care neglect, supervisory neglect, and harsh

Upload: sara-billai

Post on 10-Apr-2018

221 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/8/2019 Care Neglect, Supervisory Neglect, And Harsh

    1/16

    10.1177/1077559504273684CHILDMALTREATMENT/MAY 2005Knutson etal. /NEGLECTANDHARSHPARENTINGAREPLICATIONKnutson etal. /NEGLECTANDHARSHPARENTINGAREPLICATION

    Care Neglect, Supervisory Neglect, and HarshParenting in the Development of Childrens

    Aggression: A Replication and Extension

    John F. Knutson

    University of Iowa

    David DeGarmo

    Oregon Social Learning Center

    Gina Koeppl

    University of Iowa

    John B. Reid

    Oregon Social Learning Center

    To understand the effects of neglectful parenting, poor super-vision, and punitive parenting in the development of chil-drens aggression, 218 children ages 4 to 8 years who weredisadvantaged and their mothers were recruited from twostates to develop a sample that was diverse with respect to de-gree of urbanization and ethnicity. Multimethod and

    multisourceindices of the predictive constructs (Social Disad-vantage, Denial of Care Neglect, Supervisory Neglect, andPunitive Discipline)and the criterion construct (Aggression)were used in a test of a theoretical model using structuralequation modeling. The results established the role of care ne-glect, supervisory neglect, and punitive parenting as media-tors of the role of social disadvantage in the development ofchildrens aggression, the importance of distinguishing be-tween two subtypes ofneglect, and the need to consider the roleof discipline in concert with neglect when attempting tounderstand theparenting in thedevelopment ofaggression.

    Keywords: childrens aggression; care neglect; supervisoryneglect; punitive discipline

    Although there is littlequestion that child maltreat-ment can occasion an immediate adverse effect on

    the health and well-being of children, possible distalconsequences of maltreatment, such as problems inthe childs social adjustment, have become an impor-tant focus of research. Because there is unequivocalevidence from three National Incidence Studies(NIS; Office of Human Development Services

    [OHDS], 1981, 1988; Sedlak& Broadhurst,1996) andthe National Child Abuse and Neglect Data System(NCANDS; U.S. Department of Health and HumanServices, Administration on Children, Youth, and

    92

    CHILD MALTREATMENT, Vol. 10, No. 2, May 2005 92-107DOI: 10.1177/1077559504273684 2005 Sage Publications

    Authors Note:Supported, inpart, by funding by National Instituteof MentalHealth(NIMH)and Administration on Children,Youth,and Families (ACYF; Research Grant MH61731), John F. Knutson,principal investigator; in part by funding by National Institute onChild Health and Development (Research Grant HD 42115), Da-vid S.DeGarmo,principalinvestigator; andfunding by NationalIn-stitute on Drug Abuse and NIMH (Grants P20 DA 017592 and P3046690), respectively, JohnB. Reid,principalinvestigator. The facili-

    tation of the research by Paul Spencer (Oneida County Depart-ment of Social Services), Barry Bennett, Cheryl Whitney, MarcBatey, Mark Schmidt, and WayneMcCracken(Iowa Department ofHuman Services) is gratefully acknowledged. The assistance of Angela Anderson, Beth Boyer, Minka Dawson, Becky Fetrow,Deanna Heckenberg, Martha Early, Aubra Hoffman, EstherHoffman, Kathy Jordan, Robert Latzman, Theresa Mayne, MaryMcCarren, Donna Palmer, Nicole Shay, and Nizete-ly Valles isgreatly appreciated. Correspondenceregarding this article shouldbe sent to John F. Knutson, Department of Psychology, E-11 Sea-shore Hall, The University of Iowa, Iowa City, IA 52240; e-mail:[email protected]

  • 8/8/2019 Care Neglect, Supervisory Neglect, And Harsh

    2/16

  • 8/8/2019 Care Neglect, Supervisory Neglect, And Harsh

    3/16

    cial behavior at enrollment but significantlypredicted increases in antisocial behavior on follow-up 5 years later. From the standpoint of the theoreti-cal model that was advanced, supervisory neglect wasnotshown tobe importantin thedevelopment ofanti-social behavior in the younger cohort. For the older

    cohort, whose families were enrolled when the chil-dren were in fifthgrade a somewhatdifferent, butthe-oretically consistent, pattern emerged. Although thecare neglect construct was associated with higher lev-els of initial antisocial behaviors and marginallyassociated with increases in antisocial behavior overtime in the older cohort, the most notable differencebetween the two cohorts was the finding that thesupervisory neglect construct was related to careneglect and punitive discipline in predicting childantisocial behavior. A second notable differencebetween the cohorts was the significant link betweencare neglect and punitive discipline. Finally, and

    importantly, for the older cohort, punitive disciplineat enrollment was important in predicting antisocialstatusat enrollment, but less important as a predictorof change in antisocial behavior 5 years later whencompared to the test of the model with the youngercohort.When themodel testswere consideredin con-cert, two important findings emerged. First, theimportance of punitive discipline in predicting anti-social behavior was apparent, a finding that is consis-tent with the work by Jaffee et al. (2004). Second,there was strong evidence of the importance of careneglect in the development of childrens aggression

    and some evidence of the possible importance ofsupervisory neglect in the development of childrensaggression.

