bidirectional effects of expressed emotion and … · bidirectional effects of expressed emotion...

q American Association on Mental Retardation 229

VOLUME 111, NUMBER 4: 229–249 z JULY 2006 AMERICAN JOURNAL ON MENTAL RETARDATION

Bidirectional Effects of Expressed Emotion andBehavior Problems and Symptoms in Adolescentsand Adults With Autism

Jan S. Greenberg, Marsha Mailick Seltzer, and Jinkuk HongWaisman Center, University of Wisconsin-Madison

Gael I. OrsmondBoston University

AbstractExpressed emotion measures the emotional climate of the family and is predictive of symp-tom levels in a range of medical and psychiatric conditions. This study extends the inves-tigation of the effects of expressed emotion to families of individuals with autism. A sampleof 149 mothers co-residing with their adolescent or adult child with autism over an 18-month period was drawn from a large multiwave longitudinal study. High expressed emo-tion was related to increased levels of maladaptive behavior and more severe symptoms ofautism over time. Also, characteristics of the son or daughter influenced levels of maternalexpressed emotion over time. Although autism is a complex genetic disorder, the effect ofthe family environment in shaping the behavioral phenotype should not be underesti-mated.

In contrast to the recent proliferation of re-search on the genetic causes, brain structure andfunction, and behavioral repertoire of young chil-dren with autism (Rutter, 2005), there has beenno comparable examination of how the family en-vironment may mediate the manifestation of thecore deficits of autism. In the 1950s and 1960s,mothers of individuals with autism were blamedfor their child’s disorder, as the prevailing theorywas that poor parenting was the precipitating fac-tor (for an historical review, see Folstein & Rosen-Sheidley, 2001). This history of maternal blamewas also characteristic of theories of the etiologyof schizophrenia during the middle 20th century(Fromm-Reichmann, 1948). It is now believedthat autism and schizophrenia are complex genet-ic disorders, with environmental influences aswell, and it is widely accepted that these disordersare not caused by parenting styles. Perhaps in re-action to the history of blaming families of chil-dren with autism, there has been very limited re-

search on the impact of autism on the family orthe influence of the family environment on thedevelopment of the child. Given the centrality ofthe family in influencing the lifelong develop-ment and maintaining the quality of life of per-sons with autism, there is a need for research onthe family environment not only during the earlychildhood years but also when the person withautism reaches adolescence and adulthood. Thepresent study is based on a larger longitudinalstudy with these goals (Greenberg, Seltzer, Krauss,Chou, & Hong, 2004; Orsmond, Krauss, & Selt-zer, 2004; Seltzer et al., 2003).

In a search to identify characteristics of thefamily environment that may influence the courseof a child’s disability, researchers have paid con-siderable attention to the phenomenon of ex-pressed emotion (e.g., Butzlaff & Hooley, 1998).Expressed emotion is conceptualized as a measureof the emotional valence of the household andhas come to be defined as high levels of criticism

230 q American Association on Mental Retardation


Expressed emotion and autism J. S. Greenberg et al.

and/or marked emotional overinvolvement as ex-pressed by a family member about another familymember with an illness or a disability. Expressedemotion was originally conceptualized in terms offive dimensions of family life: criticism, hostility,emotional overinvolvement, warmth, and positivecomments. However, in an early study of the ef-fects of expressed emotion on relapse rates inadults with schizophrenia, Brown, Birley, andWing (1972) found empirically that neitherwarmth nor positive comments were indepen-dently related to relapse. Although hostility washighly correlated with level of criticism, it was thecriticism dimension that was the crucial variablein predicting symptomatic relapse, with emotionaloverinvolvement independently predicting relapsein a small number of cases.

Subsequent studies confirmed that the effectsof criticism and emotional overinvolvement arethe most important dimensions for predicting re-lapse in schizophrenia (Leff & Vaughn, 1985). Thecriticism dimension is designed to tap feelings ofnegativity expressed about the individual with thedisability. The emotional overinvolvement di-mension provides an index of extreme overpro-tectiveness or self-sacrifice by the family member.In a meta-analysis of studies of expressed emotionand psychiatric relapse, Butzlaff and Hooley(1998) reported that the mean effect size for ex-pressed emotion in predicting relapse in schizo-phrenia was .30. The effect sizes were even largerin studies in which investigators examined the ef-fect of expressed emotion on relapse for mooddisorders, r 5 .39, and eating disorders, r 5 .51.Furthermore, evidence implicates expressed emo-tion in predicting symptom levels and relapserates across a broad range of other medical con-ditions, including Alzheimer’s disease, asthma, di-abetes, and Parkinson’s disease (Wearden, Tarrier,Barrowclough, Zastowny, & Rahill, 2000).

More recently, expressed emotion has beenexplored in studies of families of individuals withdevelopmental disabilities (e.g., Beck, Daley, &Hastings, 2004; Dossetor, Nicol, & Stretch, 1994;Greedharry, 1987; Lam, Giles, & Lavander, 2003).In a cross-sectional study of mothers who had achild with a developmental disability and at leastone child without a disability, Beck et al. (2004)found that mothers evidenced higher levels of ex-pressed emotion regarding their child with dis-abilities than their other children and that thechild with disabilities had more behavior prob-lems. Similarly, Lam et al. (2003) found that chil-

dren with intellectual disabilities had more severebehavior problems when their mothers had highversus low expressed emotion, although again thiswas a cross-sectional study. Dossetor et al. (1994)examined levels of expressed emotion in 92 moth-ers of adolescents with developmental disabilitiesin another cross-sectional study. They found thata high level of maternal expressed emotion wasassociated with higher levels of behavior prob-lems. Adolescents of mothers with high emotionaloverinvolvement had more behavior problems inpublic, whereas a high level of maternal criticismwas related to agitation in the adolescent samplemembers. In the only longitudinal study to dateof families with a child with a developmental dis-ability, Hastings, Daley, Burns, and Beck (2006)found that maternal criticism but not overinvol-vement was related to more severe externalizingbehavior problems in the children. However, therelationship between criticism and externalizingbehaviors only held up cross-sectionally and notover time.

Our purpose in the present longitudinal studywas to investigate the bidirectional effects betweenexpressed emotion and maladaptive behaviors andsymptoms in families of individuals with autismspectrum disorders. The study extends our previ-ous cross-sectional study of mothers of adoles-cents and adults with autism, in which we exam-ined the predictors of criticism and emotional ov-erinvolvement and the consequences of expressedemotion for caregiving gains and strains (Ors-mond, Seltzer, Greenberg, & Krauss, 2006). Anenvironment characterized by high levels of ex-pressed emotion is likely to be experienced asstressful to any individual, but in particular to in-dividuals with autism, who often have difficultyregulating their stress responses (Prizant, Wether-by, & Rydell, 2000). In addition to investigatingthe prevalence of high levels of expressed emotionin this population, we also examined the cross-lagged relationships between maternal expressedemotion and the behavior problems and autismsymptoms in their adolescent and adult sons anddaughters with autism. This line of research is im-portant in determining how the family environ-ment affects the manifestation of the symptomsof autism over the life course as well as to eluci-date how these symptoms may reciprocally affectfamily (in this case, maternal) well-being.

The theoretical mechanism explaining howexpressed emotion leads to an escalation in symp-toms is not well-understood. Hooley and Gotlib




(2000) proposed a diathesis-stress model for ex-plaining the relationship between high expressedemotion and clinical outcomes. They suggest thatpersons with a high risk for certain mental disor-ders, such as schizophrenia, may have a height-ened genetic vulnerability or sensitivity to stress;a high level of expressed emotion is widely con-sidered a form of psychosocial stress. The resultsof studies conducted by Goldstein, Rosenfarb,Woo, and Nuechterlein (1994) suggest that per-sons with schizophrenia in high expressed emo-tion families show both more verbal (e.g., delu-sions, suspiciousness, socially inappropriate be-havior) and nonverbal (e.g., hostility) subclinicalsymptoms of psychopathology during interactionswith family members than do individuals withschizophrenia in low expressed emotion families(Goldstein et al., 1994; Rosenfarb, Goldstein,Mintz, & Nuechterlein, 1995; Woo, Goldstein, &Nuechterlein, 2004). Furthermore, when personswith schizophrenia behave in odd or unusualways, family members from high expressed emo-tion households tend to be more likely to respondto the behavior with criticism than in low ex-pressed emotion households. In turn, individualswith schizophrenia become more symptomatic inresponse to these critical comments (Woo et al.,2004). Thus, it appears that in high expressedemotion families, negative verbal and nonverbalbehavior tends to be reciprocated, which, if per-sistent over time, leads to an escalation in behav-ior problems and symptoms.

There is a debate in the literature regardingthe direction of effects between expressed emo-tion and the behavior problems and symptoms ofthe individual with disabilities. A growing numberof investigators conducting prospective treatmentstudies have reported that a reduction in expressedemotion is followed by a reduction in symptomsand a decrease in relapse rates among persons withmajor mental disorders, suggesting that the direc-tion of the causal path is from expressed emotionto behavior problems (for a review, see Butzlaff &Hooley, 1998). However, other researchers haveconcluded that expressed emotion is also a reac-tion by a family member to high levels of symp-toms and behavior problems in the individualwith the disability (Bledin, Kuipers, MacCarthy,& Woods, 1990; King, 2000; Vitaliano, Young,Russo, Romano, & Magana-Amato, 1993), sug-gesting a bidirectional pattern of effects.

