A developmental psychopathology view

In this section, a developmental psychopathology view of EE is presented. It examines how EE attitudes emerge over time in caregivers, how the patient is influenced by these attitudes at different stages of development, how his or her behavior feeds back on the attitudes and behaviors of family members, and how these dynamic processes unfold over time in the context of other risk and protective factors. The research on the origins of EE attitudes will then be discussed in light of this model.

Consider the development of EE in parental caregivers, in whom the construct has been most consistently applied. The seeds for critical or overinvolved attitudes in parents probably begin with a child who has temperamental disturbances (e.g., irritability, low frustration tolerance, mood lability, high anxiety; e.g., Akiskal & Akiskal, 1992) or compromised cognitive or behavioral functioning (e.g., Cornblatt, Obuchowski, Roberts, Pollack, & Erlenmeyer–Kimling, 1999). These early disturbances may, in part, reflect the child's genetic diathesis to psychiatric disorder. The child is paired with a parent who, by virtue of his or her own neurobiology and social history, is prone to react with frustration (resulting in frequent criticisms of the child) or guilt and anxiety (resulting in overly protective behaviors).

The expression of high-EE behaviors by parents is likely to be particularly difficult for the child when he or she begins to develop schemata for the self. Repeated exposure to criticisms during these sensitive periods may lay the foundation for future negative self-schemas (e.g., the expectation that he or she will fail at independent activities), self-doubt and self-criticism, and core beliefs about relationships as fraught with conflict. Likewise, repeated exposure to anxiety and/or overinvolvement from a parent may lead the child to internalize a sense of self as incompetent and the world as a dangerous place. These processes may result in difficulties with emotional self-regulation, as reflected in depression symptoms, anxiety, or aggression. In support of this view, Colder, Lochman, and Wells (1997) found that among healthy school-aged children, parental overinvolvement and harsh discipline were associated with high levels of depressive symptoms among children, but only when children were temperamentally characterized by high fear. Overinvolvement or excessive criticism may interfere with the child's acquisition of emotional self-regulatory skills, such as the ability to tolerate or learn to appropriately express negative affect.

In family system terms, the child's low self-esteem, behavioral inhibition, or feelings of incompetence preserve family equilibrium and distract parents from unacknowledged conflicts with each other (Minuchin, 1985). His or her negative counterreactions further fuel high-EE behaviors in parents, which in turn, contribute to the child's emotional and cognitive vulnerabilities. Thus, the biologically and genetically vulnerable child becomes an “identified patient” due to his or her temperament, and the family system contributes to the perpetuation of his or her difficulties in adjustment.

The challenge for the family is to alter its patterns of coping to address these disturbances. If the family remains fixed or rigid in its patterns of relating, then as the child ages and attempts to develop relationships outside of the family, aversive interactions with family members are likely to intensify. In recursive fashion, these dynamic processes contribute further to the child's negative views of self, which in turn, elicit greater negativity from family members.

The concept of multifinality

Although early problems in adaptation statistically portend later functional impairments (“probabilistic epigenesis”; Cicchetti & Rogosch, 2002), the model outlined above does not assume that all children with high-EE parents will have adaptational failures. Rather, the diversity of outcomes among children whose parents hold high-EE attitudes—psychiatric illness, subsyndromal symptoms, or psychological health—exemplifies the developmental psychopathology concept of multifinality: the same initial conditions may lead to a diversity of outcomes (Cicchetti & Rogosch, 2002). Numerous risk or resiliency processes will influence whether high-EE attitudes take their toll on the later functioning of the child. These factors may include whether the child is genetically predisposed to develop a psychiatric disorder; whether the child maintains self-organizational strivings such as self-reliance, self-confidence, or internal locus of control or initiates self-righting tendencies to cope with adversity (e.g., Cicchetti & Rogosch, 1997); whether there are other relatives in the family who are low in EE and/or high in warmth; the openness of the family in addressing its problems; the impact of intervening psychological or school interventions; and other “transient challengers” (e.g., sudden losses) or “transient buffers” (e.g., improved socioeconomic circumstances; Cicchetti & Lynch, 1993). Resiliency may also take the form of positive self-talk when facing criticism or overinvolvement (e.g., “Dad gets mad when I do things wrong because he cares and wants me to do my best”). Vulnerable children for whom protective influences are unavailable, or who are less resilient when facing adversity will be more strongly affected by critical or overprotective attitudes of parents. For them, criticism or overprotectiveness will be more “penetrant” and more closely associated with cognitive and emotional vulnerabilities as they age.