    Although the Knutson et al. (2004) study providedsome empirical support for a theoretical model inwhich three components of deficient parenting (careneglect, punitive discipline, and supervisory neglect)contributed uniquely to the development of chil-drens aggression and antisocial behavior, the factthat supervision only played a role in the older cohortcompromised reaching unequivocal conclusionsabout the viability of the model. A number of limita-tions of the Knutson et al. (2004) study, however, sug-

    gested that a systematic replication would be useful.First, although the sample used in that research wasdrawn from a community with elevated risk for childantisocial behavior (high juvenile arrest rates andcommunity indices of delinquency), the recruitedsample represented a relatively broad range of socio-economic strata and a relatively low base rate of fami-lies who had been identified by child protective ser-vices(CPS) agencies asbeing eitherphysically abusiveor neglectful. Thus, the sample did not reflect high

    levels of neglectful parenting nor high levels ofseverely punitive discipline. Second, and perhapsmore important, the study was imbedded in a univer-sal prevention program that was not designed to fully

    assess abusive and neglecting parenting. Thus, thesupervisory neglect construct and the care neglectconstruct were developed using proxy variables andderivative approximations of the target indicators. Inparticular, although the supervisory neglect constructwas developed using indicators that worked well foran older cohort developing into preadolescence, theindicators might nothave been suitable for tapingtheessential features of supervision in the younger years.Thus, it seems probable thatthe Knutson et al. (2004)model would fit data in which a more disadvantagedsample was tested and better indices of care neglect,supervisory neglect, and harsh discipline were avail-

    able. Thus, the current research is designed to repli-cate the theoretical model that specifies a causal rolefor care neglect, supervisory neglect, and harsh puni-tive discipline in the development of childrens anti-social behavior advanced by Knutson et al. (2004) byusing a more neglecting, abusive, and disadvantagedsample of families and stronger multimethod andmultisource indicators of the core constructs. Basedon the model advanced by Knutson et al. (2004) andtheir findings, a figure depicting the hypothesizedpaths in a structural model are shown in Figure 1.

    METHOD

    Participants

    Two hundred and eighteen mothers and theirchild who had been consecutively recruited atresearch sites in southeastern Iowa and north centralWisconsin during the first 28-month period of anongoing study of neglectful parenting and childrensaggression participated in the current study.Although some single-parent fathers and fathers in

    CHILD MALTREATMENT / MAY 2005

    94 Knutson et al. / NEGLECT AND HARSH PARENTING A REPLICATION

    ChildAggression

    SupervisoryNeglect

    PunitiveDiscipline

    SocialDisadvantage

    CareNeglect

    (Mediated Influence)

    FIGURE 1: Hypothesized Relationships for Parent DeficiencyModel of Supervisory and Care Neglect

  • 8/8/2019 Care Neglect, Supervisory Neglect, And Harsh

    4/16

    two-parent households had been recruited, the num-ber of fathers was too small to include in the analysesfor this article. Each recruited mother-child dyad hadto meet a number of eligibility criteria. First, the fam-ily had tohavea child between ages4 and 8 years resid-ing in the home at the time of recruitment. Second,

    the family had to have been a recipient of social oreconomic services at thetime of recruitment.Servicescould have been occasioned by the family being iden-tified as physically abusive or neglecting, or becauseof economic disadvantage. Families with childrenknown to have been sexually abused were notincluded in thepool of potentialparticipants. In addi-tion, because the focus of the research was on parent-child relational function, children who had been inan out-of-home placement (i.e., foster care) or wererecipients of intensive in-home services were notidentified as eligible for participation.

    The recruitment area of Iowa included two coun-

    ties that encompassed two metropolitan areas classi-fied as small urban and suburban, as well as rural agri-cultural areas and several small towns. Therecruitment area of Wisconsin included two ruralcounties characterized by several small towns, ruraldevelopment, as well as isolated rural living. In total,the sample represented a broad range of degree ofurbanization; however, l iv ing arrangementsprototypical of inner-cityenvironments (i.e., high-riseapartments) were not represented. Many of therecruited families, however, had recently relocatedfrom large urban areas of the Midwest (e.g., Chicago,

    Illinois; Detroit, Michigan; Gary-Hammond, Indiana;Milwaukee, Wisconsin), andmanyof thechildrenhadconsiderable exposure to inner-city life. Thechildrenwere slightly more racially diverse than the mothers;64% of the mothers identified themselves as White,23% African American, 5% Latino/a, 2% NativeAmerican, 1% Asian, and 4% as multiracial. Childrenwere identified by their mothers as 56% White, 23%African American, 5% Latino/a, 3% Native Ameri-can, 1% Asian, and 12% as multiracial.1

    At both of the recruitment sites, sentinel agencieswere used to identify potential research participants.Because of the different structure of the provision of

    social and economic services in the two states, slightlydifferent procedures were followed in Iowa and Wis-consin. Basically, the appropriate state or countyagency produced a list of potential research partici-pants at approximately quarterly intervals. Thoseagencies identified families who had recently beenidentified as having had a substantiated physicalabuse or neglect allegation, or who were receivingstate or county support (i.e., Temporary Assistancefor Needy Families (TANF); Title XIX). Families in

    this target population are notoriously difficult to con-tact and enroll in research. Focus groups withmembers of the target population were conducted inthe community to identify factors that could compro-mise or facilitate recruitment. Frequent addresschanges, phone disconnection or number changes,

    useof cell phones, unlisted numbers, callerID,aswellas an acknowledged reluctance to open mail thatseems official or is from an unknown source werenoted by focus group participants. As a result, therecruitment strategy that was adopted reflected someof the recommendations of those focus groups.