Disentangling the causal nature of the rela-tionship between expressed emotion, symptoms,

and behavior problems is particularly complex instudies of parents of children with autism becauseof the unique caregiving challenges these parentsface on a daily basis. Parents may feel responsiblefor anticipating their child’s needs, in part becauseof the difficulties that individuals with autismhave in communicating their needs, causing par-ents to closely monitor their child’s behavior. Forthose parents whose child has highly stereotypedand ritualistic modes of interaction, obsessionalinterests, resistance to change, and/or seems emo-tionally detached or insensitive to the feelings ofothers, expression of frustration and criticism areunderstandable responses. Some parents learn thatclose supervision of their child with autism maybe necessary to protect the child’s safety. Thus,the symptoms and behaviors associated with au-tism demand, to varying degrees, highly involvedparental behavior, and may elicit critical com-ments. As with most parent–child interactions,therefore, the relationship between expressedemotion and symptoms is most likely bidirection-al. Recognizing these bidirectional effects and pat-terns of mutual influence, in this study we ex-amined not only the effects of expressed emotionon the subsequent behavior of the individual withautism but also whether the behaviors and symp-toms of autism affect subsequent levels of mater-nal expressed emotion.

Recently, investigators have begun to explorewhich component of expressed emotion (criticismor emotional overinvolvement) has the strongereffect on the behavior of the individual with thedisability. The available research suggests thatpoorer outcomes for both adults and childrenwith disabilities have more frequently been asso-ciated with high levels of criticism than with highlevels of emotional overinvolvement. Amongstudies of adults, Tarrier, Sommerfield, and Pil-grim (1999) reported that adults with posttrau-matic stress disorder who experienced a high levelof criticism had poorer treatment outcomes thanthose who experienced low levels of criticism,whereas emotional overinvolvement was unrelat-ed to treatment outcomes. Similarly, in a 2-yearlongitudinal study, Kim and Miklowitz (2004)found that a high level of criticism was related toincreased mania and depression in adults with bi-polar disorder, but the level of emotional over-involvement was not predictive of these out-comes.

Studies of children show similar patterns. Inan investigation following young children from




preschool through third grade, Peris and Baker(2000) found that criticism was highly predictiveof whether the child had a diagnosis of ADHDin third grade, but emotional overinvolvementwas not. Even after controlling for preschool be-havior problems and maternal stress, criticism re-mained a significant predictor of ADHD symp-toms in third grade. Similarly, in a study of girlsages 6 to 12 years old, Peris and Hinshaw (2003)found that criticism had a stronger associationwith symptoms of ADHD and aggression thandid emotional overinvolvement. Stubbe, Zahner,Goldstein, and Leckman (1993), in one of the fewstudies based on a community rather than a clin-ical sample of preadolescent children, reportedthat a high level of criticism, but not emotionaloverinvolvement, was related to a child havingdisruptive behavior disorders. Hirshfeld, Bieder-man, Brody, Faraone, and Rosenbaum (1997) alsofound that a borderline or high level of criticismbut not emotional overinvolvement was associat-ed with elevations in mood and behavioral dis-orders in a sample of at-risk children.

However, a few researchers have found a re-lationship between levels of emotional overinvol-vement and child psychopathology. Asarnow,Thompson, Woo, and Cantwell (2001) found el-evated rates of emotional overinvolvement infamilies of children with comorbid ADHD andanxiety disorders. Stubbe et al. (1993) also foundthat a high level of emotional overinvolvementwas related to a diagnosis of anxiety disorder.Thus, although most researchers in this area haveconcluded that criticism is the more powerful pre-dictor of mood and externalizing disorders, a fewhave implicated high levels of overinvolvement inthe development of anxiety disorders.

The present study was based on findings ofthis past research. We examined both overall ex-pressed emotion levels as well as the separate ef-fects of criticism and emotional overinvolvementon the manifestation of behavior problems andautism symptoms. We used a longitudinal designto begin to sort out the direction of effects be-tween maternal expressed emotion and behaviorproblems and symptoms in adolescents and adultswith autism. We recruited a community-basedsample to increase the generalizability of the find-ings.

Specifically, we investigated three related re-search questions: (a) Does the level of expressedemotion, criticism, and emotional overinvolve-ment in families of adolescents and adults with

autism change over an 18-month period of time?(b) Are levels of overall expressed emotion, criti-cism, and emotional overinvolvement predictiveof subsequent behavior problems and autismsymptoms for adolescents and adults with autism?(c) Reciprocally, is the severity of the child’s be-havior problems and autism symptoms predictiveof subsequent levels of overall maternal expressedemotion, criticism, and emotional overinvolve-ment?

In examining the bidirectional relationshipbetween maternal expressed emotion and the sonor daughter’s behavior problems and symptomsof autism, we controlled for the child’s age, gen-der, and a comorbid diagnosis of mental retarda-tion. We controlled for the age of the individualwith autism because research suggests higher lev-els of family stress in adolescence than adulthood(Seltzer & Krauss, 2001; Seltzer et al., 2003) andthat there is an abatement in behavior problemsand autism symptoms as individuals with autismage (Shattuck et al., 2006). Similarly, gender of theindividual with autism was controlled because ithas been shown that mother–child interactionsacross the life course vary with the gender of thechild (Rossi & Rossi, 1990). Finally, we controlledfor a comorbid diagnosis of mental retardation be-cause individuals with mental retardation have ahigh frequency of behavior problems and func-tional limitations in communication and socialskills, which might confound the relationship be-tween expressed emotion and the dependent var-iables in the present analysis.

Based on the existing research, we hypothe-sized that children with autism living in familyenvironments characterized by high maternal ex-pressed emotion would display increasingly severebehavior problems (internalizing, externalizing,and asocial) and symptoms of autism (repetitivebehaviors and circumscribed interests, impair-ments in reciprocal social interactions, and im-pairments in communication) over time than chil-dren living in low expressed emotion families,controlling for prior levels of behavior problemsand symptoms (as well as age, gender, and mentalretardation status). In regard to the componentsof expressed emotion, we hypothesized that highlevels of maternal criticism would be related toincreasingly severe maladaptive behaviors andsymptoms of autism over time. However, basedon past research that showed emotional overin-volvement to have limited effects, we did not ex-pect levels of maternal emotional overinvolve-




ment to be related to changes in the severity ofthe child’s behavior problems or symptoms of au-tism. Recognizing the bidirectional nature of in-fluences of parents and children, we also expectedthat the severity of the child’s behavior problemsand symptoms of autism would have a significanteffect on subsequent levels of maternal expressedemotion, both with respect to criticism and emo-tional overinvolvement. Mothers of children withautism who display more severe behavior prob-lems and autism symptoms will need to be moreinvolved in monitoring all aspects of their child’sbehavior. This may lead to more self-sacrificingand overly protective behavior as well as place agreater burden on parents, which may in turn re-sult in higher levels of criticism of the child.

Method

ParticipantsThe data for this analysis came from a larger

study of 406 families of adolescents and adultswith an autism spectrum disorder who have beenfollowed in an ongoing (4-wave) longitudinalstudy (Seltzer et al., 2003). The families resided inWisconsin (n 5 202) or Massachusetts (n 5 204)when the study began. Identical recruitment pro-cedures were used in the two states (i.e., infor-mational packets sent out by agencies, schools,and diagnostic clinics as well as via announce-ments made through the media). Data for the pre-sent analysis are taken from the second and thirdwaves of the study (Time 2, conducted in 2000–2001, and Time 3, conducted in 2002–2003),which were the first two points of data collectionwhen all the variables of interest in the presentanalysis were included.

The families met three criteria when initiallyrecruited: (a) had a son or daughter age 10 or old-er; (b) child had received a diagnosis on the au-tism spectrum from a medical, psychological, oreducational professional, as reported by the par-ents; and (c) administration of the Autism Diag-nostic Interview–Revised (Lord, Rutter, & Le Cou-teur, 1994) that confirmed the parental report ofan autism spectrum disorder. Of the 406 individ-uals in the sample, 384 (94.6%) met all criteria forautism on the Autism Diagnostic Interview–Re-vised (qualitative impairments in communicationand language; qualitative impairments in recipro-cal social interaction; repetitive, restrictive, andstereotyped behaviors; and onset of symptomsprior to 36 months). The remaining 22 individuals

(5.4% of the sample) demonstrated a pattern ofimpairments on the Autism Diagnostic Interview–Revised that was consistent with their diagnosis ofAsperger’s disorder or pervasive developmentaldisorder–not otherwise specified.

The present analysis was limited to the 149mothers who lived with their son or daughter withautism continuously over the second and thirdwaves of the study. We restricted the sample inthis way because the effects of expressed emotionrequire high levels of face-to-face contact betweenparent and child (Leff & Vaughn, 1985). We re-stricted the expressed emotion data collection tomothers because of evidence that expressed emo-tion from mothers and fathers correlate differentlywith various outcomes (King & Dixon, 1995) andbecause mothers tend to be the primary caregiversof individuals with developmental disabilities (Es-sex & Hong, 2005). Participants not included inthe present analysis were mothers whose son ordaughter lived away from home at Time 2 andTime 3 (n 5 125), mothers whose child movedfrom the parental home between Time 2 andTime 3 (n 5 14), mothers who participated atTime 1 but whose son or daughter died by Time3 (n 5 4), mothers who died before Time 3 (n 55), cases for which the child’s father was the pri-mary respondent for the interview (n 5 12), moth-ers who declined participation or could not belocated at Time 2 or Time 3 (n 5 59), and casesfor which there were missing data on the ex-pressed emotion measure at either Time 2 or Time3 (n 5 20).

By restricting the sample to adolescents andadults with autism who had continuously lived athome, it is possible that we introduced a selectionbias into this analysis. Specifically, this would bethe case if high levels of expressed emotion, be-havior problems, or autism symptoms were factorsprecipitating the move from the parental home.Therefore, we explored whether families in whichthe child moved from the parental home betweenTime 2 and Time 3 (n 5 14) differed from familiesin which the child remained living at home (n 5149). The two groups did not differ significantlyon any of the measures of behavior problems orautism symptoms. Also, they did not differ onoverall expressed emotion level at Time 2 or anyof the expressed emotion measures at Time 3.However, mothers whose child moved from theparental home between Times 2 and 3 had higherlevels of criticism at Time 2 than did motherswhose child remained at home (1.14 vs. .66, re-




spectively), t 5 2.21, p 5 .028, and higher levelsof overinvolvement (.93 vs. .56), t 5 2.12, p 5.048, respectively. Given these differences, we re-ran the analysis including these 14 cases. The re-lationships between the measures of expressedemotion and its components and maladaptive be-haviors or autism symptoms did not change.Thus, we concluded that dropping these 14 casesdid not introduce a selection bias.

