Вы находитесь на странице: 1из 16

AGGRESSIVE BEHAVIOR

Volume to, pages 253-267

Family Interaction: A Process Model


of Deviancy Training
G.R, Patterson, T.J. Dishion, and L, Bank
Oregon Social Leartiing Center, Eugene

A model was presented describing the reciprocal influence of disruptions in parent


discipline practices on irritable exchanges between the target child and other family
members. Disrupted parent discipline and irritable niicrosocial exchanges within the
family were hypothesized to provide a basic training for aggression that generalizes to
other settings such that the child is identifled by peers, teachers, and parents as
physically aggressive. Physical fighting was thought to lead to rejection by the normal
peer group, which was hypothesized to feed back to further exacerbate flghting.
Multilevel assessment including interview, questionnaires, laboratory studies, and
home observations were carried out with the families of 91 preadolescent and adolescent boys. Nine indicators from the assessment battery were used to deflne the constructs Inept Parental Discipline, Negative Microsocial Exchanges, Physical Fighting,
and Poor Peer Relations. Structural equations (LISREL VI) were used to describe the
relations among the constructs. The t values for the path coefficients were significant.
A chi-square analysis showed an acceptable fit between the model and the empirical
findings.
The findings were interpreted as being consistent with the hypothesis that under
certain circumstances, family interaction may serve as basic training for aggression.
In the present study, interactions with siblings in the home seemed to serve a pivotal
role.

INTRODUCTION
There is an identifiable group of investigators in such diverse content areas as
developmental psychology, primatology, and comparative and clinical psychology
who share a common interest in the analyses of social interaction sequences [Cairns,
1979a; Sackett, 1977; Suomi et al, 1979]. Cairns 11979b] has labeled the perspective
shared by these investigators as "social interactional." A characteristic shared in
common with these studies is the search for patterns of action and reaction. Typically
these patterns occur with sufficient regularity in interaction sequences that they may
be reliably identified in samples of distressed married couples [Gottman, 1979;

Address reprint requests to Gerald R. Patterson, Oregon Social Learning Center, 207 E 5th Ave, Suite
202, Eugene, OR 97401.

1984 Alan R. Liss, Inc.

254

Patterson, Dishion, and Bank

PARENTAL
DISCIPLINE

NEGATIVE
MICROSOCIAL
EXCHANGES

POOR
PEER
RELATIONS

If
>
PHYSICAL
FIGHTING

Fig. 1. Hypothesized social interactional model of physical fighting.

Margolin, 1977] or across samples of families of antisocial children [Patterson, 1982;


Snyder, 1977]. The hypothesis examined in this report is that patterned exchanges in
social interaction may represent a vehicle for training some forms of deviant child
behaviors.
As presented in Patterson [in press), it is the patterned irritable exchanges between
the problem child, his mother, and siblings that define "basic training" for coercion.
As used here the term coercion refers to the use of aversive behaviors in both
punishment and negative reinforcement arrangements. The effect of these negative
microsocial exchanges is to shape the behavior of aggressor and victim simultaneously. The training in coercion or pain-control techniques is in turn thought to be a
necessary prelude to training in physical fighting. It is assumed that the first or basic
training phase is the result of persistent failure of the parents to use effective discipline
in controlling coercive exchanges among family members. The patterned irritable
exchanges increase in frequency and amplitude, which further disrupts parental efforts
to discipline. For the model summarized in Figure 1, the coercive exchanges and
physical fighting are coping styles that are generalized from family interaction to
interchanges with peers. The coercive style and physical aggression places the child
at risk for poor relationships with his peers. This peripheral status in turn increases
his disposition to attack.
The present report examines the correlational evidence for the structural relationships among the key constructs. In the sections that follow, each component of the
model is discussed in greater detail together with the research literature relating to the
model. The last section summarizes the findings from the current study.
GENERAL FORMULATION
In the past two decades, several hundered families of antisocial boys have been
treated at the Oregon Social Learning Center (OSLC). This experience has led to the
clinical impression that the parents of antisocial boys are "deficient" in one or more
of the following family management skills: (1) they do not adequately monitor the
whereabouts of their child; (2) they do not use effective discipline for the child's
antisocial behavior; (3) they are relatively unskilled in negotiating and problem
solving; and (4) they seldom reinforce effective prosocial survival skills (academic,
relational, work). These concepts were the focal point in the family treatment programs descibed in Patterson et al, [1975], and illustrated in audio tapes [Patterson and
Forgatch, 1976] and video tapes [Forgatch et al, 1982].
Clinical experience has also suggested that the antisocial problem child is often
characterized by two problems. Not only is he antisocial, but in addition, he is often