    Eligible families from the lists first received a letterfrom an administrator of the social service agencywho was not in a service delivery role. The letterinformed potential participants of their eligibility toparticipate in research on parenting and childrenssocial development for which they would be compen-sated at a rate of $50 per session, plus out-of-pocket

    (i.e., transportation and babysitting) expenses. Par-ents who did not respond to the initial letter receiveda follow-up letter and, later, telephone contact by aclerk employed by the project. Inthe letters,and inallof the direct contacts, potential participants wereinformed thatparticipationwas voluntary,and enroll-ment or nonenrollment in the project would have noinfluence on their receipt of any services from rele-vant state or county agencies. When the eligible par-ent indicated an interest in participation, an initialappointment was scheduled in his or her home; dur-ing that initial session the research protocol was

    explained to the participant, and informed consentwas obtained. If the parent chose to participate in theresearch, an initial structured interview was con-ducted in the home immediately following theinformed consent procedure. For families in whichmore than one child was eligible to participate, asingle child was randomly selected to be the researchparticipant.

    Of the families on the eligibility lists, approxi-mately 50% were known to have been actually con-tactedby mail orphone.Of those,approximately 55%scheduled an in-person, in-home appointment tolearn about the project. Of those who scheduled an

    in-home interview and participated in the informedconsent process, all but one agreed to participate inthe project. The mother who declined to enroll wasthe motherof twins,and she wouldnot agree toenrollif only one of the children could participate in thestudy. A few potential participants repeatedly sched-uled in-home sessions but were not home when theresearch assistantarrived at the scheduled time. Afterone third missed scheduled appointment, no furtherefforts to recruit were made. Participant recruitment

    CHILD MALTREATMENT / MAY 2005

    Knutson et al. / NEGLECT AND HARSH PARENTING A REPLICATION 95

  • 8/8/2019 Care Neglect, Supervisory Neglect, And Harsh

    5/16

    and all of the procedures were conducted under theaegis of The University of Iowa IRB-02.

    Procedures

    In developing indicators of the core constructs ofcare neglect, supervisory neglect, and harsh punitivediscipline, a number of factors were considered. Mostimportant, it was deemed appropriate that all of theindices of parenting be based on directly obtainedinformation rather than the use of administrativedata or agency records. In addition, for all of the con-structs,a multimethod and multisource approach wasadopted. With the exception of the initial structuredinterviewandin-home assessment, allof themeasureswere obtained in three laboratory sessions requiringapproximately 90 minutes each. Because of the longi-tudinal nature of the project in which the currentstudy was embedded, additional measures that are

    not the focus of this research were obtained but notdescribed in full. Only procedures and derived mea-sures relevant to the scope of the current analyses aredescribed below.

    In-home interview and environmental assessment.When a parentagreedto participate in theresearch,astructured interview designed to obtain backgroundand demographic information, as well as informationpertinent to the parenting constructs, was immedi-ately conducted by a research assistant. In most in-stances, one research assistant conducted theinterview and recorded participant responses while a

    second research assistant recorded the participantsresponses on a parallel form. This structured inter-view was based, in part, on a modification of theHome Observation for Measurement of the Environ-ment (Caldwell & Bradley, 1978) recently developedfor use in the Project on Human Developmentin Chi-cago Neighborhoods (PHDCN; Leventhal, Selner-OHagan, Brooks-Gunn, Bingenheimer, & Earls,2004) and, in part, on the information frameworkthat emerged from the recommendations of the Re-search Subcommittee of the Interagency Task Forceon Child Abuse and Neglect (see Sternberg et al.,2004). The interview also included questions regard-

    ing supervision related to injury prevention, whichwere derived from research by Peterson, Ewigman,and Kivlahan (1993) and a home safety inventory(Tymchuk, Lang, Dolyniuk, Berney-Ficklin, & Spitz,1999). Because this interview was conducted in thehome of the child, it was also possible to obtain directinformation regarding evidence of neglect manifestin the condition of the home, following the strategydeveloped by PHDCN; thatis, sleeping arrangements,cleanliness, plumbing, personal hygiene of family

    members, amountof space available to thenumberofresidents of the household, and any health or safetyhazards threatening children of the household couldbe identified by the interviewer in a direct examina-tion of thehousehold. Following theinterview, there-search assistants completed an objective checklist

    while conducting a visual survey of the physical envi-ronment in the vicinity of the participants home toidentify environmental hazards and the general con-dition of the immediate community in which thechild resides (e.g., broken glass, drug paraphernalia,houses in disrepair, boarded windows, trash). Forpar-ticipantsresiding in a municipalarea, an area withina1-block radius was examined. For rural participants,the area was generally comparable; however, the ar-rangement of rural roads and sparse populationsometimes required a greater distance to evaluate thestatus of nearby houses to complete the rating of thecommunity condition and accessible hazards.