In parallel, among children who do develop psychiatric disorders, there will be a dispersion of outcomes in terms of family reactions. Consider the case of a teenager who has always been moody and then develops a manic episode. In systems terms, the episode can create discontinuities in the functioning of the system that lead to heightened conflict, dysfunctional patterns of interaction, uncontrolled expressions of negative affect, greater enmeshment and loosening of boundaries, unhealthy alliance patterns or, at the other end of the spectrum, affective disengagement. The degree to which these discontinuities disrupt the system will depend on its capacity to adaptively self-organize and respond to changes (Cox & Paley, 1997; Minuchin, 1985). The mother may become overly invested in promoting the teenager's health and begin to “overmonitor” him or her; the father may pull away from the mother and become critical of the adolescent, believing that he or she is producing the symptoms to get attention; other siblings may become resentful of the powerful role their ill sibling has obtained in the family; and boundaries between the father and his other children may become blurred. In support of these predictions, Gilbert, Christensen, and Margolin (1984) found that distressed and multiproblem families were characterized by structural problems including poor marital alliances and crossgenerational coalitions between parents and at-risk children.

The resilient adaptation of the affected child will also affect the trajectory of these pathways. For example, the onset of a manic episode will have a different effect on caretakers if the adolescent recovers through medication, if he or she is drug compliant, if he or she is able to help promote positive interactions with caregivers, or if he or she is making a concerted effort to function adaptively. The development of EE attitudes will also be affected by the resiliency of the relative: whether he or she has overcome psychiatric problems him- or herself or has successfully dealt with health problems in other family members, the degree to which he or she obtains and uses information about the causes and treatment of the offspring's dysfunctional behavior, whether he or she is able to take a proactive stance against stigmatization from others, and whether he or she is able to react to the offspring's provocations with measured responses. Thus, the same event (the onset of a major psychiatric illness in one family member) leads to a diversity of reactions among caretakers and other family members depending on the combination of risk and protective factors in the family and social environment.

Finally, the effects of EE attitudes on the subsequent course of schizophrenia or bipolar disorder illustrate the concept of multifinality of outcomes. The reactions of family members, mediated by individual differences in genetic, biological, and psychological vulnerabilities and the success or failure of treatments, predict a dispersion of clinical end points: long-term stability, relapse, rehospitalization, or persisting symptoms; various levels of social or occupational dysfunction; and various levels of family distress or dysfunction (e.g., divorce, conflict, disengagement). In turn, the nature and persistence of the patient's dysfunctional behavior fuel further positive or negative reactions from caregivers, which then have a recursive impact on the subsequent longer term outcomes of the disorder.

This developmental view is supported by several research observations. First, not all patients from high-EE families relapse, nor do all patients from low-EE families remain well (Butzlaff & Hooley, 1998). High-EE attitudes are more common among parents whose schizophrenic offspring have been ill longer (Hooley & Richters, 1995). Premorbid variables, such as social withdrawal during adolescence, increase the probability that parents become overprotective (Miklowitz, Goldstein, & Falloon, 1983). Further examples are given below.

Implications of the model for family intervention

The developmental model suggests that successful interventions for severe psychiatric disorders must address disturbances in multiple levels of the family system. They should be timed at points of transition when the system is fluid and reorganizing, and the behavior of family members is most chaotic (Cicchetti & Toth, 1992; Minuchin, 1985). They must begin with a comprehensive assessment of targets for intervention, which will vary depending on dynamic attributes of the patient, the relatives, and their relationship. In most cases, successful interventions will encourage the psychiatric patient's use of appropriate medical treatments and services that provide extrafamilial support (e.g., groups). They should also focus on enhancing self-system processes such as autonomy, mastery, and self-determination, which promote the affected individual's striving for independence (Cicchetti & Rogosch, 1997). Finally, interventions should assist the caregiver in developing communication and problem-solving strategies to deal with fluctuations of the patient's disorder, help him or her realize that many aspects of the disorder are uncontrollable, and address his or her needs for extrafamilial support.

EE as a reaction to symptomatology among patients

In this view, an identifiable cluster of patient attributes predicts levels of criticism, hostility, or emotional overinvolvement in caregivers. The data on this issue are inconclusive. Researchers generally do not find cross-sectional correlations between levels of EE and the severity of patients' concurrent symptoms during or following acute illness episodes. Patients in high-EE families do not appear to be sicker to begin with, whether they suffer from schizophrenia (e.g., Brown, Birley, & Wing, 1972; Miklowitz et al., 1983; Nuechterlein et al., 1992; Vaughn & Leff, 1976) or bipolar disorder (Miklowitz et al., 1988).

A different picture emerges when considering EE from a developmental perspective. The frequency of critical comments among parents increases over time as a function of the longevity of the schizophrenic patient's disorder. In a cross-sectional study, Hooley and Richters (1995) found that parents who had been coping with schizophrenia in an offspring for 3–5 years expressed an average of 15 critical comments during the Camberwell Interview, whereas parents who had coped with the illness for less than 1 year expressed an average of 4.2 critical comments. After repeated exposures to episodes of illness, medical nonadherence, and low functioning, relatives may begin to blame the patient for his or her deviant behavior, or believe that the patient is unable to live independently and requires careful monitoring and protection.

