), and dropouts vs. nondropouts. In addition, a one-way analysis of variance was used to determine whether these groups differed in pretreatment levels of performance on the standardized tests.

Analyses in a repeated measures design (S(A)xB) were performed to examine the degree of change over time. There was one within-group factor, time, and one between-group factor, treatment group.

RESULTS

The following is a summary of significant findings. The significance level was set at .05 for all tests. The TC group differed from the BU group on background and demographic characteristics in that: (1) there were more males in the group ([[Chi].sup.2] = 3.646, p [less than] .05); (2) there were fewer clients from large families ([[Chi].sup.2] = 4.87, p [less than] .05); (3) there were more clients with a history of previous out-of-home placements ([[Chi].sup.2]= 9.70, p [less than] .03); and (4) clients had more experience with psychotropic medication ([[Chi].sup.2] = 8.40, p [less than] .03).

Both treatment group profiles at admission revealed clinically deviant scores on many of the variables. In terms of baseline scores, there were the following significant differences: (1) TC had less disturbed scores on Domineer-Sadistic (Devereux Scale 3), a measure of aggressive dominance of peers, F = 5.52, p [less than] .03; (2) TC had less disturbed scores on Heterosexual Interest (Devereux Scale 4), a measure of displayed interest in the opposite sex, F = 4.49, p [less than] .04; (3) TC had less disturbed scores on Poor Emotional Control (Devereux Scale 6), a measure of volatility, F = 5.63, p [less than] .03; (4) TC had poorer scores on the Positive Personal Self Scale (Tennessee, Column C), a measure of overall sense of personal worth F = 4.82, p [less than] .04; and (5) TC had a more deviant mean score on Social Anxiety (Jesness Scale 8), a measure of concern in the interpersonal sphere, F = 5.15, p [less than] .03. The group admitted to TC also had significantly higher Full-Scale IQ scores on the Wechsler, F = 8.97, p [less than] .01.

In examining the effects of time and group over the six months, it was found that 13 tests yielded significant results. The effect group was significant in that: (1) TC had a more disturbed score on Needs Approval-Dependency (Devereux Scale 7), a measure of the degree to which adult support is sought, F = 4.85,p [less than] .04; and (2) TC had a more deviant Social Anxiety score (Jesness Scale 8), a measure of concern in the interpersonal sphere, F = 8.75, p [less than] .01.

The main effect of time was significant in the direction of improvement for both groups for: (1) Positive Self Satisfaction (Tennessee), a measure of feelings about the perceived self, F = 6.30, p [less than] .04; (2) Row Variability (Tennessee), a measure of consistency of test responses, F = 4.31, p [less than] .05; (3) Personality Disorder (Tennessee Clinical Scale), a scale that differentiates personality disorders, F = 6.65, p [less than] .02; (4) Social Maladjustment (Jesness Scale 1), a measure of poor adjustment and low ego strength, F = 11.06, p [less than] .01; (5) Value Orientation (Jesness Scale 2), a measure of delinquent values, F =7.49, p [less than] .01; (6) Autism (Jesness Scale 4), a measure of egocentric preoccupations, F = 6.05, p [less than] .02; (7) Alienation (Jesness Scale 5), a measure of estrangement and disruptiveness, F = 8.57, p [less than] .01; (8) Manifest Aggression (Jesness Scale 6), a measure of difficulties with aggression, F = 8.11,p [less than] .01 and (9) Asocial Index (Jesness), a measure of the tendency for delinquent behavior, F = 5.59, p [less than] .03). Inspection of the test profiles over time indicates a reduction in the amount of clinical disturbance for both groups.

There were two variables in which the effect of the interaction of group by time was significant. These were: Positive Personal Self (Tennessee Column C), where TC improved over the 6 months and BU deteriorated, F = 4.67, p [less than] .04; and Defensive Positive (Tennessee Clinical Scale DP), a measure of defensiveness where TC's mean score became more disturbed and BU improved, F = 4.68, p [less than] .05.