The adolescents and adults with autism spec-trum disorder included in the subsample used inthis analysis ranged in age from 11.3 to 48.9 yearsat Time 2, with a mean age of 19.9 years (SD 57.84). The majority of the sample was male(75.5%), reflective of the higher prevalence of au-tism in males compared with females (AmericanPsychiatric Association, 2000). Fifty-seven percentof the participants had been given a diagnosis ofmental retardation.

The mothers ranged in age from 34.0 to 80.8years at Time 2 (M 5 49.3). Most were married(81.6%). Nearly all had completed high school(98.0%), and almost three quarters (70.1%) wereemployed either part- or full-time. The average an-nual household income in 2000–2001 was$45,800. Most mothers were White; only 7.4%were persons of color.

Procedure and MeasuresAt both Times 2 and 3, mothers participated

in a 2- to 3-hour in-home interview and complet-ed self-administered questionnaires. The twopoints of data collection were separated by 18months.

Expressed emotion. The Five-Minute SpeechSample was used to code the level of expressedemotion, following the standardized proceduresdescribed in the coding manual developed by Ma-gana et al. (1986). This measure has been usedwidely in studies of expressed emotion in a varietyof diagnostic groups (for a review see Van-Hum-beeck, Van-Audenhove, De-Hert, Pieters, &Storms, 2002). The mother was asked to speakabout her son or daughter for 5 minutes withoutinterruptions. The speech sample was tape record-ed, transcribed, and rated as high (coded 5), bor-derline (3), or low (0) along the dimensions of crit-icism and emotional overinvolvement. The moth-er was categorized as high in expressed emotionif she was rated as high on criticism and/or emo-tional overinvolvement. Although initially ex-pressed emotion was seen as a unitary dimensionreflecting high levels of criticism and/or high lev-

els of emotional overinvolvement, more recentlythe components of expressed emotion have beenevaluated separately and shown to have differenteffects on different types of behaviors and symp-toms (e.g., Peris & Baker, 2000; Stubbe et al.,1993). Therefore, we examined both the effects ofoverall expressed emotion and its components onbehavior problems and autism symptoms.

The operational definitions of the expressedemotion components of criticism and overinvol-vement in the Five-Minute Speech Sample weredeveloped based on the more extensive Camber-well Family Interview (Magana et al., 1986). Re-spondents are rated as high on criticism if they (a)make a negative opening remark, (b) describe theirrelationship with their son or daughter in negativeterms, or (c) make one or more criticisms abouttheir son or daughter during the course of the 5-Minute Speech Sample. Respondents are rated asborderline on criticism if they do not satisfy theabove requirements but, nevertheless, make oneor more statements of dissatisfaction with the re-lationship with their son or daughter. A rating oflow is given in the absence of any critical com-ments. The following transcript, which representsa composite case, provides an example of how amother classified as high in criticism speaks aboutthe relationship with her child.

What kind of a person he is? Well he’s a very uniqueindividual. How can I put this? One of the things about Jon-nie is that I really like him when he’s in a good mood butwhen he’s in a bad mood, it’s impossible for us. I still cannottrust him being alone in the house without destroying some-thing. He’s compulsive, and he’s so controlling, especiallywith me. The tension in the household gets absolutely in-credible. He is very very stubborn. He’ll tend to scream tothe point where my eardrums vibrate at that point. I supposethat’s his only way to fight something he’s scared of. Hedoesn’t have any social skills. One of the biggest problems iswhat to do to fill his time, he gets bored real easily and hasa lot of compulsive behaviors that are difficult to deal with.He hates change and doesn’t want to move but he’s becomingimpossible to live with.

Emotional overinvolvement is rated as high ifthe family member either expresses excessive self-sacrificing or overprotective feelings toward theson or daughter or experiences an emotional dis-play (e.g., crying) during the administration of theFive-Minute Speech Sample. Also, emotional ov-erinvolvement is rated as high if two of the fol-lowing behaviors are present: excessive detailabout the son or daughter’s past, a statement ofattitude (e.g., ‘‘I will do anything for my child’’),or excessive praise of the son or daughter, as in-




dicated by five or more positive remarks. A ratingof borderline emotional overinvolvement is as-signed if there is some evidence of overinvolve-ment (e.g., excessive detail about the past or ex-cessive praise), but the full criteria for high emo-tional overinvolvement are not met. Low emo-tional overinvolvement represents the absence ofsuch indicators.

The concept of overinvolvement as opera-tionalized in expressed emotion research is easilymisunderstood because many parents of childrenwith autism have been unfairly labeled by profes-sionals as overinvolved when they advocate forservices for their child. In the context of expressedemotion research, emotional overinvolvement isnot indicated by advocacy behavior or by protec-tive behaviors that are necessary and reasonablewithin the context of ensuring the safety and well-being of a child with a disability. Rather, the con-cept emotional overinvolvement is meant to cap-ture an extreme overidentification with the childthat does not allow the child to respond as anautonomous individual, and/or behavior that isoverly protective given the child’s developmentaland functional capabilities. Another aspect ofemotional overinvolvement is excessive praise ofthe child with autism. The following transcriptprovides an example of a family classified as highin emotional overinvolvement:

Susie is almost 17 years old. She is very intelligent, cre-ative, and brave. She is a very moral person. At times becauseof her needs, she is the focus, the central focus, of our family.Our communication is, a lot of times, not verbal. Throughouther life, I have been her interpreter, interpreting her to theworld and then interpreting the world to her. I’ve always beenthe one who could understand that what she was saying wasnot meaningless. I oftentimes wonder how much of my iden-tity is wrapped up in her. Where is the line? Where does Susieend and I begin?

Mothers are classified as high in expressedemotion if they are rated as high on criticism and/or emotional overinvolvement. In the presentstudy, all expressed emotion ratings were per-formed by a rater with 20 years of experience incoding the Five-Minute Speech Sample. In addi-tion, a second experienced rater independentlycoded 30 taped speech samples. Their level ofagreement was 83.3%, k 5 .67. Several researchershave found that there is good correspondence be-tween how families are rated on the Five-MinuteSpeech Sample and the full Camberwell FamilyInterview (Magana et al., 1986; Moore & Kuipers,1999).

Behavior problems. The measure of behaviorproblems was based on the Problem Behavior sub-scale of the Scales of Independent Behavior–Re-vised (Bruininks, Woodcock, Weatherman, &Hill, 1996), which consists of eight behavior prob-lems: hurtful to self, unusual or repetitive, with-drawn or inattentive, socially offensive, uncoop-erative, hurtful to others, destructive to property,and disruptive. The mother was asked whether herson or daughter manifested each of these eightbehavior problems within the last 6 months and,if so, to rate the frequency (1, less than once a monthto 5, one or more times an hour) and severity (1, notserious to 5, extremely serious) of the behavior. Stan-dardized algorithms (Bruininks et al., 1996) wereused to translate frequency and severity ratingsinto three subscales scores: Internalized Maladap-tive Behavior (hurtful to self, unusual or repetitivehabits, and withdrawn or inattentive behavior),Asocial Maladaptive Behavior (socially offensiveand uncooperative behavior), and ExternalizedMaladaptive Behavior (hurtful to others, destruc-tive to property, and disruptive behavior). Bruin-inks et al. have previously documented the reli-ability and the validity of the Scales of Indepen-dent Behavior–Revised. In our study, the corre-lations between the Time 2 and Time 3 measureswere all above .60 (Externalizing, r 5 76; Inter-nalizing, r 5 .64, and Asocial, r 5 .63), indicatinghigh levels of stability over time. Higher Scales ofIndependent Behavior–Revised scores indicatemore severe behavior problems.

Autism symptoms. The Autism Diagnostic In-terview-Revised (Lord et al., 1994) was used tomeasure autism symptoms. This measure is a stan-dardized investigator-based interview conductedwith a primary caregiver (in our case, with themother of the individual with autism). Theseitems are based on the Diagnostic and StatisticalManual of Mental Disorder-fourth edition DSM-IV-TR (American Psychiatric Association, 2000) andInternational Classification of Diseases (WorldHealth Organization, 1990) criteria for an autismdiagnosis. We administered the items that com-prise the diagnostic algorithm of the Autism Di-agnostic Interview-Revised (C. Lord, personalcommunication, February 1999). A code of 0 in-dicates no abnormality present; 1, possible abnormal-ity; and 2, definite abnormality. Some items werecoded on a 0- to 3-point scale, with a score of 3indicating extreme abnormality with respect to the be-havior. As suggested by Lord et al. (1994), we com-bined scores of 3 and 2. Items were coded twice




at Time 1, once to reflect the severity of the symp-tom when the child was age 4 to 5 and again toreflect current level of symptom severity. AtTimes 2 and 3, only current ratings were obtained,and these were the data analyzed for the presentanalysis.

The Autism Diagnostic Interview-Revisedyields ratings for the three primary symptom clus-ters used in the diagnosis of autism: repetitive be-haviors and restricted interest, impairments in re-ciprocal social interaction, and impairments incommunication. The Repetitive Behaviors andRestricted Interests domain includes seven autisticsymptoms: circumscribed interests, unusual pre-occupations, compulsion/rituals, hand and fingermannerisms, other complex mannerisms/bodymovements, repetitive use of objects/interest inparts, and unusual sensory interests. An eighthitem, verbal rituals, was dropped for this analysisbecause it is not administered to sample memberswho are nonverbal (16.1% of the present subsam-ple, n 5 24). A total score of the seven items wascalculated for Times 2 and 3.