Interestingly, levels of EE change as patients improve. Between 25 and 50% of parents of schizophrenic patients who are rated high EE, critical when their son/daughter is acutely ill are reclassified as low EE 1 year later, when patients are in remission (Brown et al., 1972; Hogarty, Anderson, Reiss, Kornblith, Greenwald, Javna, & Madonia, 1986; Tarrier, Barrowclough, Vaughn, Bamrah, Porceddu, Watts, & Freeman, 1988). Nonetheless, EE has both state and trait attributes (Hooley & Gotlib, 2000). Although levels of EE criticism drop over time, the most critical relatives are still the most critical even after the patient has remitted (Hooley, Rosen, & Richters, 1995; Simoneau, Miklowitz, Richards, Saleem, & George, 1999). Emotional overinvolvement appears to be more stable than highly critical or hostile attitudes, declining more slowly over time (Brown et al., 1972; Leff, Berkowitz, Shavit, Strachan, Glass, & Vaughn, 1990).

These findings can be interpreted in light of the foregoing developmental model. Minuchin (1985) points out that the behavior or affect of family members may be the most disturbed during periods of high stress and then becomes more stable as the stressor is removed. Some relatives appear to be temperamentally predisposed to high-EE attitudes, but these attitudes do not reach threshold until elicited by the stress associated with the patients' symptoms. It appears that the unidirectional reactivity of relatives to patients' symptoms is not an adequate explanation for the variability in relatives' EE attitudes.

EE as a product of psychopathology within the relative

Can the different reactions of relatives be explained by the relatives' own history of psychopathology? A key component of the developmental model is that high-EE parents are prone to react negatively because of their own temperamental disturbances. Thus, a relative who is under the strain of taking care of an ill son or daughter would be more likely to express a negative emotion if he or she was currently or previously depressed, anxious, or otherwise affectively dysregulated. In family system terms, the relative's negativity toward the patient involves a projection of the relative's own feelings of vulnerability deriving from a history of personal distress.

Only two studies have examined the linkage between EE and diagnosable psychopathology in relatives. Goldstein, Talovic, Nuechterlein, Fogelson, Subotnik, and Asarnow (1992) examined EE among parents of schizophrenic patients during the patient's acute illness and again 1 month after a hospital discharge. The highest rates of individual psychopathology were found among parents who were high EE across time and across measurement methods (Camberwell interview vs. a 5-min speech sample reassessment). A study by Goldstein, Miklowitz, and Richards (2002) examined EE and individual psychopathology among 82 relatives of 66 patients with bipolar disorder. No relations emerged between relatives' EE on the Camberwell interview and past or present DSM-IV psychopathology. Paradoxically, relatives with high levels of emotional overinvolvement were less likely to have a history of psychiatric disorder than those with low levels of involvement. Possibly, relatives who have a history of psychiatric disorder and recover fully may believe that the patient should be able to do the same. This reasoning may lead the relative to gradually disengage from the patient and his or her struggles with the disorder.

EE and relatives' attributions about patients' behaviors

Family systems theorists have long argued that the perceptions held by family members of each other are important in explaining their behavior in the family setting, as well as the perceptions, feelings, and behaviors of other family members (e.g., Guttman, 1991; Selvini, Boscolo, Cecchin, & Prata, 1980). Hooley and colleagues (1995; Hooley & Gotlib, 2000) postulate that relatives become critical or hostile when they make internal, personal, or controllable attributions about the causes of negative patient-related behaviors or events. Behaviors that disrupt family functioning (e.g., intense irritability, uncooperativeness), involve socially undesirable behaviors (e.g., withdrawal or embarrassing symptoms), or involve rejection of help (e.g., refusing medications, being unwilling to take advice) are the most likely targets for EE criticism. In contrast, low-EE relatives are more likely to attribute negative behaviors or events to external, universal, or uncontrollable factors, including the effects of an illness (Hooley & Gotlib, 2000). The developmental psychopathology view would predict that parents with temperamental disturbances would be more likely to make controllability attributions, especially if the ill relative has been difficult to influence and consistently shows behaviors that reflect poor judgment, misguided attempts at independence, or lack of insight.