Neither the Devereux Behavior Rating Scale nor the Locus of Control Scale showed any significant change over time.

When comparing dropouts to nondropouts, there were no significant differences in background characteristics; however, dropouts had significantly less family contact while in treatment ([[Chi].sup.2] = 10.97249, p [less than] .01). Analysis of baseline levels of the dependent variables revealed two significant differences in that nondropouts had poorer Emotional Control (Devereux Scale 6), and thus scored as significantly more volatile, F = 5.89, p [less than] .02; and dropouts had higher scores on the Neurosis Scale (Tennessee), F = 7.51, p [less than] .01.

DISCUSSION

On admission both the TC and BU clients had baseline test scores and background histories suggestive of serious problems of emotional maladjustment and atypical behavior. Despite the small number of statistically significant differences in demographic characteristics and test scores, differences do suggest distinct patterns of socioeconomic status, family background, and diagnostic category. What emerged overall was a profile of the TC client group as unhappy, anxious, immature, and neurotic. The profile of the BU group is that of impulsive, gratification-oriented youngsters who denied problems and were more hostile than happy. Despite the fact that the two programs have similar goals and propose to treat the same clients, results suggest that these groups represent two distinct clinical populations. It may be that those who refer cases may do so on the basis of their own biases about a medical vs. nonmedical facility (Westendorp, Brink, Oberson, & Ortiz, 1986).

Overall, the tests comparing dropouts to nondropouts revealed few significant differences. Baseline test scores suggest that dropouts tended to be more emotionally constricted and less volatile than other clients and somewhat more like diagnosed neurotics. It may be that these clients responded to their admission with an increase in anxiety that was beyond their level of tolerance, and thus tended to act out by running away or seeking discharge rather than express their distress. This may have been compounded by the significant absence of family contact as compared with other clients. The importance of family involvement is strongly represented in the literature of residential treatment (Kaplan & Sadock, 1991; Pfeiffer, & Strzelecki, 1990; Lewis, 1985; Riche, 1983), and the amount of contact is thought to have a bearing on treatment outcome.

The findings regarding change over time are characterized by a small number of significant changes relative to the number of measures used and by the lack of dramatic clinical shifts. There was a tendency for test scores to move in the direction of improved functioning in both groups, a trend which reached significance on nine of the scales. All clients became less maladjusted, less delinquent in their values and behavior, less egocentric, less alienated, less unable to control aggression, and less like clients diagnosed as personality disordered. In addition, overall scores on self-esteem and the consistency of self-evaluations improved for the whole sample. These shifts on self-esteem and personality functioning are judged as indicative of some important clinical improvement in the total group under study. However, there was no evidence of significant improvement on the behavioral measure nor was there change as measured by the Locus of Control instrument.

The absence of significant change on the measure of behavior deserves particularly serious consideration. One of the few universally agreed upon criteria of success for residential treatment is the alteration of dysfunctional behavior (Biase, 1984; Durkin & Durkin, 1975). Examination of the Devereux Behavior Rating Scales over time revealed that there was clinically interpretable deterioration that tended to equal or exceed positive change. On the one hand, the deterioration of behavior may be viewed as program failure. On the other hand, the relationship between treatment impact and behavior may be a complex one. It has been suggested elsewhere that admission to a residential unit provokes regression as it is one of the most intrusive of all interventions (Rutter, 1987) and provides an "intensely stimulating environment" which itself can play "a negative role and encourage going on overload" (p. 329, Small, Kennedy, & Bender, 1991). The possibility of negative change has even been cited as proof of a treatment's potency (Imber, Pilkonis, & Glanz, 1983) or indicative of a client's current stage of involvement in the treatment process (Lewis & Summerville, 1991; Mishne, 1986). Adolescence at its best is a stage fraught with the possibility of behavioral disturbance and instability (Mishne, 1986), and this may be exacerbated in a clinical population. Further, the phenomenon of "group contagion" (Brendtro & Ness, 1983; Small et al., 1991), operative in adolescence, is exaggerated in residential settings. Lastly, it may well be that the nature of the conduct-disordered adolescent's response to treatment is in fact discontinuous and nonlinear (much like normal adolescent development (Mishne, 1986), and steady consistent progress will not be visible in all areas. Curry (1991) noted that improvement in adolescent residential treatment often proceeds from symptoms to behavior to relationship skills, and then finally to academic and vocational functioning. While behavior showed no improvement in this study, self-esteem did. Perhaps the rate of change of one affects the other, or behavior by its very nature is more volatile and less likely to exhibit stable progression toward improvement.