Deviancy Training

255

retarded in such critical social survival skills as peer relations and academic and work
skills [Dishion et al, in press; Patterson, 1982]. It was assumed that the parental
failure to reinforce work and academic behaviors partially determined the poor
academic and work performance. For example, measures of parent reinforcement
and effectiveness in problem solving, described in Patterson [1983], showed significant correlations with measures of reading achievement and social competence.
Efforts to treat families of antisocial children quickly led to an emphasis on teaching
parents more effective discipline and tracking techniques. The hypothesis was that
the parent's failure to discipline and monitor the child produced increases in the
child's antisocial behavior. In the study of normal families, the parent monitoring and
discipline practices for a sample of adolescent boys was shown to correlate significantly with both self-reported delinquency and the frequency of police contacts
[Patterson and Stouthamer-Loeber, in press].
The primary assumption is that changes in the parents' use of the family management procedures produce concomitant changes in child behavior. Parents trained in
the use of family management skills have produced significant reductions in observed
antisocial behavior; and the effects have been shown to persist for at least 12 months
[Patterson and Fleischman, 1979]. Comparison studies employing random assignment
have shown that leaderless parent discussion groups [Walter and Gilmore, 1973] and
traditional family treatment [Patterson et al, 1982] are not associated with significant
changes in observed child antisocial behavior. However, as yet no studies have been
designed that demonstrate that changes in the parent family management measures
covary significantly with reductions in child problem behavior.
Parental discipline
As noted earlier, parental failure to discipline is thought to be a major determinant
for increases in antisocial child behavior. It is also the case that training parents to
use an effective punishment is one major component of the treatment for these families
[Patterson et al, 1975]. However, such an emphasis upon teaching parents to use
more effective punishment poses a paradox. Studies consistently show that parents of
antisocial children tend to be more punitive than are parents of normal children [cf
review by Berkowitz, 1973]. In fact, the correlations between parent punishment and
antisocial child behavior are so consistent as to lead some writers to take the position
that parental excesses of punishment cause the antisocial behavior of problem children
[Welsh, 1976]. Observation data collected in home have shown that in fact the parents
of problem boys are more aversive (punishing) in their reactions to the antisocial boy
than are parents of nonproblem boys [Patterson, 1982; Snyder, 1977]. However, it is
the case that the problem child engages in significantly higher rates of antisocial
behavior [Patterson, 1976, 1982]. Calculating the likelihood of the parents reacting
aversively given that the child is coercive corrects for the differences in rate of deviant
child behavior. The data show that parents of problem children are roughly twice as
likely to react aversively as are parents of nonproblem children [Patterson, 1982].
The data from these field studies are in keeping with well-established research
fmdings summarized earlier; parents of antisocial boys both punish more frequently
and are more likely to punish.
Observation studies also show that parents of antisocial children use more extreme
punishment. Mothers of socially aggressive boys are seven times more likely to hit
than are mothers of normal boys [Patterson, 1982]. Similarly, mothers of abused
children are observed to be significantly more likely to hit the problem child than are

256

Patterson, Dishion, and Bank

mothers from a matched normal group [Reid, 1983a]. In that study, daily telephone
calls showed that the mothers of abused children rated themselves as more angry in
the previous day's discipline confrontations and were more likely to use "hit, spank,
or physical restraint" as punishment.
How is it, then, that a focal point of the social learning treatment is to increase the
effectiveness of parent punishment for antisocial behavior? There are two issues that
relate to this seeming paradox. First, it is the case that a substantial proportion of the
aggressive events in families are a reaction to an aversive intrusion by a family
member. For children in normal families the likelihood of an aversive parent behavior
given a coercive child behavior is .26; for the social aggressor the likelihood is .34
[Patterson, 1982]. Given such an intrusion, the child's coercive reaction in both
samples is successful 40% of the time (ie, the other person terminates the noxious
intrusioq). What this implies is that a substantial portion of problem-child coercive
reactions in the home (21%) may be strengthened by negative reinforcement (a
response followed by withdrawal of an aversive stimulus). The assumption is that
aggressive child behaviors are maintained by mixed schedules of positive and negative
reinforcement plus frequent punishment. An experienced aggressive child has had,
then, a great deal of experience with aversive stimuli. Such events, in fact, constitute
the warp and woof for much of his coercive behavior. Parental "nattering" such as a
mild threat, disapproval, or scolding as a consequence is likely to serve as "just
another" aversive stimulus. Such a parental consequence may well be met with yet
furtlier aggression, and may become part of the negative reinforcement escalation
pattern.
For families of both normal and problem children, the vast bulk of the parental
punishment is mildly aversive. The next reaction of the problem child to such a
"punishment" is very likely to be that of accelerating his ongoing coercive behavior.
For example, Reid [1983a] found that the likelihood of the child continuing with his
coercive behavior, given that the parent reacted aversively to the prior one, was only
.14 for normal children, but .35 for antisocial children, and .53 for abused antisocial
children. All of the differences among these samples were statistically significant.
Antisocial boys are significantly more likely than normal boys to continue being
coercive when mildly punished by parents. Furthermore, the disposition to continue
coercive behavior when punished may prove to be a gauge of how far the target
child's fight training has progressed. The disposition to continue in the face of parental
aversive reactions correlated .466 (P < .001) with the overall measure of the child's
coerciveness in the home [Patterson, 1979].
The analyses of sequential data shows that aversive consequences to function as if
they suppress ongoing prosocial behaviors [Patterson and Cobb, 1971], but the same
consequences are associated with acceleration for coercive events! The findings for
acceleration to mild punishment led us to differentiate parental nattering from parental
discipline. Both kinds of parental reactions would be judged by the child as being
aversive; but we believe that only one of them, discipline, reduces the strength of a
stimulus-response (S-R) connection.
These findings from field studies are in agreement with those from the programmatic studies of laboratory aggression reviewed in Follick and Knutson [1977]. They
showed that aggression elicited by an aversive stimulus ("irritable aggression")