The EE/attribution model has received support in a number of studies (for review, see Hooley & Gotlib, 2000). For example, Weisman, Nuechterlein, Goldstein, and Snyder (1998) examined Camberwell interviews from relatives of young adult schizophrenic patients and found that high-EE relatives were more likely than low-EE relatives to criticize negative “deficit” symptoms (e.g., apathy, withdrawal) or long-standing character attributes (e.g., introversion) than positive symptoms of schizophrenia (e.g., hallucinations). In a bipolar sample, Wendel, Miklowitz, Richards, and George (2000) found that when interacting with their ill spouse or offspring, high-EE relatives were more likely than low-EE relatives to make statements attributing negative patient behaviors (e.g., withdrawal, irritability, hypomanic behavior) to personal and controllable factors (e.g., “You don't get your work done because you just don't care about it”).

Why would some relatives make controllability attributions whereas others blame the illness? Hooley and Hiller (2000) found that high-EE relatives of schizophrenic patients score higher than low-EE relatives on measures of personality inflexibility. High-EE relatives are also more likely to have an internal locus of control for their own successes and failures (Hooley, 1998). Possibly, high-EE relatives are more likely than low-EE relatives to be intolerant of a family member who violates societal norms and does not appear to be working hard to improve his or her situation. There may also be cultural differences in attributional patterns. For example, Weisman, Gomes, and Lopez (2003) found that Latino relatives of schizophrenia patients were rarely rated high in EE. These relatives consistently attributed the patient's illness to external factors such as a neurological illness, life events, or spiritual causes.

Putting the attribution model together with the data on psychopathology among relatives yields the following predictions. A history of psychiatric problems or temperamental disturbances in a caregiver would be associated with causal attributions of the form “I handled my own problems. Why can't my son/daughter?” In this sense, high criticism or hostility reflects the endpoint of a developmental pathway from a personal history of illness, to recovery, to a specific set of beliefs about personal effort as fundamental to recovery, to negative reactions to another family member who does not appear to be effortful in dealing with problems. If the patient, in turn, overvalues an illness explanation (e.g., “I'd be able to finish college if I didn't have bipolar disorder”) or external factors (“My medicines make me not want to work”) then an attributional mismatch has occurred between the high-EE caregiver and the patient. These differences in belief systems may be reflected in negative back and forth, escalating patterns of verbal interaction, as discussed below.

A clinical implication of the EE/attribution research is that psychosocial treatments need to consider the role of belief systems among caregivers and patients regarding the causes of symptomatic behavior. Some of these beliefs may be culturally based. Contemporary models of therapy for marital dysfunction focus on modifying partners' evaluations of their own and their partners' behaviors (e.g., Baucom, Epstein, & LaTaillade, 2002). Family educational programs for psychiatric disorder may also have an impact on attributions. Among relatives of schizophrenic patients taking part in a psychoeducational program, reductions in EE were correlated with increases in universal and uncontrollable illness attributions (e.g., “Many people with schizophrenia have trouble holding a job”; Brewin, 1994).

Nonetheless, we should be mindful that encouraging patients or relatives to make attributions to a purely biological cause will not necessarily improve the patient's self-view or reduce the relatives' fears or resentments. Such attributions can make the patient feel helpless and hopeless to exert control or make life changes, except by taking medications. They may also lead to the same social exclusionary practices that attributions to demonic possession produced in an earlier era. As Hinshaw and Cicchetti (2000, p. 567) caution,

… both demonologic and medical-model perspectives convey the attributional consequence of lack of personal control or responsibility for disordered behavior. Whereas such attribution should theoretically lead to more benign reactions, fear of the disturbed behavior, blame for the person's susceptibility to influence by evil or illness, and the typical perception that behavior is indeed controllable have all led instead to extremely harsh reactions to both conceptions of the disorder.

EE and family interactions

Fundamental to most models of family systems, particularly the strategic models (Madanes, 1991), is the view that a family is best understood as a moving picture, or a series of predictable reactions and counterreactions by family members that are reciprocally dependent on one another. As Cox and Paley (1997) state, “the behavior of any one individual in the family is best understood according to their functioning within the network of interpersonal relationships in which the individual is embedded, rather than according to individual behavior considered alone” (pp. 254–255). In this view, high-EE attitudes, even if measured in one individual, are leading indicators of patterns of interaction in the family as a whole.

Studies of family intervention in schizophrenia

The investigations of family intervention for schizophrenic disorder can be roughly divided into “first-generation” and “second-generation” studies (Goldstein & Miklowitz, 1995). The first generation studies evaluated whether family therapy adds to the efficacy of neuroleptic regimens in delaying psychotic relapses during maintenance treatment. Typically, the comparison groups have been patients receiving “customary care” (pharmacotherapy with case management) or pharmacotherapy with individual therapy. The second generation studies evaluated different forms of family intervention against each other.