    Laboratory Assessments

    Within a median of 26 days of the home visit, thefirst of a series of laboratory sessions was completed.Subsequent sessions were arranged as soon as possi-ble; however, scheduling difficulties and missedappointments occasionally resulted in protocol com-pletion more than 60 days after enrollment. Duringthe first session, the parent was administered theReading and Spelling subtests of the Wide RangeAchievement Test (WRAT;Jastak Associates, 1993) toestablish literacy. For mothers who were not suffi-

    ciently literate to complete the paper-and-pencil self-report instruments of the protocol, thoseinstruments were administered in an interview format by aresearch assistant.

    Parent-child interaction task. Microsocial indices ofparent-child interaction were coded from videotapedinteractions between the mother andchild while theyparticipatedin a structured laboratory task duringthefirst session. This 45-minute parent-child laboratorytask was conducted in a room outfitted to approxi-mate a living room setting, with comfortable stuffedfurniture, an end table, a coffee table, and appropri-

    ate appointments. So coders are uninformed with re-spect to the source of the families; the laboratoryrooms at the Wisconsin laboratory and the Iowa labo-ratory are identical with respect to size, furnishings,carpeting, and other appointments. The task beginswith a communication test in which the child firstplays an unfamiliar (i.e., commercially unpopular)age-appropriate board game with an experimenter.At that time, the parentcompletes a questionnaire onchild behavior issues in an adjacent room. After 5

    CHILD MALTREATMENT / MAY 2005

    96 Knutson et al. / NEGLECT AND HARSH PARENTING A REPLICATION

  • 8/8/2019 Care Neglect, Supervisory Neglect, And Harsh

    6/16

    minutes playing the game, the parent is brought tothe room and left alone with the child with the in-struction to learn from the child what had just tran-spired. The parents task is to learn as much aspossible about what the child did with the experi-menter and the details of the game. After 5 minutes,

    the child is removed from the room, and the experi-menter interviews the parent to determine how suc-cessfully she was able to obtain information from herchild.Thesecondcomponent of thelaboratory task isa social problem-solving task involving role-playingand pretend activities by the parent and child basedon instructions given to the parent (e.g., preparingfor the visit of another family with a child who is dis-liked by the participant child). The third componentinvolves theparent leading a discussion regarding themost important of the child issues identified in thequestionnaire completed while the child played thegame with the experimenter. Fourth, the parent andchild engage in free play for 10 minutes. Finally, theparent is to induce the child to straighten the roomand put the toys away. All of the parent-child interac-tions are recorded using a remote control domecamera.

    Coding of parent-child interactions. The videotapes ofthe structured parent-child interactions were codedby a professional team of four behavior coders at Ore-gon Social Learning Center (OSLC) using the Inter-personal Process Code (IPC; Rusby, Estes, & Dishion,1991). The IPC provides a real-time assessment of vir-tually all verbal and nonverbal interactions among

    family members by recording the Activity (the globalcontext or setting in which the interactions occur),Content(adescription of each verbal, nonverbal,andphysicalbehavior),and Affect(the emotionaltone ac-companying each content code) of the mother-childinteractions. TheIPCwasused to index thefrequencyof maternal aversiveand physically negative behaviors(e.g., contempt,anger, hitting with hand, hitting withan object, pinching, ear flicking, kicking, grabbing,restraining, spitting, shoving) as well as those of thechild. To enhance reliability of the coding, the familyinteraction tapeswere aggregated andcoded in a ran-

    dom order. Of the sessions, 15% were randomly se-lected for reliability checks throughout the time ofcoding. Contrasts are made between coders who hadbeen most recently established as reliable and thosefor whom the longest time had elapsed since the lastreliability assessment. This procedure permits the de-tection of the potential problem of so-called observerdrift,a phenomenon in which codershave acceptablelevels of agreement across time but are simulta-

    neously shifting steadily away from an establishedstandard (DeMaster, Reid, & Twentyman, 1976). Theeffective agreement for occurrence was .89 for Con-tent codes with a kappa coefficient of .80. Theeffective agreement score for Affect codes was .93,with a kappa of .76. The Activity code agreement was

    .93.

    Confidentiality and Child Safety

    The entire project was conducted under a Certifi-cate of Confidentiality issued by the office of theNational Institutes of Mental Health. The Certificatewas obtained with the provision in the informed con-sent documents that the investigators would report torelevant CPS agencies any circumstances that weredeemed to occasion risk to the child. Such circum-stances could be identified from child interviews, par-ent interviews, or direct observations. It should benoted that the submission of such reports to CPSagencies did not result in any loss of participants inthe protocol described in this article.