Deviancy Training

257

followed by a mild aversive consequence is likely to be accompanied by yet further


aggression. However, when the aversive consequence was of at least the same
duration and intensity as the antecedent stimuli then the effect was suppression of
aggression. We believe that parent use of time-out or withdrawal of privileges
represents an analogous increment in the intensity of aversive consequence. A series
of laboratory studies has demonstrated that a few minutes of time-out or withdrawal
of privileges will not only suppress ongoing aggression but probably weaken the S-R
connections as well [see Patterson, 1982]. One factor to consider in connection with
the effectiveness of time-out is that the duration of a burst of child aversive behavior
in more than 90% of normal families is approximately one to 11 seconds [Reid, in
press]. Thus even a one-minute time-out may be seen as a fourfold increase in
aversiveness over the preceding child behavior. This may explain why the consistent
use of a minute or less of time-out was shown to be more effective with preschool
children than merely stating the rules [Hobbs et al, 1978]. The Forgatch and Toobert
study [1979] showed that furnishing the parents with a pamphlet describing time-out
and a brief modeling of its use were sufficient to produce changes in child behavior
that persisted for at least a month. The general set of findings buttresses the assumption that parents can be trained to use time-out and withdrawal of privileges as an
effective backup for parental requests, threats, and scolding. Presumably, the result
will be to increase the effectiveness of parent "stop" commands; this in turn eventually reduces the need for their use.
In keeping with this formulation, the pre-post treatment comparison by Taplin and
Reid [1977] showed that, following parent training, the mothers of antisocial boys
were significantly less likely to use aversive consequences for coercive child behavior.
Reid [1983b] also found that home observation data and parent daily telephone reports
showed significantly less parental hitting following parent training in abusive families.
Presumably, the use of more effective discipline techniques reduces not only the rates
of deviant child behavior but increases the effectiveness of parent command to "stop."
Microsocial exchange: Training for fighting
While various writers might agree upon the central role played by parental discipline in relation to antisocial child behavior, there is a good deal of disagreement as
to how this relation comes about. For example, one could think of the parent as
modeling the very aggressive behavior he or she seeks to control by frequently
scolding and threatening and occasionally physically abusing the child. An alternative
hypothesis presented here is that the parental failure to discipline is the first step in a
process that eventually leads to increases in physical fighting within and outside of
the home. The second step in this process is defined by increases in the rate of
coercive exchanges. The increases in rate are accompanied by an increase in frequency for extended coercive chains. The next step is an increase in hitting; this is
particularly likely during longer chains as shown by Reid [1983a]. Most of the hitting
occurs in exchanges between the target child and siblings [Patterson, in press]. It is
in this sense that one might think of the siblings as playing a key role in fight training.
Three dispositions are thought to be of critical importance in determining extended
coercive chains: start-up, which describes the likelihood that one member will
commence conflict; punishment, which describes the likelihood the other member

258

Patterson, Dishion, and Bank

will react aversively; and continuance, which describes the likelihood that the first
member will continue being aversive no matter what the reaction to his aversive
initiation.*
Synchronicity is introduced as a concept to explain the escalations in levels during
step 1 among all family members. This refers to the disposition to react in kind to
aversive initiations by other family members. The studies reviewed by Patterson
[1982] showed that distressed couples and members of distressed families were more
likely to react in kind (aversively) to aversive interactions. The increase in likelihood
of synchronous reactions ranged from 18% to 42% for dyads in families of social
aggressors as compared to dyads from normal families. If one member increases, the
other "matches up." A simple means for testing this idea is to standardize each of the
Start-up, Punishment, and Continuance scores and sum them to form a composite
"irritability" total score for each member of the dyad. For the normal sample, it was
found that the target child and his siblings were closely matched, as shown by the
correlation of .57 (P < .001) for their irritability scores. The comparable correlation
for the mother and the target child was .35 (P < .001), and for the father and the
target child .29 (P < .05). These findings support the idea of a kind of mutuality in
the irritable exchanges for family members.
Rejection by Peers
Clinical experience in treating several hundred antisocial boys suggest that most of
them are relatively lacking in the social skills required to maintain friendship with
members of the normal peer group. A review of the peer sociometric studies gives
strong support to the hypothesis; most of the studies showed that peers reject antisocial
boys [Patterson, 1982]. The relation between being rejected by normal peers and
status as an oppositional child has been known for some time [for example, Moore,
1967]. Gottman's [1977] programmatic work on friendship in younger children
heavily emphasizes the role of negative coercive behaviors in determining peer
rejection. The amount of interaction a boy has with the peer group increases as a
function of age. For this reason rejection by peers may have a greater impact as the
child grows older. For example. Barker and Wright [1954] found that as much as
50% of an 11-year-old boy's social activity involved other children.
The child's failure to develop effective relations with peers may be the outcome of
several processes. On the one hand, he tends to generalize the coercive interpersonal
style he learned in the family to settings such as the school. Presumably his abrasive
manner of interacting leads to rejection and conflict in the normal peer group
*In regard to stability of these behaviors, test-retest correlations were calculated for a sample of 21 boys
from normal families. They were observed in the homes for six sessions, and then again for two more
sessions about eight weeks later. The test-retest correlations for the Start-Up score were .71 (P < .001),
for Counterattack .44 (P < .10), and for Continuance .65 (P < .10). The score described only the
target boy's reaction to his mother, and was based upon the combined data for his protocol (as target)
and her's. The comparable stability correlations for her reactions to him were .39 (P < . 10), .54 (P <
.10), and .37 (ns), respectively.
It should be noted that the only scores considered were from individuals where the digit(s) in the
denominator exceeded 4. The findings suggest that for the normal sample, three sessions are an absolute
minimum for stable estimates of these conditional P values. Four of five sessions would probably increase
the reliability considerably.