Impairments in Reciprocal Social Interactionwere measured by the sum of the ratings of 14items measuring qualitative impairments in recip-rocal social interaction (direct gaze, social smiling,range of facial expression, interest in people, re-sponse to others’ approaches, friendships, direct-ing attention, offering to share, sharing enjoymentwith others, use of other’s body, offering comfort,quality of social overtures, inappropriate facialexpressions, and appropriateness of social re-sponse).

The impairments in the Communication do-main of the Autism Diagnostic Interview–Revisedconsists of 11 symptoms of autism. However, 7of the symptoms are not administered if the in-dividual is nonverbal and, thus, were not analyzedhere. The 4 remaining items (i.e., pointing to ex-press interest, gestures, nodding, and head-shak-ing) were administered to all sample members(verbal as well as nonverbal) and were summed toderive an overall score indicating the adolescent’sor adult’s current level of impairments in nonver-bal communication.

The interviewers who administered the Au-tism Diagnostic Interview–Revised participated inan approved training program for conducting thisstructured interview. All interviews were tape re-corded. Interrater reliability for individual AutismDiagnostic Interview–Revised items between the

interviewers and two supervising PhD clinical psy-chologists experienced in the diagnosis of autismand in the use of the Autism Diagnostic Inter-view–Revised averaged 89%, mean k 5 .81. Pastresearch has demonstrated the test–retest reliabil-ity, diagnostic validity, convergent validity, andspecificity and sensitivity of the Autism Diagnos-tic Interview–Revised (Hill, Bolte, & Petrova,2001; Lord, Pickles, & McLennan, 1997).

Characteristics of the son/daughter with autismspectrum disorder and of the mother. We examinedthree characteristics of the son or daughter withautism: age, gender, and a comorbid diagnosis ofmental retardation. Age was coded in years. Gen-der was coded 0 (male) or 1 (female). Mental re-tardation status was determined using a variety ofsources of information. Standardized intelligenceand adaptive behavior measures were adminis-tered by interviewers trained and supervised by aPhD clinical psychologist. Intelligence was as-sessed through use of the Wide Range IntelligenceTest (Glutting, Adams, & Sheslow, 2000) andadaptive behavior was assessed by the VinelandScreener (Sparrow, Carter, & Cicchetti, 1993). In-dividuals with standard scores of 70 or below onboth measures were classified as having mental re-tardation, consistent with diagnostic guidelines(Luckasson et al., 2002). Those with scores above75 on either of the measures were classified as nothaving mental retardation.

For those sample members with scores be-tween 71 and 75 on one or both measures or forwhom either of the measures was missing, a re-view of records combined with a clinical consen-sus procedure was used to determine mental re-tardation status. This determination was conduct-ed by three psychologists (one master’s level andtwo PhDs), who reviewed each case file indepen-dently. Information in the file included the stan-dardized measures (Wide Range Intelligence Testor Vineland, when available), parent report of pri-or diagnoses, intellectual functioning, adaptive be-havior, and clinical and school records (whenavailable). Agreement among the three indepen-dent raters was reached on 77.8% of cases follow-ing the initial review. All cases in which there wasdisagreement among the raters were discussed un-til consensus was reached as to whether the indi-vidual had mental retardation. For this analysis, acomorbid diagnosis of mental retardation wascoded 1 or 0 if the child did not have a diagnosisof mental retardation.




Table 1. Distribution of Expressed Emotion,Criticism, and Emotional Overinvolvement in theSample (N 5 149)

Expressedemotion &subscales

Time 2

n %

Time 3

n %

Expressed Emotion

Low 108 72.5 117 78.5High 41 27.5 32 21.5

Criticism

Low 78 52.3 63 42.3Borderline 44 29.5 61 40.9High 27 18.1 25 16.8

EmotionalOverinvolvement

Low 81 54.4 87 58.4Borderline 53 35.6 51 34.2High 15 10.1 11 7.4

Results

Prevalence of High Levels of ExpressedEmotion

For our first research question, we askedabout changes in the level of expressed emotionover an 18-month period in families of adoles-cents and adults with autism. As shown in Table1, high levels of expressed emotion were found inonly about a quarter of the families at Time 2 andabout a fifth of the families at Time 3. There wasconsiderable stability in levels of expressed emo-tion over the 18-month study period. Approxi-mately 61.7% had low expressed emotion at bothwaves of data collection, and 10.7% had high ex-pressed emotion at both data points. About 17%of the sample showed a reduction in expressedemotion (high at Time 2 but low at Time 3) and10.7% showed an elevation in expressed emotion(low at Time 2 and high at Time 3). Similarly,relatively few mothers had high ratings on eithercriticism or overinvolvement. However, as shownin Table 1, mothers were approximately twice aslikely to be classified as high on criticism as onoverinvolvement. Whereas relatively few familieswere high on criticism or overinvolvement, a larg-er number fell in the borderline category. Ap-proximately 30% of the mothers had borderlinescores on the criticism dimension at Time 2,which increased by Time 3 to about 40%. The

percentage of families in the borderline emotionaloverinvolvement group was stable over time, withabout a third of the sample classified as borderlineat each of the two waves of data collection. Thus,although we found relatively low rates of highoverall expressed emotion, criticism, or overinvol-vement, a more substantial number of families fellinto the borderline categories.

Bidirectional EffectsOur second and third research questions con-

cerned the cross-lagged effects between behaviorproblems, autism symptoms, and expressed emo-tion. A cross-lagged correlational approach hasbeen used in other studies of the bidirectional ef-fects of the family environment and the behaviorof children with developmental disabilities (e.g.,Mink & Nihira, 1987; Orsmond, Seltzer, Krauss,& Hong, 2003) and in studies of expressed emo-tion effects (e.g., King, 2000). To estimate bidirec-tional effects, we used path analysis for a two-wavecross-lagged effects model (Finkel, 1995) usingLISREL 8.54 to obtain maximum likelihood esti-mates of the paths. Each variable at Time 3 ispredicted by the Time 2 value of the variable aswell as the Time 2 value of the cross-lagged vari-able and control variables (child’s age, gender, andwhether child had a diagnosis of mental retarda-tion). The variables measured at the same pointin time were allowed to be correlated. A separatepath model was performed for each measure ofbehavior problems and autism symptoms. Be-cause all possible paths are estimated, the modelis saturated so that LISREL fit indices indicate aperfect fit of the data. As an alternative estimateof the fit of the models, we report the percentageof variance explained for each of the Time 3 out-comes.

The Pearson product-moment correlationsamong all study variables at Time 2 and Time 3are shown in Table 2. However, the path analysiswas based on a mixed matrix of Pearson correla-tions and polychloric/polyserial correlations(available from the first author). We computedpolychloric/polyserial correlations for correlationsinvolving expressed emotion, criticism, and emo-tional overinvolvement because these variables aremeasured as dichotomous or ordinal variables butconceptually represent underlying continuousconstructs. All other variables analyzed were basedon Pearson product-moment correlations.

Our preliminary analyses indicated that theage of the child with autism had no effect on any




Table 2. Pearson Correlations Among Study Variables by Time

Variablea 1 2 3 4 5 6 7 8 9

1. T2 EE 1.02. T2 Criticism .57 1.03. T2 EOI .36 2.18 1.04. T2 SIB-R Externalizing .13 .29 2.05 1.05. T2 SIB-R Internalizing .22 .22 .10 .46 1.06. T2 SIB-R Asocial .19 .18 .11 .62 .55 1.07. T2 ADI-R .09 .09 2.01 .32 .37 .29 1.0

Repetitive Behaviors8. T2 ADI-R .03 .13 2.13 .19 .39 .18 .28 1.0

Social Reciprocity9. T2 ADI-R Nonverbal 2.05 2.08 2.08 2.00 .14 .03 .03 .59 1.0

CommunicationImpairment

10. T3 EE .26 .26 .08 .03 2.02 .03 .11 .05 2.0511. T3 Criticism .26 .36 2.03 .21 .06 .19 .12 .10 .0112. T3 EOI 2.00 2.18 .32 2.10 2.04 .03 2.05 2.09 2.0313. T3 SIB-R Externalizing .28 .31 .04 .76 .47 .58 .27 .24 .0614. T3 SIB-R Internalizing .27 .36 .08 .41 .64 .37 .27 .39 .0815. T3 SIB-R Asocial .33 .31 .16 .54 .37 .63 .14 .13 .0416. T3 ADI-R .14 .26 2.03 .31 .44 .38 .59 .33 .08

Repetitive Behaviors17. T3 ADI-R .16 .18 2.06 .20 .33 .15 .27 .87 .54

Social Reciprocity18. T3 ADI-R Nonverbal 2.09 2.11 2.07 .02 .07 .04 .07 .57 .83

CommunicationImpairment

19. Genderb .09 .04 2.08 .03 .01 2.03 .06 .02 .0320. MR statusc .08 .11 2.01 .08 .18 .08 .15 .39 .3821. Child’s age 2.06 2.12 2.07 2.17 2.05 2.14 2.02 .21 .24Mean .28 .66 .56 104.98 113.37 110.50 4.66 14.79 4.69SD .45 .77 .67 9.79 9.55 11.39 2.09 6.12 2.59

Note. All correlations of .16 to .21 significant at p , .05; .22 to .25, p , .01; .26 and above, p , .001.aT2 5 Time 2, EE 5 expressed emotion, EOI 5 emotional overinvolvement, SIB-R 5 Scales of Independent Behavior–Revised, ADI-R 5 Autism Diagnostic Interview–Revised, T3 5 Time 3. b1 5 female, 0 5 male. c1 5 child has mentalretardation diagnosis, 0 5 otherwise.

measures of Time 3 expressed emotion, autismsymptoms, or behavioral outcomes, and droppingit from the model did not have any effect on thefindings. Therefore, to achieve a more parsimo-nious model, we dropped age from the final mod-el. Figure 1 provides a diagram of the conceptualmodel we analyzed.