    OPERATIONAL INDICATORS OF

    THEORETICAL CONSTRUCTS

    Social Status

    Four indicators constituted the social status con-struct. Two scales were from the Home EnvironmentQuestionnaire (HEQ; Laing & Sines 1982; Sines,Clarke, & Lauer, 1984). The HEQ is a true-false ques-tionnairethat uses objectively phrased items to obtain

    information about the childs environment from thechilds parent.Developed usingtherational-statisticalapproach (Loevinger, 1956), the HEQ has eightempirically derived scales measuring specific dimen-sions of a childs psychosocial environment,with eachof the scales assessing the presence of environmentalattributes and events that are theoretically importantfor the expression of deviant and nondeviant childbehaviors (Murray & Sines, 1996). For the purpose ofthe current research, two scales were selected as indi-cators of the Social Status construct. The Achieve-ment scale measures family conditions that model or

    provide support for achievement on the part of thechild, and the Socioeconomic Status scale comprisesitems that refer to activities and attitudes related toacademic and intellectual pursuits, as well as partici-pation in community affairs. Two additional indica-tors were mothers education andoccupation. Educa-tion ranged from 1 (never reached high school) to 8(graduate or professional degree). Occupation rangedfrom 1 (unskilled laborer) to 7 (professional). Students,

    CHILD MALTREATMENT / MAY 2005

    Knutson et al. / NEGLECT AND HARSH PARENTING A REPLICATION 97

  • 8/8/2019 Care Neglect, Supervisory Neglect, And Harsh

    7/16

    homemakers, and unemployed working status wererecoded to 1 to reflect economic contribution to thesocial status indicator of occupational status.

    Supervision and Tracking

    As noted by Dishion and McMahon (1998), aware-

    ness of child activities is a critical component ofparental monitoring. Thus, for the young children ofthe current study, it was hypothesized that parentalawareness measured by congruence between parentreport and child report would serve as the optimalindex of supervision. Two congruence or concor-dance scoreswere derived to measure effective super-vision and tracking skills of the parent. As one mea-sure, the parent and the child independentlycompleted the Childrens Experience and Excite-ment Scale (CEES: Selner, 1992; Selner & Knutson,1990) during the first laboratory session. The CEESconsists of 44 slides depicting children engaging in a

    range of activities. To minimize sex-role responding,there is one form for boys, with male actors, andanother form for girls, with female actors. In an inter-view format, child participantsare asked whether theyhave ever engaged in the depicted activity. If they hadnot engaged in the activity, they were asked whetherthey have had the opportunity to do so. Without hav-ing any knowledge of their childs response to theCEES slides, parents complete the CEES in a self-report format by indicating whether their child hasever engaged in the pictured activity. If they indicatetheir child has not engaged in the activity, they indi-

    cate whether the child ever had the opportunity to doso. Concordance between child and parent reports ofexperiences and opportunity provide an index ofsupervision by adding mother-child affirmativematched pairsand mother-child negative pairs acrossall slides. The index ranged from 12 to 39 agreements(M= 30.94, SD= 4.75).

    The second index of supervision was the corre-spondence between child report and parent reporton the Childrens Reinforcement Survey Schedule(RSS; Clement & Richard, 1976). The RSS, adminis-tered by interview, asks the children to identify thepeople with whom they spend the most time, their

    favorite foods, the toys they use most often, and theplaces they spend the most time. They are also askedto identify toys they do not have but would like, placesthey would like to spend more time, and people withwhom they would like to spend more time. The par-entcompletesa paper-and-pencil RSS form. Based onthe work of Hall (1986), the effective agreement foroccurrence statistic between the child report and theparent report isused anindex of supervision (M=.35,SD = .09).

    Harsh Punitive Discipline

    The punitive discipline construct was measuredwith five indicators from multiple sources of datainvolving direct observation, analog procedures, andself-report. The self-report measures included a scale

    of inconsistent discipline consisting of 11 items usedin Knutson et al. (2004) indicating a pattern of incon-sistent or erratic discipline. Sample items includedget angry when punishing child, how often parentdoes not follow through on punishment, threatenpunishment to get child to do something, and so on.Cronbachs alpha was .98. The second reported scalewas an index of abusive discipline resulting in injuriesto the child as a result of discipline. Ten items fromthe home interview were scored as 0 (did not occur)or 1 (has occurred) and then summed. Sample itemsinclude red marks that lasted more than 24 hoursfrom being spanked, bruises after being disciplined,

    broken bones from being disciplined, child requiredstitches after discipline, child spanked with objectother than a hand (e.g.,belt,paddle, tree branch,hairbrush, other), unusual punishment (e.g., child tiedup, lock in closet, physically restrained, push ups,etc.). The observed score ranged from 0 to 6 (M=1.56, SD= 1.18).