First-generation studies

The first-generation studies found that adding a family intervention to medication cuts relapse rates to one-third of their expected rates. A review of eight first-generation studies (Miklowitz & Tompson, 2003) revealed that 6-month to 1-year relapse rates among schizophrenia patients receiving family psychoeducational treatment ranged from 0 to 33% (mean = 12.5%), and among schizophrenia patients treated with customary care, 17 to 61% (average = 42%). Crosscultural replications of this result have been reported, notably two studies in China and one in Spain (for review, see Baucom et al., 1998). Nonreplications have also been reported (e.g., Kottgen, Sonnichsen, Mollenhauer, & Jurth, 1984; Linszen, Dingemanns, Van der Does, Nugter, Scholte, Lenior, & Goldstein, 1996; Vaughan, Doyle, McConaghy, Blaszczynski, Fox, & Tarrier, 1992) including one Mexican American sample (Telles, Karno, Mintz, Paz, Arias, Tucker, & Lopez, 1995).

Representative of the first-generation studies were randomized trials by Falloon, Boyd, McGill, Williamson, Razani, Moss, Gilderman, and Simpson (1985) and Hogarty and associates (1986; Hogarty, Anderson, Reiss, Kornblith, Greenwald, Ulrich, & Carter, 1991). Falloon and colleagues examined 44 schizophrenic patients who were stabilized on neuroleptic regimens after being randomly assigned to a 21-session behavioral family therapy or a comparably intensive individual supportive therapy. The behavioral family therapy began with a functional assessment of the family's communication styles and conflict areas, followed by intensive education for the patient and relatives about schizophrenic disorder and training in communication and problem-solving skills. The individual therapy was supportive and involved education and social skills training. Results over 2 years indicated that patients in family therapy had fewer days of illness, a greater probability of remaining episode free for the full 2 years (83 vs. 17%), and better social and familial relations (Falloon et al., 1985).

A post hoc analysis of the Falloon study (Doane, Goldstein, Miklowitz, & Falloon, 1986) revealed that parents in family therapy showed greater decreases in negative affective style (measured from verbal behavior in family interactions) and greater increases in problem-solving behavior during treatment than parents of patients assigned to individual therapy. Relapses in the individual therapy condition were more common in the first 9 months of follow-up if the parents showed an increase in negative affective style behavior during treatment. Thus, consistent with systemic views of therapeutic change, the efficacy of behavioral family therapy was in part mediated by whether families were able to modify their styles of interaction.

Hogarty et al. (1991) conducted a 2-year study of 103 schizophrenic patients who, after hospital discharge, received neuroleptic maintenance alone or neuroleptic maintenance with psychoeducational family treatment, individual social skills training, or family treatment plus social skills training. All of the families were rated high EE at baseline. Family treatment began with a one-session educational workshop. Then, families attended biweekly family sessions over the next 6–12 months, which focused on the patient's reintegration into social and familial roles. The therapists employed techniques of traditional family systems therapy, including reframing and relabeling of family problems and encouraging boundaries between parental and offspring subsystems. At 2 years, rates of relapse were lowest in the family treatment plus medication condition (29%) and the combined family, individual, and medication condition (25%), suggesting no advantage of adding individual therapy to family therapy. Rates of relapse in the individual social skills training plus medication condition were only slightly lower at 2 years (50%) than rates in the medication-only condition (62%).

Across the groups, no patient relapsed in the first year whose parents changed from high to low EE, whereas 36% of the patients whose parents remained high EE relapsed. Only the combination of family therapy and individual social skills training delayed relapses in the subgroup whose parents remained high EE (Hogarty et al., 1986). The combined treatment appeared to have blunted the effects of high EE on schizophrenia patients, perhaps through teaching patients and family members conflict resolution skills and ways to set limits on unproductive interactions.

Second-generation studies

Most of the second-generation studies compared single-family treatment formats with multiple family group formats. The rationale is that equivalent outcome results would support the use of family groups given the presumably lower costs of these modalities. More central to our purposes is the question of whether dynamic changes in family systems can be accomplished with several families at once.

Among schizophrenic patients who were in high face to face contact with their high-EE relatives, Leff et al. (1990) compared multifamily psychoeducation groups (n = 11) with home-based individualized family therapy focused on psychoeducation and support (n = 12). Families in both conditions began with two education sessions while the patient was hospitalized. Then, family or group sessions were held biweekly for 9 months, with maintenance sessions thereafter. All patients received antipsychotic medications. The psychoeducation groups excluded the patients, whereas the home-based family therapy sessions included patients and followed a model similar to Hogarty et al.'s (1991).

Over 2 years, the effects of the two modalities on relapse rates were comparable: 36% of the patients whose relatives were assigned to the psychoeducation groups relapsed, versus 33% of patients in family therapy. However, the dropout rate was higher in the relatives' groups, which were only attended regularly by 6 of the 11 assigned families. Families in the single-family format were more adherent with biweekly sessions, possibly because treatment included the patients and was held in the families' homes.