Deviancy Training

259

[Gottman, 1977]. His lack of work, academic, and recreational skills further exacerbates the situation [Dishion et al, in press]. As shown in Figure 1, the assumption is
that poor peer relations are primarily the outcome of practiced coercive exchanges in
the family and only indirectly related to disruptions in parental discipline. As shown
in an earlier study, the failure to develop these social skills may be due to the parent's
inept use of positive reinforcement (and problem solving). Moderate correlational
support for this hypothesis was provided in the report by Patterson [1983].
PRESENT STUDY
A proper test for the model requires both a longitudinal design and experimental
manipulations that demonstrate, for example, that changes in parent di.scipline produce the sequence of effects described in the model. Such studies are currently
underway; in that context the present report is best thought of as a first-stage effort.
It examines the correlational data collected in a cross-sectional study to determine the
fit of the measures to the model. As pointed out by Bentler [1980], one of the prime
functions of the LISREL analysis is that it permits one to evaluate the viability of the
hypothesized model (i.e., the internal coherence of the structure). A lack of fit
between the empirical findings would mean that the model should be altered prior to
its being tested in a longitudinal design.
Wherever possible the constructs described in the model were defined by multiple
indicators [Bentler, 1980]. For two of the four constructs it was possible to include
reports from different agents. Overall, the indicator for the model sampled six
different modes of assessment: self-report, parent report, teacher report, molecular
observation data, global ratings by observers, and peer-nomination. The data were
based on the multilevel assessment of 91 families of boys in their early and mid
adolescence.
The relative accuracy of the representation was examined within the structural
modeling format. The LISREL VI program was used to determine whether the Inept
Discipline, Negative Microsocial Exchange, Poor Peer Relations, and Physical Fighting constructs related to each other in the manner specified in Figure 1.
METHODS
Subjects
The data reported here are part of a three-year planning study prepared for a
longitudinal study on the development of antisocial behavior. Two hundred ten
families were recruited from 21 schools in an Oregon town of about 100,000 people.
These schools provided names and addresses of all families with boys in the fourth,
seventh, and tenth grades. Approximately 1,000 families initially received letters
explaining the research procedures and soliciting their participation. The introductory
letters were followed by telephone calls from the project staff and, in the majority of
cases, subsequent home visits. Three hundred families agreed to particpate in all
phases of the study, and 210 of these families were assessed. The 210 families included
74 fourth-grade boys (ages 9-10), 78 seventh-grade boys (ages 12-13), and 58 tenthgrade boys (ages 15-16), and their parents. All but one of the boys were White. The
socioeconomic levels of the 210 families were computed with the Hollingshead Index
of Socioeconomic Status (SES) [unpublished manuscript]. The sample included 8.6%

260

Patterson, Dishion, and Bank

in level 1 (unskilled laborers), 23.3% in level 2 (semiskilled workers), 29,0% in level


3 (clerical and skilled workers), 32.9% in level 4 (minor professional), 3.8% in level
5 (professional education and employment). Single-parent families comprised 28.1%
of the sample.
Evidence from other longitudinal studies on antisocial behavior show that families
most likely to refuse participation or to drop out are likely to have a higher proportion
of antisocial children [Ghodsian et al, 1980; Lefkowitz et al, 1977; Rutter et al,
1970]. We therefore expected the same to be true with this research.
In our final sample, 10.68% of the boys had a court record, whereas a search
through the court files and an estimation of the number of boys in fourth, seventh,
and tenth grades in the area from which the sample was drawn revealed that we
should have had 16.33% ofthe subjects with a court file, had the sample been random
in this respect. Even though there were fewer delinquents in our sample than was
expected on the basis of the base rate in the population, our delinquent sample did not
statistically differ in number of offenses, number of nonstatus offenses, or number of
delinquent episodes from a randomly drawn sample of court files of boys in the same
age groups.
Procedures
The data-collection procedures for this planning study were categorized into three
phases. Phase 1 involved the collection of teacher ratings and peer-nominations.
Phase 2 consisted of a three-hour structured interview at the Oregon Social Learning
Center. Within the three-hour session, the parent(s) and child were interviewed
separately, completed questionnaires, and participated in a 30-minute structured
videotaped family problem-solving task. The parents also completed the Child Behavior Checklist, a standardized questionnaire assessing child behavior problems [Achenback, 1978], During phase 3 ofthe data collection, telephone interviews and home
observations were conducted. Home observations were comprised of three one-hour
sessions of trained observers coding family interactions with the MOSAIC observation code [Toobert, Patterson, Moore, and Halper, unpublished manuscript]. Due to
budgetary constraints, only 91 of the 210 study families could be observed in their
home.*
Below is a description ofthe latent variables included in the process model.
Operational Definitions for Constructs
Poor Diseiptine (Eta 1). This constuct defines the extent that the parent was
observed to be consistent and to follow through on his/her discipline practices in the
home, and his/her use of effective discipline practices. The indicators were based
upon ratings made by the observers following each of the three sessions in the home.
They rated the mother on, "Did the mother follow up on commands?" This score

*To determine if there were systematic differences between the 91 families observed and the original
sample of 210, an analysis of variance (ANOVA) was computed comparing the partial sample (N = 9 1 )
with the remaining sample (N = 19) on the three indicators of physical fighting. These univariate
analyses revealed no statistically reliable differences between the two subsamples.