    The third indicator was a factor score of physicalresponses and escalation in angry discipline inresponse to Analog Parenting Task (APT) firstdescribed by Zaidi, Knutson, and Mehm (1989). The APT, consists of 28 slides, each depicting a childengaging in a developmentally appropriate or devel-opmentally inappropriate activity that could be irri-tating or concerning to a childs caretaker. Sevenscenesdepicta child engaged in destructive acts (e.g.,stepping on a calculator, tearing pages from a book),sevenscenes depict dangerous activities (e.g., loadinga revolver, hanging out the window of a moving car,sitting on the edge of a roof), and seven depict rule-violating behaviors (e.g., theft, drinking an alcoholicbeverage, smoking). The remaining scenes includeage-appropriate acts (e.g., spilling a jar of salsa, messyplay with toys). In response to each scene, the motheris asked to imagine that she is charged with the

    responsibility of caring for thedepicted child, to indi-cate her emotional reaction to the depicted child(e.g., anger, worry, annoyance, amusement), and toclassify the depicted behavior (e.g., sloppy, destruc-tive, dangerous, fine). After rating and classifying thedepicted behavior theparticipant is asked to select, ina closed-set format, the disciplinary response shewould use if she were attempting to alter the childsbehavior. Disciplinary choices included such acts asignoring, verbal reprimands, restricting of privileges,

    CHILD MALTREATMENT / MAY 2005

    98 Knutson et al. / NEGLECT AND HARSH PARENTING A REPLICATION

  • 8/8/2019 Care Neglect, Supervisory Neglect, And Harsh

    8/16

    spanking, striking other than spanking, and striking with objects. Although the more severe acts werepotentially injurious, the possible injurious conse-quences of the acts were not specified in the responsechoices. After selectinga disciplinary strategy,thepar-ticipant was asked to indicate how many times she

    would permit the child to engage in the depictedbehavior before changing her disciplinary response.If theparticipant indicatedthat shewould changeherdisciplinary tactic, she was asked to indicate what thatnext disciplinary alternative would be.

    There are three primary dependent measuresobtained from the APT. The first was the frequencywith which the participants initial disciplinary choiceinvolved the use of physical discipline. The seconddependent measure was the use of escalated disci-pline as described by Knutson and Bower (1994).Escalated discipline is a circumstance where the par-ticipant shiftsfroma nonphysical form of discipline to

    physical discipline if thedepicted child were to persistin the displayed behavior. Escalated discipline canalso occur when the participant shifts from minorphysical discipline(e.g., spanking) to potentially inju-rious discipline (e.g., striking with an object) within ascene. The computed analog discipline indicator wasa factor score of total physical responses ranging from0 to 15 (M = .48, SD = 1.46) and total escalatedresponses ranging from 0 to 13 (M= 1.42, SD= 2.70),because of the skewed distributions before combin-ing, each score was log transformed (plus a constantof 0.5 for zero values).

    Based on the work of Greenwald, Bank, Reid, andKnutson (1997), a third measure from the APT con-tributed to the anger indicator in the punitive disci-pline construct. This APTangerscorewasthe numberof scenes that evoke an anger response from the par-ent. Using the framework of Averills (1982) work onnormative and nonnormative anger experiences,mothers completing the APT were asked to indicatetheir emotional reaction to the depicted child scene,with anger being among the possible responses. Thenumber of scenes evoking an anger response rangedfrom 1.0 to 3.83 (M= 1.98, SD= .58, = .91). The sec-ond anger indicator was the State-Trait Anger Scale

    (STAS: Spielberger, Jacobs, Russell, & Crane, 1983)that was administered during the second or third lab-oratory session. The STAS score ranged from 19 to 65(M=29, SD= 6.15). Thefinal anger composite indica-tor was a factor score of the APT pictorial angerresponse and the STAS.

    The final indicator was from the IPC observationalcoding system described above. Two scores consti-tuted the indicator, the log-transformed frequencycount of total aversive behaviors (contempt, anger,

    etc.), and the log-transformed count of total negativephysicals (e.g., hit, pinch, slap, etc.) of the motherdirected to the child.

    Care Neglect and Environmental Neglect Index

    In the Knutson et al. (2004) study, the assessment

    of parental care neglect was relatively limited andcomprised a summative index using six items from ahome visitor and from teacher observations (e.g.,childs hygiene).One focusof thecurrent studywastodevelop a multisource measure of care neglect. Thus,the assessment battery was expanded to include par-ent report and objective observer ratings of careneglect (e.g., child does not have a toothbrush), andhousehold environmental conditions that wouldoccasion social (household is overly crowded, inade-quate illumination) and physical risks to a child(unsafe stairs, inadequate plumbing, animal fecespresent, accessible pharmaceuticals). Items were all

    scored in a direction to indicate neglect and thensummed. Items were chosen based on their inclusionin the research literature or recommendations fromthe Interagency Task Force on Defining Child Mal-treatment (see Sternberg et al., 2004) and wereobtained from parent interviews and also fromobserver ratings of the home environment. Proximalcircumstances outside the home that could occasionrisk to the child that were observed during the homevisit (e.g.,broken glass, drugparaphernalia) werealsoincluded in theindex. In selectingthis summativeriskindex, we considered item relevance or face validity,

    avoided operational confounding with other con-structs, and considered the importance of compre-hensiveness as suggested by Turner and Wheaton(1995). The obtained total neglect index scoresranged from 1 to 20 (M= 9.56, SD= 3.67).