Table 3 shows the relationships between ex-ternalizing, internalizing, and asocial maladaptivebehaviors and the measures of expressed emotion.Model 1 is based on the overall expressed emo-

tion score, with Models 2 and 3 reporting the co-efficients for the criticism and emotional overin-volvement domains separately. As shown in Table3, there was considerable stability in the Scales ofIndependent Behavior–Revised measures of be-havior problems over time. The stability coeffi-cient for externalizing behavior was .71. For in-ternalizing behaviors, the coefficient was .58, andfor asocial behaviors, .55. The stability coefficientsfor Scales of Independent Behavior–Revised mea-sures in the remaining models ranged from .57 to




Table 2. Extended

10 11 12 13 14 15 16 17 18 19 20 21

1.0.68 1.0.35 .02 1.0.20 .35 2.04 1.0.13 .21 2.09 .57 1.0.15 .31 2.04 .66 .50 1.0.12 .20 2.11 .34 .43 .33 1.0

.12 .16 2.08 .29 .38 .19 .29 1.0

2.06 2.04 .12 .05 .07 2.01 .10 .51 1.0

.17 .07 .10 2.02 2.02 .04 .04 .09 .08 1.02.01 .14 2.19 .18 .18 .23 .23 .42 .26 2.03 1.02.12 2.09 2.15 2.16 2.12 2.15 2.04 .18 .25 2.11 .23 1.0

.22 .75 .50 106.09 113.20 111.11 4.62 14.01 4.56 .77 .57 19.94

.41 .73 .63 11.41 10.11 12.16 2.30 6.03 2.54 .43 .50 7.84

Figure 1. Conceptual model of relationship be-tween expressed emotion and behavior problemsand symptoms of autism.

.75 and were highly significant in each model. Ex-pressed emotion and the two dimensions of crit-icism and emotional overinvolvement were lessstable (ranging from .41 to .46), but all of thecoefficients were highly significant, ps , .001.

Next, we examined the cross-lagged effects ofthe three Scales of Independent Behavior–Revisedmaladaptive behavior subscales and overall levelof expressed emotion, as shown in Model 1 ofTable 3. Consistent with our hypothesis, a highlevel of expressed emotion at Time 2 was signifi-cantly related to increasing levels of externalizingmaladaptive behavior in the son or daughter withautism 18 months later at Time 3, b 5 .21. How-ever, externalizing behavior problems at Time 2




Table 3. Standardized Path Coefficients for Two-Wave Cross-Lagged Models for Subscales of the Scalesof Independent Behavior-Revised (SIB-R)

Modela

Behavior

Externalizing Internalizing Asocial

Model 1: Expressed EmotionsStability effects

T2 EE to T3 EE .43*** .46*** .43***T2 SIB-R to T3 SIB-R .71*** .58*** .55***

Cross-lagged effects

T2 EE to T3 SIB-R .21*** .18*** .27***T2 SIB-R to T3 EE 2.04 2.15* 2.05

Control variables

Gender of child to T3 SIB-R 2.06 2.04 .02Gender of child to T3 EE .21*** .21*** .21***MR to T3 SIB-R .10* .06 .15**MR to T3 EE 2.05 2.03 2.05

R2

T3 EE .24 .26 .24T3 SIB-R .63 .45 .50

Model 2: CriticismStability effects

T2 Criticism to T3 Criticism .41*** .45*** .41***T2 SIB-R to T3 SIB-R .72*** .57*** .57***


T2 Criticism to T3 SIB-R .10 .26*** .22***T2 SIB-R to T3 Criticism .09 2.07 .13

Control variables

Gender of child to T3 SIB-R 2.04 2.03 .05Gender of child to T3 Criticism .06 .06 .06MR to T3 SIB-R .11* .05 .15**MR to T3 Criticism .11 .12 .11

R2

T3 Criticism .23 .22 .23T3 SIB-R .59 .48 .48

Model 3: Emotional Overinvolvement (EOI)Stability effects

T2 EOI to T3 EOI .42*** .43*** .43***T2 SIB-R to T3 SIB-R .75*** .63*** .60***


T2 EOI to T3 SIB-R .08 .02 .11T2 SIB-R to T3 EOI 2.10 2.05 .01

(Table continued)




Table 3. Continued

Modela

Behavior

Externalizing Internalizing Asocial

Control variables

Gender of child to T3 SIB-R 2.03 2.02 .07Gender of child to T3 EOI .17** .17** .17**MR to T3 SIB-R .12* .07 .18***MR to T3 EOI 2.19*** 2.19** 2.20***

R2

T3 EOI .25 .24 .24T3 SIB-R .59 .42 .44

aExpressed emotion, SIB-R 5 Scales of Independent Behavior–Revised, T2 5 Time 2, T3 5 Time 3.*p , .05. **p , .01. ***p , .001.

did not have a reciprocal cross-lagged effect onexpressed emotion at Time 3, b 5 2.04. Withrespect to internalizing behaviors, high levels ofexpressed emotion at Time 2 were related to in-creasingly severe internalizing behavior problemsat Time 3, b 5 .18. In addition, internalizing be-havior problems had a significant reciprocal effecton levels of expressed emotion at Time 3, b 52.15, but the direction of the effect was the op-posite of our expectation. A high level of inter-nalizing behavior problems at Time 2 was relatedto decreasing expressed emotion at Time 3. Final-ly, a high level of expressed emotion at Time 2was related to increasing asocial maladaptive be-haviors at Time 3, b 5 .27, but asocial maladap-tive behavior at Time 2 was unrelated to levels ofexpressed emotion at Time 3, b 5 2.05. Themodel explained approximately 25% of the vari-ance in levels of expressed emotion at Time 3, and63%, 45%, and 50% of the variance in Time 3externalizing, internalizing, and asocial maladap-tive behaviors, respectively.

Model 2 displays the cross-lagged effects ofcriticism and the three behavior problems sub-scales. Contrary to our hypothesis, there were nosignificant cross-lagged effects between levels ofcriticism and externalizing behavior. With respectto internalizing behaviors, a high level of criticismat Time 2 was related to an increase in the severityof internalizing maladaptive behaviors at Time 3,but internalizing maladaptive behaviors at Time 2did not have a cross-lagged effect on criticism atTime 3. Similarly, a higher level of criticism atTime 2 was related to a significant increase in theseverity of the child’s asocial behaviors at Time 3,

but the severity of the child’s asocial behaviorproblems at Time 2 did not have an effect onlevels of criticism at Time 3.

Model 3 reports the cross-lagged effects ofemotional overinvolvement and the three behav-ior problems subscales. As expected, emotionaloverinvolvement at Time 2 had no effect on theseverity of externalizing, internalizing or asocialbehaviors at Time 3. In addition, there was nocross-lagged effect between externalizing, internal-izing, or asocial behavior problems at Time 2 andlevels of emotional overinvolvement at Time 3.

Turning from behavior problems to autismsymptoms, Table 4 reports the path coefficientsbetween overall expressed emotion and its com-ponents and the three Autism Diagnostic Inter-view–Revised subscales: Repetitive Behaviors andRestricted Interests, Impairments in Social Reci-procity, and Impairments in Nonverbal Commu-nication. Again, there was a high level of stabilityin the symptoms manifested by the adolescent oradult with autism. The stability coefficientsranged from .55 to .85, quite similar in magnitudeas the stability coefficients for the Scales of In-dependent Behavior–Revised subscales.

As shown in Model 1, a high level of ex-pressed emotion at Time 2 was associated withincreasing impairments in reciprocal social inter-action at Time 3. Contrary to our hypothesis,there were no cross-lagged effects in either direc-tion between levels of expressed emotion and theseverity of repetitive behaviors and restricted in-terests or impairments in nonverbal communica-tion. The model explained approximately 25% ofthe variance in levels of expressed emotion at




Table 4. Standardized Path Coefficients for Two-Wave Cross-Lagged Models for Autism DiagnosticInventory-Revised (ADI-R) Measures of Symptoms

Modela

Repetitivebehaviors &

restricted interests

Impairments inreciprocal social

interactions

Impairments innonverbal

communication

Model 1: Expressed EmotionsStability effects

T2 EE to T3 EE .41*** .42*** .42***T2 ADI-R to T3 ADI-R .56*** .83*** .84***


T2 EE to T3 ADI-R .11 .18*** 2.07T2 ADI-R to T3 EE .11 .08 2.03

Control variables

Gender of child to T3 ADI-R 2.01 .05 .06Gender of child to T3 EE .20*** .21*** .21***MR to T3 ADI-R .14* .08* 2.06MR to T3 EE 2.07 2.08 2.04

R2

T3 EE .25 .25 .24T3 ADI-R .38 .80 .69

Model 2: CriticismStability effects

T2 Criticism to T3 Criticism .43*** .43*** .44***T2 ADI-R to T3 ADI-R .55*** .82*** .84***


T2 Criticism to T3 ADI-R .23*** .07 2.04T2 ADI-R to T3 Criticism .07 2.00 .01

Control variables

Gender of child to T3 ADI-R 2.00 .07 .05Gender of child to T3 Criticism .06 .06 .06MR to T3 ADI-R .12 .10* 2.06MR to T3 Criticism .10 .11 .11

R2

T3 Criticism .22 .22 .22T3 ADI-R .42 .78 .69

Model 3: EmotionalOverinvolvement (EOI)

Stability effects

T2 EOI to T3 EOI .43*** .44*** .44***T2 ADI-R to T3 ADI-R .57*** .84*** .85***

Cross-lagged effect

T2 EOI to T3 ADI-R 2.02 .06 .00T2 ADI-R to T3 EOI 2.04 .08 .12

(Table continues)




Table 4. Continued

Modela

Repetitivebehaviors &

restricted interests

Impairments inreciprocal social

interactions

Impairments innonverbal

communication

Control variables

Gender of child to T3 ADI-R .01 .08* .05Gender of child to T3 EOI .17** .17** .16*MR to T3 ADI-R .15* .10* 2.07MR to T3 EOI 2.19*** 2.23*** 2.24***

R2

T3 EOI .24 .25 .25T3 ADI-R .37 .78 .69

aExpressed emotion, ADI-R 5 Autism Diagnotic Interview-Revised, T2 5 Time 2, T3 5 Time 3.*p , .05. **p , .01. ***p , .001.