The major finding of this study was the relative lack of significant differences over time in the rate of change in the two treatment groups. The only instrument which proved sensitive to statistically significant differential change over time was the Tennessee Self-Concept Scale. It appears that the TC group experienced some positive change in both their sense of personal worth and healthy defensiveness while the BU group deteriorated significantly. Given the earlier observation that the TC group tended to present as more unhappy and neurotic on admission, while the BU clients were characterized by hostility, denial, and a relative tendency to deny discomfort, these findings are in the expected direction. Clinically, they can both be interpreted as progress as they may represent both the lessening of self-deprecation in the hospitalized group and the breaking down of the bravado and pseudo-adequacy of the BU group.

Limitations of the Current Study

There are limitations of this study that are due to individual methodological problems, while others are inherent to outcome research in residential treatment.

Random assignment to the two treatments and inclusion of a non-treated control group was impossible, given that ethical consideration and results suggest that the two groups investigated represent two distinct populations. For this reason, comparisons must be made cautiously. It is always difficult to demonstrate in an adolescent population that maturation did not contribute in some way to any of the changes observed (Sarafino & Armstrong, 1986). In addition, the time intervals chosen did not permit assessment of postdischarge adaptation, an important indicator of treatment effectiveness. However, it was judged that the posttreatment functioning is influenced by complex individualized ecological factors (Evans, 1987; Lewis, 1984; Wells, 1991b) that were beyond the scope of the current study.

The presence of numerous scales measuring outcome can actually increase the appearance of spurious change. The small number and consistency of results suggests that this was not an important factor in this study. However, the problem of regression toward the mean (Mcleary, Gordon, Mcdowal, & Maltz, 1979) can never be completely dismissed as an alternative explanation for improvement in a deviant population.

The integrity of the treatment programs under study is always a concern in outcome research (Quay, 1979). The administration of the Moos (1974) scales represented an attempt in this study to confirm that the models were in fact implemented; however, the consistency with which the treatment was delivered is difficult to assess. Lastly, the measurement of outcome in a residential treatment environment is complex. It is impossible in quantitative outcome research to pinpoint which elements of a program contribute to change.

Recommendations for Future Research

The results of this study raise questions about the utility of comparative designs in promoting understanding of residential treatment models and about the course of responses to treatment in a conduct-disordered adolescent population. There is a need for research investigating underlying processes, stressing similarities rather than differences between different models. Qualitative methods that go beyond test scores to consider perceptions, attitudes, and personal experiences can begin to identify nonspecific factors and components on the microscopic level (Curry, 1991) that contribute to clinical improvement across models. Treatment outcome, especially in an adolescent population, must be viewed as a highly individual process, not a product. A series of such studies is needed so that a common knowledge base can begin to unify a field of literature and practice that is currently fragmented, poorly developed (Evans, 1987), and misunderstood despite growing numbers of adolescents in residential settings (Small et al., 1991; Zimmerman, 1990).

REFERENCES

Auchenbach, T.M. (1981). Behavioral disorders. In R. Henley (Ed.), Encyclopedia of clinical assessment, Volume 1. San Francisco: Jossey-Bass.

Ainsworth, F., & Fulcher, L. (1981). Introduction. In F. Ainsworth, & L. Fulcher (Eds.), Group care for children - Concept and issues. London: Tavistock.