Deviancy Training

261

consisted of the average number of "yes" responses across sessions. They also rated
each parent on an item reading, "Mother's consistency in discipline style." Mothers
were scored if they were observed to be erratic or inconsistent, and the average across
sessions was used as the indicator score.
Negative Microsociat Exchange (Eta 2). This construct was intended to define
the level of coercive and physically aggressive interactions between the study child
and his mother and siblings as observed in the home. Six out of ten scores showed at
least moderate intercorrelations; these six scores were entered into the measurement
model:
Person X Punishes Person Y: This score represented the likelihood that the target
person reacted coercively given the interactant previously initiated a coercive response to him. There were 16 content codes defined a priori as coercive: Command,
Command Ambiguous, Complain, Criticize, Defend Self, Destructiveness, Disapproval, Euture Redirect, Get-Off-My-Back, Grab, Physical Negative, Provoke, Refutation-Denial, Restrain, Threat, and Yell.
There were four scores for this variable describing child to mother, mother to
child, sibling to child, and child to sibling. The sibling interchanges entered wto the
database were all instances of any sibling interacting with the target child.
Physical Coercion: This score described the proportion of the interaction that was
physically aggressive between the target child and the other person. Four content
codes define physical aggression: Destructiveness, Grab, Physical Negative, and
Restrain. There were two scores used in the analyses. One described the hitting in the
target child and mother exchanges and the other described the exchanges between
siblings and the target child.
Poor Peer Relationships (Eta 3). The quality of the child's relationships with
peers was assessed by three measures based upon data obtained from peers, teachers,
and child's self-report. In the structured interview, the child completed a skills
checklist questionnaire including the item, "How well do you get along with other
kids?" The child rated this item on a 5-point scale ranging from 1 (very well) to 5
(very poorly). Teachers rated the item, "This boy is competent in making friends
with his peers," A 5-point rating scale was used ranging from 1 (totally accurate) to
5 (very inaccurate). Peers nominated two to three children who best fit the following
descriptor: "Kids who don't get along with most other kids," The number of
nominations received was standardized by administration group to control for variation due to class size.
Physical Fighting (Eta 3). This is the latent variable assessing the child's use of
physical fighting across settings (ie, the home and school). It was also based upon
three measures. The mother's ratings on two items from the Child Behavior Checklist
were used to assess the extent each subject child physically fought in the home
environment. Mothers rated the following two items on how true they were for their
child (0 = not true, 1 = sometimes true, 2 = often true): "Gets in many fights" and
"Physically attacks people," Peers nominated two to three children in their classroom
who best fit the following descriptor: "Kids who fight a lot with other kids." The
number of nominations received was standardized by administration group. The
teacher rated each child on the following item on a 5-point scale: "This boy physically
fights with other children," with 5 representing "almost always" and 1 "almost
never." These scores were standardized by grade level for all subjects.

262

Patterson, Dishion, and Bank

RESULTS
The assessment procedures were designed to provide an approximate fit to broad
social learning concepts concerning the role of the family in the etiology of antisocial
behavior. Specific family management constructs included those in Eigure 1 and
emerged after the data were collected [Patterson, 1982]. Eor this reason, in the post
hoc search for multiple indicators there was often some difficulty in establishing three
good indicators for each latent variable. Three is hardly a magic number but rather a
guide that most researchers who use structural equations in the LISREL format appear
to agree is reasonable [Bentler, 1980], It is, however, at times appropriate to test a
model with some of the constructs defined with two or even one indicator [eg, see
Huesmann et al, in press]. As shown in Eigure 2, Poor Peer Relations and Physical
Eighting were each defined by three indicators.
The Negative Microsocial Exchange construct was defined by two indicators
representing the likelihood of punishment between the siblings and the target child
using each as both the target and interactant. The mother's likelihood of punishment
score did not sufficiently correlate with the sibling scores to permit inclusion as an
indicator for this construct. Another possible indicator for this construct, the proportion of physical coercion between the child and his siblings, was dropped because of
its extremely low base rate and small standard deviation.

% \

CONSISTENCY

SIBLING
PUNISHES
CHILD

CHILD
PUNISHES
SIBLING

PEERS

TEACHERS

CHILD
SELF-REPORT

,487

Fig, 2, A social interactional process model of discipline, microsocial exchanges, and peer relationships
as they relate to fighting.