McFarlane, Lukens, Link, Dushay, Deakins, Newmark, Dunno, Horen, and Toran (1995) assigned families of 172 schizophrenia patients from six public hospitals to a single-family treatment or a multifamily group format that focused on psychoeducation, problem solving, and support. Treatments began after the patient had been discharged from the hospital and continued biweekly for 2 years. Families in both conditions began with three weekly sessions of psychoeducation. Both treatments included relatives and patients, focused on problem solving, and encouraged expansion of the family's natural social network. Over 2 years, patients assigned to the multifamily groups had somewhat lower cumulative rates of relapse (16%) than patients in single-family treatment (27%). No group differences in patients' attrition rates or medication adherence were found. The multifamily groups were found to be more cost effective than the single-family treatments in terms of rehospitalization expenses.

In a multisite study of 313 patients, Schooler et al. (1997) examined the combination of 12 months of single-family therapy (using the Falloon et al. home-based behavioral format) with 2 years of monthly clinic-based multifamily groups for relatives only. The comparison group received 2 years of the monthly relatives' groups. Patients were also randomly assigned to one of three neuroleptic maintenance strategies. The design enabled the investigators to determine whether the traditional advantages of single-family therapy (involving the patient, individualized goal setting, modifying idiosyncratic patterns of interaction) were of incremental value when the relatives were already receiving support in a group setting.

No differences emerged between the two family groups in patients' relapse or rehospitalization rates. The study raises the possibility that the individualized learning of skills or restructuring of transactional processes are not essential therapeutic ingredients of family approaches to schizophrenia. Providing education and support in a group format and reducing the family's social isolation may be sufficient to observe improvements in schizophrenia patients. However, this study did not include a medication-only condition. Thus, it was impossible to determine whether the addition of family therapy to medications was effective beyond what could have been accomplished with medications alone.

Implications and future directions

The literature on schizophrenia is consistent in demonstrating that family therapy, in various forms, is an efficacious adjunct to neuroleptic treatment in staving off psychotic relapses and improving family functioning. Clearly, family intervention should be considered an important component of the outpatient management of schizophrenia patients.

Several questions remain unanswered. First, what theoretical orientations of treatment lead to the most effective results for schizophrenia patients? Some of the studies examined family interventions that were supportive and educational (e.g., Hogarty et al., 1991; Leff et al., 1990; Leff, Kuipers, Berkowitz, & Sturgeon, 1985), whereas others examined more purely behavioral interventions (e.g., Falloon et al., 1985; Schooler et al., 1997). One study examined an “enactive” family therapy (skills training through role playing and between-session rehearsal) versus a “symbolic” family therapy (skills taught through discussion and instruction), and found no differences in efficacy (Tarrier, Barrowclough, Porceddu, & Fitzpatrick, 1994). Currently, there is no convincing evidence that single-family approaches of different theoretical orientations produce different effects, as long as treatments are matched in duration and intensity (Baucom et al., 1998). The exception seems to be psychodynamically oriented family interventions, which in two studies have been shown to be ineffective for the families of schizophrenic patients (Kottgen et al., 1984; McFarlane, Link, Dushay, Marchal, & Crilly, 1995).

Second, by what mechanisms do family interventions for schizophrenia achieve their effects? It does not appear that the clinical benefits of family psychoeducation are a direct function of increased medication compliance among patients (Falloon et al., 1985; Hogarty et al., 1991; Leff et al., 1985, 1990; McFarlane, Lukens, et al., 1995). Although most studies find that family therapy leads to reductions in EE or negative family interactions (e.g., Falloon et al., 1985; Hogarty et al., 1991; Leff et al., 1985, 1990; Tarrier, Barrowclough, Vaughn, et al., 1988), few studies show that such reductions are a necessary condition for the clinical improvement of patients. In fact, high- and low-EE families appear to benefit equally from family therapy (e.g., Linszen et al., 1996).

In all likelihood, the mechanisms of change associated with family interventions are complex. The core element of all family interventions is education. Providing didactic information about the disorder, examining members' affective reactions to this information (e.g., hopelessness, anger), addressing their causal attributions, and teaching new communication and problem-solving skills may lead to decreases in reciprocal patterns of criticism, guilt induction, or blame. Likewise, addressing caregivers' feelings of shame, burden, and social isolation through group support may translate into more effective problem solving and less family tension. Acquainting families with practical strategies such as how to intervene at the first signs of relapse may help to prevent rehospitalizations. In other words, examining changes in EE or interactional behavior gleaned from laboratory assessments may not capture the multiple ways in which families change as a result of psychoeducational interventions.