Time 3, and 38%, 80%, and 69%, of the variancein Time 3 repetitive behaviors, impairments in re-ciprocal social interactions, and impairments innonverbal communication, respectively.

Regarding the components of expressed emo-tion, level of criticism at Time 2 was significantlyrelated to worsening Autism Diagnostic Inter-view–Revised repetitive behaviors at Time 3(Model 2). However, the reciprocal cross-laggedeffect was not significant. In addition, there wereno effects in either direction between maternalcriticism and either the Autism Diagnostic Inter-view–Revised reciprocal social interaction or non-verbal communication impairment domains.

Model 3 in Table 4 shows the relationshipbetween emotional overinvolvement and thethree Autism Diagnostic Interview–Revised sub-scales. Consistent with our expectation, higherlevels of emotional overinvolvement at Time 2were not related to an elevation in repetitive be-haviors and restricted interests, impairments in re-ciprocal social interactions, or nonverbal com-munication impairments at Time 3. Also, therewere no cross-lagged effects between the symp-toms of autism at Time 2 and levels of overin-volvement at Time 3.

In each model reported above, the associa-tions between the control variables (gender andmental retardation status) and the measures of be-havior problems and autism symptoms are dis-played. Although the coefficients (and levels ofsignificance) varied somewhat from model tomodel, the general pattern was the same. As pre-dicted, those who had a diagnosis of mental re-tardation had higher levels of externalizing and

asocial maladaptive behavior, more severe repeti-tive behaviors, and more severe impairments insocial reciprocity. Female sample members weremore impaired in reciprocal social interactions.

We also examined the extent to which thecontrol variables predicted expressed emotion inall models. Mothers of female sample membershad higher levels of overall expressed emotion aswell as higher levels of emotional overinvolve-ment. Mothers of sample members with mentalretardation had lower levels of overinvolvement.Thus, characteristics of the son or daughter withautism conditioned maternal expression of over-involvement.

DiscussionWe found strong support for the hypothesis

that high levels of maternal expressed emotionwould have detrimental effects on the well-beingof individuals with autism. For four out of six de-pendent variables, high levels of overall expressedemotion led to increasing levels of maladaptivebehaviors and/or more severe autism symptomsover time. Consistent with the general literatureon expressed emotion, a high level of overall ex-pressed emotion in mothers of adolescents andadults with autism was related to increasing sever-ity and intensity of externalizing, internalizing,and asocial maladaptive behaviors over time, con-trolling for prior levels of behavior problems.High levels of overall expressed emotion had sim-ilar, though weaker, effects on the symptoms ofautism, leading to intensifying impairments in re-ciprocal social interactions.




Consistent with prior research, the level ofcriticism appears to be driving the relationship be-tween overall expressed emotion and elevation insymptoms. A high level of criticism was signifi-cantly related to increasing levels of internalizingand asocial maladaptive behaviors and to increas-ing severity of repetitive behaviors over time. Itwas only with respect to impairments in nonver-bal communication (i.e., gestures, pointing, nod-ding, and head-shaking) that neither overall ex-pressed emotion nor criticism had a significant ef-fect. Our findings are consistent with the moregeneral literature on the effects of parental criti-cism on maladaptive behavior in children (e.g.,Ge, Best, Conger, & Simons, 1996; G. Patterson,1982).

Mink and Nihira (1987) were among the firstto examine the bidirectional parent/child influ-ences in families of children with mental retar-dation. They found that the direction of influencebetween parent and child depended on the typeof family, with the influence flowing from parentto child in highly cohesive families and from childto parent in control-oriented and child-orientedfamilies. Our findings here parallel their findingsfor cohesive families. Although it would havebeen informative to replicate their family typolo-gy in our sample, we did not have the data to doso. We were, however, able to examine the levelsof family cohesion based on data collected at thefirst wave of our study, when the Family Cohe-sion subscale from the FACES II (Olson et al.,1983) was administered. The average level of fam-ily cohesion was 60.4 (scale ranges from 16 to 80),indicating a moderate level of family closeness. Inaddition, most mothers in our study reportedclose relationships with their child who had au-tism (Orsmond et al., 2006). Thus, our findingsappear to lend some support to Mink and Nihira’s(1987) earlier findings about the direction of ef-fects in cohesive families. An important goal forfuture researchers is to examine whether the rela-tionship between expressed emotion and behaviorproblems and symptoms of children with autismis moderated by family type.

Although our focus in this paper was on thedetrimental effects of high levels of expressedemotion on the behavior problems and symptomsof adolescents and adults with autism, we notethat the majority of families in our sample did notevidence elevated levels of expressed emotion. Onaverage, the levels of expressed emotion in thesefamilies were lower than those reported in families

dealing with psychiatric disorders (e.g., Barrow-clough & Parle, 1997; P. Patterson, Birchwood, &Cochrane, 2000; Scazufca & Kuipers, 1996). Suchpatterns may indicate a pattern of family strengthand effective coping with the stresses of autismand suggest a positive influence of the effect ofthe family environment on the development ofindividuals with autism spectrum disorders.

Another explanation for the low rates of ex-pressed emotion in our sample is suggested by re-search on the relationship between expressedemotion and parental attributions about the de-gree to which the child has control over his or hersymptoms. There is growing evidence that highexpressed emotion occurs when the family mem-bers perceive that the individual with disabilityhas control over his or her symptoms and behav-iors (Greenley, 1986; Hooley, 1987). As part ofthe larger study, we asked mothers to tell us theirbeliefs regarding why their child developed au-tism. Many mothers attributed the cause of theirchild’s condition to factors beyond their child’scontrol, such as birth complications, genetics, orenvironmental toxins. Thus, the relatively low rateof high expressed emotion in our sample may berelated to the fact that many of the mothers inour study attributed their child’s behavioral symp-toms to factors beyond his or her control. It alsomay explain why we did not replicate King’s(2000) finding of a relationship between negativesymptoms and higher levels of maternal criticism.If mothers perceive their son or daughter’s behav-iors and symptoms as beyond the child’s control,we would not expect that they would be highlycritical when their son or daughter displayed be-havior problems and autism symptoms.

There was some evidence supporting the ex-pectation that the characteristics of the individualwith autism would have an influence on maternalexpressed emotion, but the influence in this di-rection was weaker. Mothers had higher levels ofoverall expressed emotion and overinvolvementwith daughters than with sons, reflecting genderedpatterns of maternal vigilance. Overinvolvementwas found to be lower when the individual withautism also had a comorbid diagnosis of mentalretardation. We speculated that this finding mayhave been the result of a higher level of serviceutilization by those who had a comorbid diag-nosis of mental retardation because higher levelsof service utilization make it possible for mothersto share responsibility with the formal service sys-tem and not shoulder the caregiving burden




alone. In a follow-up analysis, we compared sam-ple members who had a comorbid diagnosis ofmental retardation with those who did not andfound that the former group received significantlymore services than did the latter group, as ex-pected (means at Time 3 5 7.68 vs. 4.01), t(1, 147)5 6.70, p , .001. One possible implication of thisfinding is that providing a rich array of servicesto individuals with autism may have the indirecteffect of reducing maternal emotional overinvol-vement.

Finally, mothers tended to display declininglevels of overall expressed emotion when the sonor daughter had higher levels of internalizing mal-adaptive behavior, a category that includes beingwithdrawn, inattentive, and self-injurious. Perhapsthis reflects a pattern of mutual or reciprocal emo-tional disengagement, a finding that warrants ad-ditional research given the high prevalence of de-pression among individuals with autism (Ghaziud-din, Ghaziuddin, & Greden, 2002; Howlin, 2002).

One of the unsettled issues in the field iswhether expressed emotion represents an under-lying personality trait or a transient state. In astudy comparing relatives who were high versuslow on expressed emotion, Hooley and Hiller(2000) found that those classified as high in ex-pressed emotion were less flexible, less tolerant,and had lower levels of empathy and achieve-ment, as measured by the California PsychologicalInventory, than those classified as low in ex-pressed emotion. Hooley (1998a) also found thatrelatives high in expressed emotion were morelikely to have an internally based locus of controlcompared to relatives low in expressed emotion.Our finding of a moderate level of stability ob-served for expressed emotion over the 18-monthperiod along with weak cross-lagged effects of be-havior problems and autism symptoms on ex-pressed emotion are consistent with conceptuali-zations of expressed emotion as a personality trait.

However, other researchers have found evi-dence that expressed emotion is reactive to char-acteristics of the adult with disabilities, in partic-ular fluctuations in behavior problems and, thus,expressed emotion also may have state-like quali-ties (King, 2000). Still others reported evidencethat expressed emotion has characteristics of botha state and trait (Schreiber, Breier, & Pickar, 1995).In our study, even though expressed emotion re-mained quite stable over time, approximately 30%of the sample shifted from low to high expressedemotion or vice versa between Times 2 and 3. We

agree with Hooley (1998b) and others (e.g., Ro-senfarb et al., 1995) that attention should be fo-cused on understanding the mechanisms under-lying the relationship between expressed emotionand symptom exacerbations rather than continu-ing the trait versus state debate.