Almond, R. (1974). The healing community. New York: Jason Aronson.

Bale, R. N. (1979). Outcome research in therapeutic communities for drug abusers: A critical review, 1963-1975. International Journal of the Addictions, 14(8), 1053-1074.

Beschner, E. W., & Friedman, A. S. (1985). Treatment of adolescent drug abusers. International Journal of the Addictions, 20(6 & 7), 91-93.

Biase, D. V. (1984). A drug abuse prevention program developed within a therapeutic community. Journal of Psychoactive Drugs, 16(1), 63-68.

Bratter, T. E., Collabolletta, E. A., Fossbender, A. J., Pennchia, M. C., & Rubel, J. R. (1985), The American self-help residential therapeutic community: A pragmatic treatment approach. In T. E. Bratter, & G. G. Forrest (Eds.), Alcoholism and substance abuse. New York: Free Press.

Brendtro, L. K., & Ness, A. E. (1983). Reeducating troubled youth: Environments for teaching and treatment. New York: Aldine.

Buros, O. K. (1972). The seventh mental measurements yearbook. Highland Park, NJ: Gryphon.

Curry, J. F. (1991). Outcome research on residential treatment: Implications and suggested directions. American Journal of Orthopsychiatry, 61(3) 348-357.

DeLeon, G., & Rosenthall, M. S. (1979). Therapeutic communities. In R. I. Dupont, A. Goldsteing, & J. O'Donnell (Eds.), Handbook on drug abuse. Washington: National Institute on Drug Abuse.

Durkin, R. P., & Durkin, A. B. (1975). Evaluating residential treatment programs for disturbed children. In M. Guttentag, & E. L. Swening (Eds.), Handbook of evaluation research. Beverly Hills, CA: Sage.

Evans, J. (1987). Residential care of youth-convergent versus divergent thinking. Journal of Adolescence, 10, 255-267.

Fitts, W. (1965). Manual for the Tennessee Self-concept Scale. Nashville, TN: Counselor Recordings and Tests.

Gillian-Mall, D., & Judd, P. (1986). The effectiveness of residential Care facilities for adolescent boys. Adolescence, 21(82), 307-321.

Hunt, D., Butler, L., Noy, J., & Rosser, M. (1979). Assessing conceptual level by the paragraph completion method. Toronto: OISE.

Imber, S. D., Pilkonis, P. A., & Glanz, L. (1983). Outcome studies in psychotherapy. In E. Walker (Ed.), The handbook of clinical psychology. Homewood, IL: Dow Jones-Irwin.

Jesness, C. F. (1972). Manual for the Jesness Inventory. Palo Alto: Consulting Psychologists Press.

Kaplan, H. I., & Sadock, B. I. (1991). Synopsis of psychiatry (6th ed.). Baltimore: Williams & Wilkins.

Lennard, H. L., & Allen, D. D. (1974). The treatment of drug addiction: Towards new models. In G. DeLeon (Ed). Phoenix House: Studies in a therapeutic community (1968-1973). New York: MSS Information.

Levenson, H. (1981). Differentiating among internality, powerful others and chance. In H. Lefcourt (Ed), Research with the locus of control construct. New York: Academic Press.

Lewis, M. (1985). Residential treatment. In M. I. Kaplan, & B. J. Sadock (Eds), Comprehensive textbook of psychiatry (4th ed.) Baltimore: Williams & Wilkins.

Lewis, M., & Summerville, J. (1991). Residential treatment. In M. Lewis (Ed.), Child and adolescent psychiatry. Baltimore: Williams & Wilkins.

Macaller, P. (1993). Reaffirming rehabilitation in juvenile justice. Youth and Society, 25(1), 104-125.

Manning, N. (1972). Evaluating the therapeutic community. In R. D. Hinshelwood, & N. Manning, Therapeutic communities: Reflections and progress. London: Routledge and Kegan Paul.