Third, what is the most effective format for delivering family treatment? Although multifamily groups appear to be more cost effective than single formats, at least one study (Leff et al., 1990) found that families were less adherent with multifamily groups. Thus, the question of format is not just an economic one. Turning to the theoretical models outlined early in this article, single-family and multiple-family group formats carry different assumptions about the role of the family in schizophrenia or other psychiatric disorders. Systems theory holds that family interventions must address relationship systems in which individual pathology is expressed (Cox & Paley, 1997). Single-family treatments usually have individualized objectives such as reopening lines of communication, changing idiosyncratic patterns of interaction, breaking taboos against discussing mental illness, addressing problematic coparenting strategies, restructuring patterns of dysfunctional alliance, or supporting subsystem boundaries.

In contrast, multiple-family groups begin with the assumption that family members are all coping with a brain disorder in their offspring or spouse. In the worst-case scenario, these groups “overbiologize” the causes of psychiatric disorders and neglect the individual family system in favor of common sense approaches to illness management, which may not even require the patient's participation. Thus, the family's dynamic mechanisms of growth, change, or self-regulation may receive minimal attention in multifamily groups.

In some cases, only single representatives of a family (i.e., a caregiving parent) attend multifamily groups, and the groups become support sessions for that individual. Exclusion of patients (or the implication that they need not participate) carries the risk that patients will not learn to take responsibility for their own care. In fact, an Australian study found that schizophrenic patients whose caregivers attended relatives-only groups did not differ in outcome from patients receiving medication and supportive counseling (Vaughan et al., 1992).

Multifamily formats may be more effective than single-family formats in addressing the problems of illness stigma, shame, and social isolation. Hinshaw and Cicchetti (2000) have reviewed the long history of discrimination, punitive interventions, and blame that persons with mental illness have endured along with their family members. Multifamily groups have the potential to reduce illness stigma, and the guilt and fear reactions of family members that often accompany societal rejection, by acquainting parents with other caregivers who have experienced similar dilemmas in coping with their son or daughter's disorder. They also have the potential to broaden the social network of relatives or patients. In the McFarlane et al. study, the family groups became “quasi-natural social networks” (McFarlane, Link, et al., 1995, p. 131). Participants modeled coping strategies for one another, offered solutions to each others' problems, and helped reduce each others' isolation. Thus, the support that parents receive in groups may translate into better coping, more mutual understandings of the illness and its triggers, and improvements in the parent/offspring relationship.

Possibly, single- and multiple-family groups are equally effective in relapse prevention but operate through different mediating mechanisms. Single-family formats may operate through enhancing communication and affective expression, decreasing conflict, strengthening healthy alliances and boundaries, improving the family's hierarchical organization, or increasing the patients' commitment to a pharmacotherapy regimen. Multifamily formats may operate through reducing the impact of stigma on the family, increasing social support, or enhancing the caregivers' ability to solve problems relevant to the mental health system. These hypothesized differences in change processes require examination in randomized, controlled designs comparing single and multiple family formats.

The efficacy of various treatment formats may also be moderated by differences in the developmental phases of patients' illnesses. For very chronic, treatment-resistant patients, there may be greater value in support groups that assist relatives in managing the disorder. In support of this possibility, McFarlane, Lukens et al. (1995) found that multifamily groups were particularly effective among patients who did not achieve full remissions of their illness following a hospital discharge. In contrast, relatives of younger, first- or second-episode patients may be unsure what level of functioning to expect from their relative in the future. They may have difficulty hearing the stories of other relatives who have endured their family member's disorder for a long time. They may also prefer the confidentiality of single-family formats. In parallel, patients with new onsets of schizophrenia may be more capable of incorporating the self-management techniques central to individualized family therapy than chronic or highly recurrent patients with significant cognitive impairment. Thus, treatment planning should take into account the duration of the disorder, the willingness of family members to share their stories and hear the stories of others, and the likelihood that the patient's participation will benefit the process (Goldstein & Miklowitz, 1995; Tompson, Goldstein, & Rea, 1996).

Family-focused treatment

The literature on family treatment of bipolar illness has lagged far behind the literature on schizophrenia, but inroads have been made (Clarkin, Carpenter, Hull, Wilner, & Glick, 1998; Miklowitz, George, Richards, Simoneau, & Suddath, 2003; Rea, Tompson, Miklowitz, Goldstein, Hwang, & Mintz, 2003). The existing treatment models have assumed no unique pathology of the family system, but rather that bipolar illness brings about disruptions to the regulatory processes of a healthy family. As for schizophrenia, the primary goal of family intervention is to encourage adaptation of the family or marital unit to the developmental transitions brought about by the illness.