In research on families of individuals withschizophrenia, expressed emotion has been shownto be modifiable through psychoeducational in-terventions. For example, Leff and Vaughn (1985)reported that a 9-month psychoeducational groupfor relatives of adults with schizophrenia had a54% success rate in reducing expressed emotionlevel from high to low. Tarrier et al. (1988) foundthat psychoeducational behavior-oriented treat-ment significantly reduced levels of criticism andoverinvolvement. Based on the evidence of thepresent investigation, there is reason to developand evaluate the effects of similar interventionsfor families of individuals with autism, particularlyduring the early childhood years when family in-teraction patterns may be more fluid, and targetedto families who score high on expressed emotion.Family psychoeducational interventions are likelyalso important for families of adolescents andadults with autism, as our data suggested thatthere are continued significant detrimental effectsof high levels of expressed emotion, even in adult-hood. Because our data are correlational and donot permit strong inferences regarding causality,intervention studies are a necessary first step todetermine whether family treatments designed tolower caregiver expressed emotion in families ofpersons with psychiatric disabilities have a similarbeneficial effect on adolescents and adults withautism.

Several limitations of the study must be ac-knowledged. First, our sample of mothers of ad-olescents and adults with autism consisted of vol-unteers who were predominantly White. Conse-quently, the representativeness of the sample andthe generalizability of the findings are not known.Second, the 18-month time span analyzed in ourdata represents a very small slice in the life of theinteraction between mothers and their adolescentor adult child with autism. The correlations be-tween parental characteristics and child behaviorsreflect bidirectional transactional processes thatunfold over many years. For example, Bates, Pet-tit, and Dodge (1995) found that an infant’s earlycharacteristics elicited harsh parenting at age 4,which in turn predicted externalizing problemswhen children were young adolescents. Thus, a




longer time period may be needed to detect theinfluence of the son or daughter’s behavior prob-lems or symptoms on the mother. Third, the mea-sures of maladaptive behavior problems, and au-tism symptoms were based on data collectionfrom the mothers. This likely increased the asso-ciations among these variables over time, makingit more difficult to detect the effect of behaviorproblems and autism symptoms on expressedemotion. Fourth, as noted above, overall levels ofexpressed emotion were low relative to other stud-ies, with very few mothers in our sample classifiedas high on emotional overinvolvement. The re-striction in range on the emotional overinvolve-ment measure at both Time 2 and Time 3 maypartially explain why we were not able to replicatelongitudinally our earlier cross-sectional finding ofa relationship between impairments in reciprocalsocial interactions and emotional overinvolve-ment, based on a larger sample drawn from thesame study (Orsmond et al., 2006). In addition,although Leeb et al. (1991) found considerablestability in ratings based on the Five-MinuteSpeech Sample on two occasions 5 weeks apart,the evidence for the long-term stability of Five-Minute Speech Sample ratings is mixed. King(2000) reported considerable stability in thesespeech sample ratings over an 18-month period,with stability coefficients for criticism higher thanthose for emotional overinvolvement (rangingfrom .69 to .71 vs. 53 to .60, respectively). Perisand Baker (2000) found good stability over a 2-year period on the overall rating of expressedemotion but lower levels of stability for criticismand emotional overinvolvement. Clearly, more re-search needs to be conducted on the stability ofFive-Minute Speech Sample ratings in longitudi-nal research.

Past research on families of individuals withautism has confirmed that such families experi-ence greater levels of stress than do families whosechildren have other types of developmental dis-abilities (Seltzer et al., 2000). Early interventionfor young children with autism has been shownto be an effective strategy for improving skills andreducing the level of impairments (Lord & Mc-Gee, 2001). Results of the present study point toanother potentially effective avenue for improvinglife-course outcomes for this population, becausethe sons and daughters of mothers who do notexpress high levels of criticism, and to a lesser ex-tent, who are not overly vigilant, tend to becomeless impaired over time. Although there is increas-

ing evidence that autism is a complex genetic dis-order, the effect of the family environment inshaping the behavioral phenotype should not beunderestimated.

References

American Psychiatric Association. (2000). Diag-nostic and statistical manual of mental disorders(4th ed., text rev.). Washington, DC: Author.

Asarnow, J. R., Thompson, M., Woo, S., & Cant-well, D. P. (2001). Is expressed emotion a spe-cific risk factor for depression or a nonspecificcorrelate of psychopathology? Journal of Ab-normal Child Psychology, 29, 573–583.

Barrowclough, C., & Parle, M. (1997). Appraisal,psychological adjustment and expressed emo-tion in relatives of patients suffering fromschizophrenia. British Journal of Psychiatry,171, 26–30.

Bates, J., Pettit, G., & Dodge, K. (1995). Familyand child factors in stability and change inchildren’s aggressiveness in elementaryschool. In J. McCord (Ed.), Coercion and pun-ishment in long-term perspectives (pp. 124–138).New York: Cambridge University Press.

Beck, A., Daley, D., & Hastings, R. P. (2004).Mothers’ expressed emotion towards childrenwith and without intellectual disabilities. Jour-nal of Intellectual Disability Research, 48, 628–638.

Bledin, K., Kuipers, L., MacCarthy., B., & Woods,R. (1990). Daughters of people with demen-tia: Expressed emotion, strain and coping.British Journal of Psychiatry, 157, 221–227.

Brown, G. W., Birley, J. L. T., & Wing, J. K.(1972). Influence of family life on the courseof schizophrenic disorders: A replication.British Journal of Psychiatry, 121, 241–258.

Bruininks, R. H., Woodcock, R. W., Weatherman,R. F., & Hill, B. K. (1996). Scales of IndependentBehavior–Revised. Itasca, IL: Riverside.

Butzlaff, R. L., & Hooley, J. M. (1998). Expressedemotion and psychiatric relapse: A meta anal-ysis. Archives of General Psychiatry, 55, 547–552.

Dossetor, D. R., Nicol, A. R., & Stretch, D. D.(1994). A study of expressed emotion in theparental primary carers of adolescents with in-tellectual impairment. Journal of IntellectualDisability Research, 38, 487–499.

Essex, E. L., & Hong, J. (2005). Older caregivingparents: Division of household labor, marital




satisfaction, and caregiver burden. Family Re-lations, 54, 448–460.

Finkel, S. E. (1995). Causal analysis with panel data(Sage University Paper series on QuantitativeApplications in the Social Sciences, 07–105).Thousand Oaks, CA: Sage.

Folstein, S. E., & Rosen-Sheidley, B. (2001). Ge-netics of autism: Complex aetiology for a het-erogeneous disorder. Nature Reviews Genetics,2, 943–955.

Fromm-Reichmann, F. (1948). Notes on the de-velopment of treatment of schizophrenics bypsychoanalytic psychotherapy. Psychiatry, 11,263–274.

Ge, X., Best, K. M., Conger, R. D., & Simons, R.L. (1996). Parenting behaviors and the occur-rence and co-occurrence of adolescent depres-sive symptoms and conduct problems. Devel-opmental Psychology, 32, 717–731.

Ghaziuddin, M., Ghaziuddin, N., & Greden, J.(2002). Depression in persons with autism:Implications for research and clinical care.Journal of Autism and Developmental Disorders,32, 299–306.

Glutting, J., Adams, W., & Sheslow, D. (2000).Wide Range Intelligence Test. Wilmington, DE:Wide Range.

Goldstein, M. J., Rosenfarb, I., Woo, S., & Nuech-terlein, K. (1994). Intrafamilial relationshipsand the course of schizophrenia. Acta Psychia-trica Scandinavica, 90 (Suppl.), 60–66.

Greedharry, D. (1987). Expressed emotion in fami-lies of the mentally handicapped: A pilot study.British Journal of Psychiatry, 150, 400–402.

Greenberg, J. S., Seltzer, M. M., Krauss, M. W.,Chou, R., & Hong, J. (2004). The effect ofquality of the relationship between mothersand adult children with schizophrenia, au-tism, or Down syndrome on maternal well-being: The mediating role of optimism. Amer-ican Journal of Orthopsychiatry, 74, 14–25.

Greenley, J. R. (1986). Social control and ex-pressed emotion. Journal of Nervous and Men-tal Disease, 174, 24–30.

Hastings, R. P., Daley, D., Burns, C., & Beck, A.(2006). Maternal distress and expressed emo-tion: Cross sectional and longitudinal rela-tionships with behavior problems of childrenwith intellectual disabilities. American Journalof Mental Retardation, 111, 48–61.

Hill, A., Bolte, S., & Petrova, G. (2001). Stabilityand interpersonal agreement of the interview-

based diagnosis of autism. Psychopathology, 34,187–191.

Hirshfeld, D. R., Biederman, J., Brody, L., Fara-one, S. V., & Rosenbaum, J. F. (1997). Asso-ciations between expressed emotion and childbehavioral inhibition and psychopathology:A pilot study. Journal of the American Academyof Child and Adolescent Psychiatry, 36, 205–213.

Hooley, J. M. (1987). The nature and origins ofexpressed emotion. In M. J. Goldstein & K.Hahlweg (Eds.), Understanding major mentaldisorder: The contribution of family interaction re-search (pp. 176–194). New York: Family Pro-cess Press.

Hooley, J. M. (1998a). Expressed emotion and lo-cus of control. Journal of Nervous and MentalDisease, 186, 374–378.

Hooley, J. M. (1998b). Expressed emotion and psy-chiatric illness: From empirical data to clinicalpractice. Behavior Therapy, 29, 631–646.

Hooley, J. M., & Gotlib, I. H. (2000). A diathesis-stress conceptualization of expressed emotionand clinical outcome. Applied and PreventativePsychology, 9, 135–151.

Hooley, J. M., & Hiller, J. B. (2000). Personalityand expressed emotion. Journal of AbnormalPsychology, 109, 40–44.

Howlin, P. (2002). Autistic disorders. In P. Howlin& O. Udwin (Eds.), Outcome in neurodevelop-mental and genetic disorders (pp. 136–168).Cambridge, UK: Cambridge University Press.

Kim, E. Y., & Miklowitz, D. J. (2004). Expressedemotion as a predictor of outcome among bi-polar patients undergoing family therapy.Journal of Affective Disorders, 82, 343–352.