McLeary, R., Gordon, A., McDowal, D., & Maltz, M. R. (1979). How a regression artifact can make any delinquency intervention program look effective. In L. Sechrest, S. G. West, M. A. Phillips, R. Redner, & W. Yeaton. Evaluation studies review annual. Volume 4. Beverly Hills, CA: Sage.

Milham, & Bullock (1987). A holistic approach to evaluation of residential institutions. Child and Youth Services, 8(3/4), 5-8.

Mishne, J. M. (1986). Clinical work with adolescents. New York: Free Press.

Moos, R. (1977). The human context - Environmental determinants of behavior. New York: Wiley.

Moos, R. M. (1974). Correctional institutions environment scale. Palo Alto, CA: Consulting Psychologist Press.

Munson, R. F., & Blincoe, M. M. (1984). Evaluation of a residential treatment center for emotionally disturbed adolescents. Adolescence, 19(74), 253-261.

Munson, R. F., & LaPaille, K. (1984). Personality tests as predictors of success in a residential treatment center. Adolescence, 19(78), 697-701.

Paulhus, D., & Christie, R. (1981). Spheres of control: An interactionist approach to assessment of perceived control. In M. Lefcourt (Ed.), Research with the locus of control construct. Volume 1. New York: Academic Press.

Pfeiffer, S. I., & Strezeleck, B. A. (1990). Inpatient psychiatric treatment of children and adolescents: A review of outcome studies. Journal of the American Academy of Child and Adolescent Psychiatry, 29(6), 847-853.

Quay, H. (1979). The three faces of evaluation. In L. Sechrest, S. G. West, M. A. Phillips, R. Redner, & W. Yeaton (Eds.), Evaluation studies review annual, Volume 4. Beverly Hills, CA: Sage.

Reed, T. (1978). Outcome research on treatment and the drug abuser: An exploration. International Journal of the Addictions, 13(1), 149-171.

Reitsma-Street, M. (1982). A critical review of the conceptual level matching model and its relevance for the treatment of maladjusted youth. The Differential View, 12, 49-58.

Riche, M. (1983). Integrating families into a healing community: The use of structural and strategic family therapy in a psychodynamically oriented hospital. Residential Group Care and Treatment, 1(4), 67-84.

Rutter, M. (1987). Temperament, personality and personality disorder. British Journal of Psychiatry, 150, 443-458.

Sarafino, E., & Armstrong, J. W. (1986). Child and adolescent development (2nd ed.). St. Paul, MN: West Publishing.

Small, R., Kennedy, K., & Bender, B. (1991). Critical issues in residential treatment: The view from within. American Journal of Orthopsychiatry, 61(3), 327-338.

Spivack, G., Spotts, J., & Haines, P. E. (1967). Manual for the Devereux adolescent behavior rating scale. Devon, PA: The Devereux Foundation.

Turner, E. (1985). Centralized assessment at Youth Horizons. The Differential View, 15.

Velasquez, J. S., & Lyle, C. G. (1985). Day versus residential treatment for juvenile offenders: The impact of program evaluation. Child Welfare, 54(2).

Wells, K. (1991a). Long-term residential treatment for children: Introduction. American Journal of Orthopsychiatry, 61(3), 324-326.

Wells, K. (1991b). Placement of emotionally disturbed children in residential treatment: A review of placement criteria. American Journal of Orthopsychiatry, 61(3), 339-347.

Weschsler, D. (1974). Manual for the Weschsler intelligence scale for children - revised. New York: Psychological Corporation.

Westendorp, F., Brink, R. L., Roberson, M. L., & Ortiz, I. E. (1986). Variables which differentiate placement of adolescents into juvenile justice or mental health systems. Adolescence, 21(81), 24-37.

Wolff, S. (1984). The concept of personality disorder in childhood. Journal of Child Psychology and Psychiatry, 25(1), 5-13.

Zimmerman, D. P. (1990). Notes on the history of inpatient and residential treatment. Adolescence, 25(97), 9-38.