    Child Aggression

    Five indicators comprised the child aggressionconstruct. The first was the Child Behavior Checklistlong form (CBCL; Achenbach, 1992) completed bythe mothers during the first or second laboratory ses-sion.The 1992 CBCL T scores were converted to 2001norms using the Achenbach System of Educational

    Based Assessment (ASEBA; Achenbach & Rescorla,2000, 2001). Similarly, teachers directly working witheach child completed the Teacher Report Form(TRF; Achenbach, 1991) behavior rating scales dur-ing the spring semester following participant enroll-ment, at time when they should be most familiar withthechildsbehavior.The1991TRF wasalsoconvertedto 2001 ASEBA norms.

    As a third indicator, the child vignette was a mea-sure of the childs aggressive responses to four social

    CHILD MALTREATMENT / MAY 2005

    Knutson et al. / NEGLECT AND HARSH PARENTING A REPLICATION 99

  • 8/8/2019 Care Neglect, Supervisory Neglect, And Harsh

    9/16

    situational vignettes. For example, in one vignette achild is hit in the head with a ball during play. Theindex score summed the childs endorsement ofeither behavioral retaliation or hostile intent of thechild in the scenario, or both, thus the score couldrange from 0 to 8 (M= 1.68, SD= 1.76). As a fourthindicator, the interviewer rating of aggression was asummative index of dichotomous items rated by theinterviewer in the home and at each laboratory visit.Seven items constituted the checklist of itemsrescored to indicate aggression: child struck parent,positive to interviewer, cooperative, angry-irritable,

    screamed or yelled, noncompliant, friendly to par-ents. The final aggression indicator was from the IPCobservational coding system. The indicator was thefactor score of log-transformed negative physicalbehaviors directed to the parent and the log-transformed measure of aversive behaviors directedto the parent.

    Analytic Strategy

    We employed structural equation path modeling(SEM) to test a replication of thehypothesized associ-ations in the theoretical model. SEM is a latent vari-

    able regression technique that simultaneously com-bines factor analyses with path analyses under theassumptions of multivariate normality. SEM is partic-ularly suited for multimethod multisource covariancematrices used in the current study because of its abil-ity to partial measurement error of constructs and,more specifically, to specify error between same-source indicators, thus controlling forcertain types ofmono-method, mono-agent biases. Two control vari-ables were used in the analyses. Age of the child mea-

    sured in numerical years and sex of the child coded 1for female and 2 for male.

    We estimated the SEM path models using FullInformation Maximum Likelihood (FIML).FIMLuti-lizes all portions of data in a covariance matrix to esti-mate parameters in the model. SEM parameters aretypically estimated with list-wise deletion or mean-substitution matrices. However, under assumptionsof random missing data, FIML produces optimallyefficient estimates of standard errors (Arbuckle,1996; Wothke, 2000). Therefore, in the first step, weconducted a missing value analysis of the variablesmodeled in the SEM covariance matrix, includingcontrols. Littles test of missingness uses the EM algo-rithm to compare estimated mean and variance val-uesof partial data cases with complete data cases. Theanalysis indicated that the data were missing com-pletely at random (MCAR), 2 = 237.94(261), p= .84,and that FIML was the optimal estimation procedurefor SEM.

    RESULTS

    Results of the first model specifying theassociation

    of mothers social status affecting levels of childaggression is presented in Figure 2. Standardizedcoefficients are presented for the regression paths ofthe latent variable factor loadings and regressionpaths for prediction between the theoretical latentconstructs. The model obtained excellent fit to thedata with no significant difference between the speci-fied theoretical model and the observed covariancematrix, 2 = 43.30(46), p= .59,

    2/df= .94, CFI = 1.00.For estimation, the first indicator of social status (i.e.,

    CHILD MALTREATMENT / MAY 2005

    100 Knutson et al. / NEGLECT AND HARSH PARENTING A REPLICATION

    .13

    ChildAggression

    D1

    CBC-L T e5

    Ageof Child

    Sexof Child

    Mothers'SocialStatus

    Social StatusHEQSS

    e2

    MotherEducation

    e1

    AchievementHEQACH

    e3

    -.28*

    r2

    .08

    MotherOccupation

    e4

    .80***

    .44

    .73***

    .36***

    ChildScenarios

    e6

    Interviewer

    Rating

    e7

    log IPCChild toParent

    e8

    TRF T e9

    -.22*

    .54

    .39*

    .40*

    *

    .31*.4

    8**

    FIGURE 2: Child Aggression Factor Predicted by Mothers Social StatusNOTE: HEQ SS = Home Environment Questionnaire - Social Status scale; HEQ ACH = Home Environment Questionnaire - Achievementscale; CBCL-T = Child Behavior Checklist Aggression Scale T-Score; IPC = Interpersonal Process Code; TRF T = Teacher Report Form Ag-gression Scale T-Score.

    2 = 43.30(46), p= .59, 2/df= .94, CFI = 1.00.

    *p< .05. **p< .01. ***p< .001.