Deviancy Training

263

Inept Discipline was to be defined by observer's global ratings of parental inconsistency as well as the lack of followup on commands. The relation between these two
variables, however, was unusual. Scores for inconsistency and the lack of followup
were gathered at the same home observation sessions, but were found to be negatively
correlated. It was thought that the negative correlation might be due to the fact that
parents who were highly consistent in their discipline practices were far less likely to
need to followup on their commands during the home observations. Conversely, those
parents who were not effective and/or not consistent with their discipline needed to
followup on their commands. Thus the consistency variable during the home observation period was thought to provide the most accurate indication of effective parent
discipline and followup was discarded for these analyses. Figure 2 contains the
hypothesized model with each of the nine indicators used in the final analysis. Means
and standard deviations for the nine indicators and their intercorrelations are presented
in Table I.
The analyses were conducted using LISREL VI fJoreskog and Sorbom, 1978].
Given that the Inept Discipline construct was represented entirely by a single indicator, there was no residual (error) for that latent variable; technically, zeta 1 was fixed
at 0. The model, as tested, hypothesizes a bidirectional process between Inept
Discipline and Negative Microsocial Exchanges as well as a similar process between
Physical Fighting and Poor Peer Relations. The paths from Negative Microsocial
Exchanges to Physical Fighting and to Poor Peer Relations were both hypothesized to
be unidirectional processes. The analysis converged, and resulted in a satisfactory fit
of the model to the data (x24 = 27.80, P > .32). With the exception of the path
coefficient from Microsocial to Peer Relations, all of the paths were statistically
significant. With this one exception, the model represented in Figure 1 is a resonable
one. A second analysis was performed with this path deleted (see Fig. 2). That
analysis resulted in a X25 = 27.97, P > .31. This model provides an excellent fit to
the data, though the difference between this model and the previous one including
beta 32 is not a statistically significant one (xi = 0.17, P > .2). Thus this second
model, while not providing a significantly better fit, does provide a more parsimonious desciption of the data. As of this writing, these two competing models both
require further testing.

TABLE I. Means, Standard Deviations, and Intercorrelations Among Indicator Variables


1
Discipline
Inconsistency
1.0
Microsocial
Child to sib
.23
Sib to child
.04
Poor Peer Relations
Self
.07
Peers
.24
Teachers
.19
Physical Fighting
Mother
.18
Teacher
.16
Peers
.16

1.0
.43

1.0

.17
.02
.04

.09
,21
-.03

1.0
.19

.25

.32

.10
.15
.26

.20
.18
.29

.00
.10
.01

.24
.45
.54

1.0
1.0
.17
.23
.18

1.0
.32
.22

1.0
.53

1.00

sd

3.11

0.820

.02
.00

1.010
0.770

1.91
-.05
2.24

0.94
1.07
0.73

.34
-.06
-.03

0.73
0.884
I.OO

264

Patterson, Dishion, and Bank

While model two was taken as the best fit at this time, other competing models
were also considered. For example, a path from Inept Discipline directly to Poor Peer
Relations was considered but found to be statistically nonsignificant. A path directly
from Inept Discipline to Physical Fighting was added to model 2; it proved to be
untestable at this time due to a dependency within the data set.*
The final form of model 2, with its indicators, indicator residuals, and path
coefficients, is presented in Figure 2, The t scores for the path coefficients and
indicator loadings on the latent variables were statistically significant (P < ,05; onetailed test),
DISCUSSION
These fmdings are consistent with the model that family interaction may serve as a
basic training for fighting. Disruptions in parents' use of discipline were associated
with increases in target child-sibling irritable exchanges; these, in turn, were related
to increased disruption in discipline practices. What is of theoretical interest is that
the effect of inept discipline upon fighting was mediated by the measure of microsocial
process. Also, the effect of inept discipline upon peer relations was mediated by
physical fighting through microsocial process. In both instances, the microsocial
mechanism seems to function as the important mediating function ascribed to it within
coercion theory [Patterson, 1982], This preliminary analysis is sufficiently supportive
of the model to warrant an effort to develop additional indicators for the discipline
construct and employ it in the longitudinal studies currently underway.
The current social interactional perspective contributes to the accumulated body of
research on the acquisition of aggressive behavior in boys by emphasizing the role of
siblings in generating coercive chains that lead to physical fighting. The training
seems to generalize to the school since both teacher and peer (classmate) ratings
loaded significantly on the Peer Relations and Physical Fighting latent factors. Thus
it may well be that family interaction mechanisms in part determine the highly stable
and generalizable trait aggressiveness [Loeber, 1982; Olweus, 1979].
In part the contribution of the present report may be primarily a methodological
one. Other investigators have implicated the family as a possible determinant for
aggressive behavior [Eron et al, 1971; Farrington, 1978; Lefkowitz et al, 1977;
McCord et al, 1963], Most often the database has consisted of global ratings from a
single interview. However primitive such data might be, they have generated lowlevel, positive correlations for child-rearing characteristics with later aggression and
implicated variables such as parents in conflict, poor supervision, and inconsistent
discipline as important. The inclusion of home observation data enables us to elaborate on our understanding of why some of these earlier findings might have appeared
so consistently. For example, in the current report parent discipline can be seen as
only a first step that may lead to a deviant training process.
The second methodological point is more implied than explicit. It is assumed that
indicators sampling across assessment modes and agents might provide more power
*The actual problem in the LISREL analysis was one of an unidentified parameter. Structural equation
modeling has not been used much with observational data thus far; the use of this statistical technique
clearly presents a challenge to researchers using observational techniques because it is often difficult to
collect such data while maintaining independence of each observation. No doubt, such data sets will be
more accurate and valuable if this challenge can be met.