Two randomized trials have tested the efficacy of a family-focused psychoeducational treatment (FFT) for adult bipolar patients (Miklowitz & Goldstein, 1997; Miklowitz et al., 2003; Rea et al., 2003). FFT is adapted from Falloon et al.'s (1985) model, and consists of 21 outpatient sessions of family assessment, education, communication enhancement training, and problem-solving skills training. The first study, conducted at the University of Colorado, randomized 101 acutely ill patients and their relatives (parents or spouses) to FFT (conducted in family's homes) and medications or a less intensive psychoeducational comparison called crisis management (CM; Miklowitz et al., 2003). Patients in CM received two home-based sessions of family education followed by crisis intervention sessions as needed for 9 months. Pharmacotherapy was provided by community physicians. Over the 2-year follow-up, patients in FFT had fewer relapses and significantly longer remissions than patients in CM (means = 73.5 and 53.2 weeks, respectively). Patients in FFT also had lower levels of depression and mania symptoms at follow-up.

Secondary analyses of the Colorado trial revealed that patients in FFT had higher levels of medication adherence over the 2-year follow-up than patients in CM. Notably, drug adherence mediated the effects of family intervention on mania symptoms, but not depression symptoms. Most relevant to family systems conceptualizations, patients and relatives in FFT showed greater improvements in communication and problem-solving skills than patients and relatives in CM. Furthermore, improvements among patients in positive interactional behavior with relatives predicted improvements in their mood disorder symptoms over 1 year, most notably their depression symptoms (Miklowitz et al., 2003; Simoneau et al., 1999). Levels of negative interactional behavior between relatives and patients remained stable over time despite the patients' symptomatic improvement. In fact, the best predictor of the number of negative statements made by patients or relatives after treatment was the number they made before treatment. As discussed earlier, temperamental variables, or perhaps the mismatching of temperaments between caregivers and patients may help explain why some families or couples become locked into negative relational patterns despite the patients' improvement.

A limitation of the Colorado trial was that the experimental and control groups were not matched on the amount of therapist/patient contact. This potential confound was controlled in a second trial of FFT at the University of California, Los Angeles (Rea et al., 2003). Fifty-three hospitalized manic patients were randomized to FFT (21 sessions over 9 months) plus medications or an individual therapy focused on education, stress management, problem solving, and compliance enhancement (also 21 sessions). By 1 year, patients in both treatment groups had equivalent rates of relapse. However, during a 1-year posttreatment follow-up, patients in FFT were less likely to relapse (28%) or be rehospitalized (12%) than patients in individual therapy (60 and 60%, respectively).

The delayed effects of the family intervention relative to individual therapy suggests that families and patients require a period of absorption of communication or problem solving skills before the benefits to patients can be observed. In systemic terms, the family's emotional climate, organization, and conflict resolution skills have to become more complex, adaptable, and differentiated in response to challenges (Minuchin, 1985). Adoption of new patterns of family functioning requires time.

Application of FFT to adolescent bipolar patients

Most of the studies of family psychoeducation have involved adult psychiatric patients, with the implicit assumption that they can be adapted to younger populations. This assumption has rarely been tested. Consider the case of childhood and adolescent bipolar disorder, which has a distinctive set of features when compared to adult bipolar disorder, such as chronicity of mood states, rapid cycling, psychosis, severe grandiosity, sudden deteriorations in functioning, and a predominance of mixed symptoms (Geller & Luby, 1997). The disorder may affect as many as 1% of all adolescents (Lewinsohn, Klein, & Seeley, 1995). The disorder's onset and treatment requirements interfere with important developmental tasks of adolescence, such as identity consolidation, forming intimate relationships with peers, academic achievement, and psychological autonomy (Cicchetti & Rogosch, 2002; Masten & Coatsworth, 1998).

Miklowitz, George, Axelson, Kim, Birmaher, Schneck, Beresford, Craighead, and Brent (in press) have developed a form of FFT that is sensitive to the developmental requirements of bipolar adolescents. Although the basic structure—family education, communication enhancement, and problem solving—has been retained, a number of adjustments have been made. For example, the reformulated FFT takes into account the rapid and intense onset and offset of episodes characteristic of juvenile patients. Family members become attuned to what sometimes appear to be trivial triggers for the child's mood swings (e.g., not being able to eat a favorite meal), which may lead to explosive outbursts of rage, uncontrolled crying spells, extreme silliness and giddiness, or suicidal gestures. Adolescents are encouraged to take responsibility for their own treatment to the extent that their cognitive and emotional development and symptoms allow. Parents are coached to set limits on oppositional behavior, use appropriate rewards and set consequences, and avoid allowing the child's mood and behavior to dictate the moods of other family members.

The efficacy of FFT for bipolar adolescent patients has been evaluated in a 20-case open trial at the University of Colorado. All patients entered the open trial following an acute episode of bipolar illness, and received a 9-month program of FFT and mood stabilizing medications. Results over 12 months suggested positive effects of FFT and mood stabilizers in reducing internalizing and externalizing symptoms (Miklowitz et al., in press). A two-site randomized controlled trial is now in progress (NIMH Grant MH62555).