King, S. (2000). Is expressed emotion cause or ef-fect in the mothers of schizophrenic youngadults. Schizophrenia Research, 45, 65–78.

King, S., & Dixon, M. J. (1995). Expressed emo-tion, family dynamics and symptom severityin a predictive model of social adjustment forschizophrenic young adults. Schizophrenia Re-search, 14, 121–132.

Lam, D., Giles, A., & Lavander, A. (2003). Carers’expressed emotion, appraisal of behaviorproblems and stress in children attendingschools for learning disabilities. Journal of In-tellectual Disability Research, 27, 456–463.

Leeb, B., Hahlweg, K., Goldstein, M. J., Feinstein,E., Mueller, U., Dose, M., & Magano-Amato,A. (1991). Cross-national reliability, concur-rent validity, and stability of a brief method




for assessing expressed emotion. Psychiatry Re-search, 39, 25–31.

Leff, J., & Vaughn, C. (1985). Expressed emotion infamilies: Its significance for mental illness. NewYork: Guilford Press.

Lord, C., & McGee, W. (2001). Educating childrenwith autism. Committee on Educational Inter-ventions for Children With Autism. Divisionof Behavioral and Social Sciences and Educa-tion. Washington, DC: National AcademyPress.

Lord, C., Pickles, A., & McLennan, J. (1997). Di-agnosing autism: Analyses of data from theAutism Diagnostic Interview. Journal of Au-tism and Developmental Disorders, 27, 501–517.

Lord, C., Rutter, M., & Le Couteur, A. (1994).Autism Diagnostic Interview–Revised: A re-vised version of a diagnostic interview forcaregivers of individuals with possible perva-sive developmental disorders. Journal of Au-tism and Developmental Disorders, 24, 659–685.

Luckasson, R., Borthwick-Duffy, S., Buntinx, W.H. E., Coulter, D. L., Craig, E. M., Reeve, A.,Schalock, R. L., Snell, M. E., Spitalinik, D.M., Spreat, S., & Tasse, M. J. (2002). Mentalretardation: Definition, classification, and systemsof supports (10th ed.). Washington, DC: Amer-ican Association on Mental Retardation.

Magana, A. B., Goldstein, M. J., Karno, M., Mik-lowitz, D. J., Jenkins, J., & Falloon, I. R. H.(1986). A brief method for assessing expressedemotion in families of psychiatric patients.Psychiatry Research, 17, 203–212.

Mink, I. T., & Nihira, K. (1987). Direction of ef-fects: Family life styles and behavior of TMRchildren. American Journal on Mental Retarda-tion, 92, 57–64.

Moore, E., & Kuipers, E. (1999). The measure-ment of expressed emotion in relationshipsbetween staff and service users: The use ofshort term speech samples. British Journal ofClinical Psychology, 38, 345–356.

Olson, D. H., McCubbin, H. I., Barnes, H., Lar-sen, A., Muxen, M., & Wilson, M. (1983).Families: What makes them work. Beverly Hills,CA: Sage.

Orsmond, G. I., Krauss, M. W., & Seltzer, M. M.(2004). Peer relationships and social and rec-reational activities among adolescents andadults with autism. Journal of Autism and De-velopmental Disorders, 34, 245–256.

Orsmond, G. I., Seltzer, M. M., Greenberg, J. S.,& Krauss, M. W. (2006). Mother–child rela-

tionship quality among adolescents andadults with autism. American Journal of MentalRetardation, 111, 121–137.

Orsmond, G. I., Seltzer, M. M., Krauss, M. W.,& Hong, J. (2003). Behavior problems inadults with mental retardation and maternalwell-being: Examination of the direction ofeffects. American Journal of Mental Retardation,108, 257–271.

Patterson, G. R. (1982). Coercive family process. Eu-gene, OR: Castalia.

Patterson, P., Birchwood, M., & Cochrane, R.(2000). Preventing the entrenchment of highexpressed emotion in first episode psychosis:Early developmental attachment pathways.Australian and New Zealand Journal of Psychi-atry, 34 (Suppl.), S191–S197.

Peris, T. S., & Baker, B. L. (2000). Applications ofthe expressed emotion construct to youngchildren with externalizing behavior: Stabilityand prediction over time. Journal of Child Psy-chology and Psychiatry and Allied Disciplines, 41,457–462.

Peris, T. S., & Hinshaw, S. P. (2003). Family dy-namics and preadolescent girls with ADHD:The relationship between expressed emotion,ADHD symptomatology, and comorbid dis-ruptive behavior. Journal of Child Psychologyand Psychiatry and Allied Disciplines, 44, 1177–1190.

Prizant, B. M., Wetherby, A. M., & Rydell, P. J.(2000). Communication intervention issuesfor young children with autism spectrum dis-orders. In A. W. Wetherby & B. M. Prizant(Eds.), Autism spectrum disorders: A transaction-al developmental perspective (pp. 193–224). Bal-timore: Brookes.

Rosenfarb, I. S., Goldstein, M. J., Mintz, J., &Nuechterlien, K. H. (1995). Expressed emo-tion and subclinical psychopathology observ-able within the transaction between schizo-phrenic patients and their family members.Journal of Abnormal Psychology, 104, 259–267.

Rossi, A. S., & Rossi, P. H. (1990). Of human bond-ing: Parent–child relations across the life course.Hawthorne, NY: Adine de Gruyter.

Rutter, M. (2005). Aetiology of autism: Findingsand questions. Journal of Intellectual DisabilityResearch, 49, 231–238.

Scazufca, M., & Kuipers, E. (1996). Links betweenexpressed emotion and burden of care in rel-atives of patients with schizophrenia. BritishJournal of Psychiatry, 168, 580–587.




Schreiber, J. L., Breier, A., & Pickar, D. (1995).Expressed emotion: Trait or state? British Jour-nal of Psychiatry, 166, 647–649.

Seltzer, M. M., & Krauss, M. W. (2001). Qualityof life of adults with mental retardation/de-velopmental disabilities who live with family.Mental Retardation and Developmental Disabili-ties Research Reviews, 7, 105–114.

Seltzer, M. M., Krauss, M. W., Orsmond, G. I.,& Vestal, C. (2000). Families of adolescentsand adults with autism: Uncharted territory.In L. M. Glidden (Ed.), International review ofresearch on mental retardation (pp. 267–294).San Diego: Academic Press.

Seltzer, M. M., Krauss, M. W., Shattuck, P. T.,Orsmond, G., Swe, A., & Lord, C. (2003). Thesymptoms of autism spectrum disorders in ad-olescence and adulthood. Journal of Autismand Developmental Disorders, 33, 565–581.

Shattuck, P., Seltzer, M. M., Greenberg, J. S.,Lounds, J., Kring, S., Bolt, D., & Lord, C.(2006). Changes in autism symptoms andmaladaptive behaviors in adolescents andadults with an autism spectrum disorder. Jour-nal of Autism and Developmental Disorders.Manuscript submitted for publication.

Sparrow, S. S., Carter, A. S., & Cicchetti, D. V.(1993). Vineland Screener: Overview, reliability,validity, administration, and scoring. New Ha-ven, CT: Yale University Child Study Center.

Stubbe, D. E., Zahner, G. E. P., Goldstein, M. J.,& Leckman, J. F. (1993). Diagnostic specificityof a brief measure of expressed emotion: Acommunity study of children. Journal of Psy-chology and Psychiatry and Allied Disciplines, 34,139–154.

Tarrier, N., Barrowclough, C., Vaughn, C., Bamrah,J. S., Porceddu, K., Watts, S., & Freeman, H. L.(1988). The community management of schizo-phrenia: A controlled trial of a behavioral in-tervention with families to reduce relapse. Brit-ish Journal of Psychiatry, 153, 532–542.

Tarrier, N., Sommerfield, C., & Pilgrim, H. (1999).Relatives’ expressed emotion (EE) and PTSDtreatment outcome. Psychological Medicine, 29,801–811.

Van-Humbeeck, G., Van-Audenhove, C., De-Hert,M., Pieters, G., & Storms, G. (2002). Expressedemotion: A review of assessment instruments.Clinical Psychology Review, 22, 321–341.

Vitaliano, P. P., Young, H. M., Russo, J., Romano,J., & Magana-Amato, A. (1993). Does ex-pressed emotion in spouses predict subse-quent problems among care recipients withAlzheimer’s disease? Journals of Gerontology,48, 202–209.

Wearden, A. J., Tarrier, N., Barrowclough, C.,Zastowny, T. R., & Rahill, A. A. (2000). Areview of expressed emotion in health care.Clinical Psychology Review, 20, 633–666.

Woo, S. M., Goldstein, M. J., & Nuechterlein, K.H. (2004). Relatives’ affective style and the ex-pression of subclinical psychopathology inpatients with schizophrenia. Family Process,43, 233–247.

World Health Organization. (1990). The ICD–10classification of mental and behavioral disorders:Clinical descriptions and diagnostic guidelines.Geneva: Author.

Received 4/15/03, accepted 2/7/06.Editor-in-charge: William E. MacLean, Jr.

Support for the preparation of this paper was pro-vided by grants from the National Institute onAging (R01 AG08768), National Institute of Men-tal Health NIMH (R01 MH 55928), National In-stitute of Child Health and Human Development(P30 HD03352), all through the Waisman Centerat the University of Wisconsin-Madison; NationalInstitute for Disability and Rehabilitation Re-search (H133B031134) through the RehabilitationResearch and Training Center on Aging and De-velopmental Disabilities at the University of Illi-nois at Chicago; and NIMH (P30 MH068579)through the Center for Mental Health ServicesResearch at the George Warren Brown School ofSocial Work, Washington University in St. Louis.Requests for reprints should be sent to Jan S.Greenberg, University of Wisconsin-Madison,Waisman Center, Madison, WI 53705. E-mail:[email protected]

bidirectional effects of expressed emotion and … · bidirectional effects of expressed emotion...

Documents