Deviancy Training

265

in understanding the relation among the various constructs in this and other tnodels
as well. The multirespondent measures of Physical Fighting and Poor Peer Relations
are good examples of this approach. In the large-scale study now underway, a
systematic examination will be made of the relative utility of multimode/multiagent
indicators for latent constructs.
The acceptance of the present model does not, of course, preclude the acceptance
of alternative models. Furthermore, the agreement between this structural model and
the covariance matrix does not imply a verification of an implied causal process. This
point has been made by methodologists fBentler, 1980] and theorists [eg, Baumrind,
in press] alike. For example, Hirschi and Selvin [1967] suggested that in order to
infer that 'A' causes ' B ' , it must be shown that: (1) 'A' covaries with "B' (2) variable
'A' temporally precedes the occurrence of variable 'B', and (3) 'A' and 'B' are not
mutually caused by a third variable, ' C . The current study satisfies some of the
requirements for step 1. The longitudinal studies currently underway will perhaps
satisfy some of the requirements for step 2. As pointed out by Baumrind [in press]
and others, it is not possible to "prove" step 3. We can only test the hypothesis that
A causes B by rigorous efforts to disprove it [Popper, 1972]. For example, in a
treatment manipulation (experiment), improvements in parental discipline practices
should be followed by reductions in observed child and sibling negative interactions,
and also by subsequent reductions in physical fighting within the home. For a
comparison group where .school personnel employed more effective discipline for
fighting [eg. Walker, 1979], there would be no change in target child-sibling exchanges. In this broader context, then, the current analysis is seen as a necessary but
not sufficient step in establishing a process model of explaining physical aggression
in boys.
In that same vein, the experience of the present writers is that one of the main
utilities of the structural modeling approach lies in the effect the process has upon the
thinking of the investigators. Being engaged in this formatexamining a m o d e l forces the participants to clarify their definitions of constructs as well as the nature of
the model itself. This explanatory process occurred at several points in the present
analyses. For example, examination of the measurement model for the microsocial
construct showed clearly that for children aged 9-16, the irritable exchanges of
mother and target child added very little beyond that which was contributed by
measures of irritable exchanges with siblings. Patterson's [1980] earlier formulation
about the critical role played by the mothers' irritable exchange was based upon a
younger sample. At the very least the current analyses force one to reconsider the
generalizability of the idea of the mother as trainer across age groups.
Aside from the lack of replication data, the primary weakness of the current
analyses lies in the inadequate definition for the Inept Parent Discipline. As previously
noted, the use of a single mode of assessment by a single agent is hardly a satisfactory
state of affairs. Studies currently under way employ more behavioral items for the
interviews and in addition a powerful measure of parent discipline devised by Reid
[1983b]. Reid's daily telephone interview with the parent identifies not only the
antisocial behavior occurring within the last 16 hours, but also the parent punishment
for that event. In his sample of clinical cases the daily telephone report correlates
with the observer impression measure of Inept Parent Discipline employed in the
current report. It seems reasonable to believe that the definition of this key construct
will be improved in future studies.

266

Patterson, Dishion, and Bank

ACKNOWLEDGMENTS
Support for this study was provided by grants No. 1 ROl MH 32857 and 2 ROl
MH 37940 from the Center for Studies of Crime and Delinquency, National Institute
of Mental Health, US PHS.
REFERENCES
Achenbach TM (1978): The child behavior profile,
I: Boys aged 6-11. Journal of Consulting and
Clinical Psychology 46:478-488.
Bandura A (1973): Social learning theory of
aggression. In Knutson JF (ed): "The Control
of Aggression." Chicago: Aldine.
Baumrind D (In press): Specious causal attributions in the social sciences: The reformulated
stepping-stone theory of heroin use as exemplar. Journal of Personality and Social
Psychology.
Barker RG, Wright HF (1954): "Midwest and Its
Children." Evanston, IL: Row Peterson.
Bentler PM (1980): Multivariate analyses with latent variables: Causal modeling. Annual Review of Psychology 31:419-456.
Berkowitz L (1973): Control of aggression. In
Caldwell BM Riecuter (eds): "Review of Child
Development Research." Chicago: University
of Chicago Press.
Cairns RB (1979a): The Analysis of Social Interaction: Methods. Issues, and Illustrations.
Hillsdale, NJ: Lawrence Eribaum Associates.
Cairns RB (1979b):"Social Development: The
Origins and Plasticity of Interchanges." San
Francisco W.H. Freeman and Company.
Dishion TJ. Stouthamer-Loeber M, Loeber R, Patterson G (In press): Skill deficits and male
adolescent delinquency. Journal of Abnormal
Child Psychology.
Eron LO, Walder LO, Lefkowitz MM (1971):
"Learning of Aggression in Children." Boston: Little-Brown.
Farrington DP (1978): The family backgrounds of
agressive youths. In Herson LA, Berger M,
Schaffer D (eds):"Aggressive and Antisocial
Behavior in Childhood and Adolescence."
Follick MJ, JF Knutson (1977): Punishment of
Irritable aggression. Aggressive Behavior 4:
1-17.
Forgatch MS, Patterson GR, Chamberlain P, Kavanagh K, Gabrielson P (1982): "Time-out."
30-minute videotape. Eugene, OR: Castalia
Publishing Co.
Forgatch MS. Toobert DT (1979): A cost-effective
parent training program for use with normal
preschool children. Journal of Pediatric Psychology 4:129-145.