  • 8/8/2019 Care Neglect, Supervisory Neglect, And Harsh

    10/16

    education) and child aggression (i.e., CBCL-T) werefixed at 1.0 in the unstandardized parameters to scale

    the rest of the factor loadings. The remaining freelyestimated factor loadings for each construct were sig-nificant and ranged from 0.31 to 0.80; meaning eachindicator significantly contributed to the commonvariance underlying the theoretical construct.

    Although this sample is rather homogeneous interms of being from lower socioeconomic strata, theindicators operationalizing the social status constructobtained factor variance that was significantly differ-ent from zero (VAR= .44, SE= .07,p< .001); meaningthat although a relatively homogeneous sample wasrecruited by design, there was a range in the mothers

    achieved and ascribed social status variables. Like-wise, the factor measuring child aggression using fiveindependent sources of data (i.e., observer, teacher,parent, child, and interviewer), obtained good con-vergence and significant factor variance (VAR = 2.42,SE= .79, p< .01). Regarding the effect of social statuson child aggression, the findings supported the theo-retical model; higher levels of maternal social statuswas associated with predicted lower levels of childaggression ( = .22,p< .05). Among the control vari-

    ables, older children were estimated to be lower on

    the multimethod construct of aggression compared

    to younger children (= .28,p< .05).Overall,13% ofthe variance in child aggression was explained by

    social status, and the findings in Figure 2 were consis-

    tent with prior developmental research on harsh con-

    textual factors of social disadvantage and protective

    environmental factors of socioeconomic status and

    consistent with the theoretical model that wasused to

    structurethe current analyses (Knutson etal., 2004).

    Wenext modeled thebehavioral parentingmecha-

    nism hypothesized to link social status to child aggres-

    sion. These factors included multimethod indicators

    of care neglect, supervision tracking, and harsh puni-

    tive discipline. To evaluate generalizability of the the-oretical model, paths among the constructs were

    specified in accordance to the associations modeled

    in Knutson et al. (2004) using the longitudinal pre-

    ventive intervention sample from Oregon. Results for

    the effects of parenting mechanisms in the current

    sample are displayed in Figure 3 using standardized

    path coefficients. For visual clarity, the paths for the

    control variables were suppressed in the final figure.

    CHILD MALTREATMENT / MAY 2005

    Knutson et al. / NEGLECT AND HARSH PARENTING A REPLICATION 101

    .43

    ChildAggression

    D4

    .16

    Supervision-Tracking

    CRSSCongruence

    e6

    CEESCongruence

    e5

    .29

    PunitiveDiscipline

    log AbusiveIndex

    e8

    CBCL T e13

    D3D1

    SocialStatus

    Social StatusHEQSS

    e2

    MotherEducation

    e1

    -.07

    .12

    Care &Environmental

    Neglect

    NeglectIndexe7

    D2

    .24*

    Achieve-HEQACH

    e3

    .01

    r2 r2

    r2

    1.00

    r2

    -.12

    MotherOccupation

    e4

    -.21

    -.06

    AnalogPhysicals -Escalation

    e9

    AngerSTAS - APT

    e10

    .80***

    .44

    .46

    .29*

    *

    .63***

    .73***

    .36***

    .03 Inconsistent

    Disciplinee11

    .36*

    ChildScenarios

    e14

    InterviewerRating

    e15

    log IPCChild toParent

    e16

    TRF T e17

    .33**

    .71

    .38***

    .46**

    log IPCMother to

    Childe12

    .25*

    .22*

    -.29*

    -.53*

    .4

    8**

    .46*

    .70

    .33**

    .32*.21*

    FIGURE 3: Child Aggression Predicted by Mediating Parenting Constructs Controlling for Age and Sex of the ChildNOTE: CEES = Childrens Experienceand Excitement Scale; CRSS = Concordance Scores from the Reinforcement Survey Schedule; HEQSS = Home Environment Questionnaire - Social Status scale; HEQ ACH = Home Environment Questionnaire - Achievement scale; STAS -APT = SpielbergerState-TraitAnger ScaleAnalog Parenting Task; IPC = Interpersonal Process Code;CBCL-T = ChildBehaviorChecklistAg-gression Scale T-Score; TRF T = Teacher Report Form Aggression Scale T-Score.

    2 = 166.64(138), p= .05, 2/df= 1.21, CFI = .90.

    *p< .05. **p< .01. ***p< .001.

  • 8/8/2019 Care Neglect, Supervisory Neglect, And Harsh

    11/16

    The fully specified path model obtained adequatefit to the data, 2 = 166.64(138), p= .05,

    2/df= 1.21,CFI = .90, the 2 minimization coefficient comparingthe theoretical model and the observed covariancewas not significantly different, the comparative fit was.90 or higher and the 2 ratio 2/dfwas below the con-

    servative value of 2.00 (Byrne, 1989). Each of therespective factor loadings obtained significant contri-butions to the communality of factor variance.

    The theoretical model was supported by the data.Starting from left to right with social status, the directpath from social status to aggression was renderednonsignificant on entering the intervening mecha-nisms.Further, social statuswas marginally associatedwith lower levels of punitive discipline ( = .21, p