Ghodsian M, Fogelman K, Lambert L, Tibbenham


A (1980): Changes in behavior ratings of a
national sample of children. British Journal of
Social and Clinical Psychology 19:247-256.
Gottman JM (1979): "Marital Interaction: Experimental Investigations." New York: Academic
Press.
Gottman JM (1977): Toward a definition of social
isolation in children. Child Development
48:513-517.
Hirschi T, Selvin HC (1967): "Delinquency Research: An Appraisal of Analytic Methods."
New York: The Free Press.
Hobbs SA, Forehand R, Murray RG (1978): Effects of various deviations of time-out on noncompliant behavior of children. Behavior
Therapy 9:652-656.
Huesmann LR, Eron LD. Lefkowitz MM, Walder
L (In press): The stability of aggression over
time
and
generations.
Developmental
Psychology.
Joreskog KG, Sorbom D (1978): "LISREL IV."
Chicago: National Education Resources.
Lefkowitz MM, Eron LD, Walder LP, Huesmann
LR (1977): "Growing up To Be Violent." New
York: Pergamon Press.
Loeber R (1982): The stability of antisocial and
delinquent child behavior: A review. Child Development 53:1431-1446.
Margolin G (1977): "A Sequential Analysis of
Dyadic Communication." Paper presented at
the meeting of the Association for the Advancement of Behavior Therapy. Atlanta,
Georgia, December.
McCord J, McCord W. Howard A (1963): Family
interaction as an antecedent to the direction of
male aggressiveness. Journal of Abnormal and
Social Psychology 66:239-242.
Moore S (1967) Correlates of peer acceptance in
nursery school children. Young Children
22:281-297.
Olweus D (1979): Stability of aggressive reaction
patterns in males: A review. Psychological
Bulletin 86:852-875.
Patterson GR (In press): The contributions of siblings to training for fighting: A microsocial
analysis. In Block J, Olweus D, Radke-Yarrow
M (eds): "Development of Antisocial and Pro-

Deviancy Training
social Behavior." New York: Academic Press.
Patterson GR (1983): The unattached mother: A
process analysis. In Hartup W, Rubin Z (eds):
"Relationships: Their Role in Children's Development." Cambridge, MA: Harvard University Press.
Patterson GR (1982): "Coercive Family Process."
Eugene, OR: Castalia Publishing Co.
Patterson GR (1980): Mothers: The unacknowledged victims. Monographs of the Society for
Research in Child Development 45(5,Serial
No. 186).
Patterson GR (1979): A performance theory for
coercive family interaction. In Cairns B: "The
Analysis of Social Interactions: Methods, Issues, and Illustrations." Hillsdale, NJ: Lawrence Eribaum.
Patterson GR (1976): The aggressive child: Victim
and architect of a coercive system. In Hamerlynck LA, Handy LC, Mash EJ (eds): "Behavior Modification and Families: Theory and
Research" Vol. 1. New York: Brunner/Mazel.
Patterson GR, Chamberlain P, Reid JB (1982): A
comparative evaluation of a parent-training
program. Behavior Therapy 13:638-650.
Patterson GR, Cobb JA (1971): A dyadic analysis
of "aggressive" behaviors. In Hill JP (ed):
"Minnesota Symposia on Child Psychology,"
Vol. 5. Minneapolis, MN: University of Minnesota Press.
Patterson GR, Fleischman MJ (1979): Maintenance of treatment effects: Some considerations concerning family systems and follow-up
data. Behavior Therapy 10:168-185.
Patterson GR, Forgatch MS (1976): "Family Living Series. Part 1." (Five cassette tapes.) Eugene, OR: Castalia Publishing Co.
Patterson GR, Reid JB, Jones RR, Conger R
(1975): "A Social Learning Approach to Family Intervention: Parent Training," Vol. I. Eugene, OR: Castalia Publishing Co.
Patterson GR, Stouthamer-Loeber M (In press):

267

The correlation of family management practices and delinquency. Child Development.


Popper KR (1972): "Objective Knowledge: An
Evaluationay Approach." London: Oxford
Press.
Reid JB (In press): Social-interactional patterns in
families of abused an nonabused children. In
Zahn-Waxler C, Cummings M, Radke-Yarrowm (eds): "Social and Biological Origins of
Altruism and Aggression. Cambridge, MA:
University Press.
Reid JB (1983a); Final report: Child abuse: Developmental factors and treatment. No. 7 ROl
MH 37938. National Insitute of Mental Health,
U.S. Public Health Service.
Rutter M, Tizard J, Whitmore K (1970): "Education, Health, and Behavior." New York: Wiley.
Sackett GP (1977): the lag sequential analysis of
contingency and cyclicity in behavioral interaction research. In Osofsky J (ed): "Handbook
of Infant Development." New York: Wiley.
Snyder, JJ (1977): Reinforcement analysis of interaction in problem and nonproblem families.
Journal of Abnormal Psychology 86:528-535.
Suomi S, Lamb M, Stephenson G (1979): "Social
Interaction Analysis: Methodological Issues."
Madison, WI: University of Wisconsin Press.
Taplin PS, Reid JB (1977): Changes in parental
consequation as a function of family intervention. Journal of Consulting and Clinical Psychology 4:973-981.
Walker HM (1979): "The Acting-out Child: Coping With Classroom Disruption." Boston: Allyn & Bacon, Inc.
Walter HI, Gilmore SK (1973): Placebo versus
social learning effects in parent training procedures designed to alter the behaviors of aggressive boys. Behavior Therapy 4:361-377.
Welsh RS (1976): Severe parental punishment and
delinquency: A developmental theory. Journal
of Clinical Child Psychology 5:17-21.

Вам также может понравиться