You are on page 1of 29

Psychological Bulletin Copyright 1992 by the American Psychological Association, Inc.

1992, Vol. 111, No. 1,127-155 0033-2909/92/$3.00

Externalizing Behavior Problems and Academic Underachievement


in Childhood and Adolescence: Causal Relationships
and Underlying Mechanisms
Stephen P. Hinshaw
University of California, Berkeley

Conceptual and measurement issues surrounding externalizing behavior problems and academic
underachievement, the strength and specificity of the covariation between these domains, and the
viability of explanatory models that link these areas are reviewed. In childhood, inattention and
hyperactivity are stronger correlates of academic problems than is aggression; by adolescence,
however, antisocial behavior and delinquency are clearly associated with underachievement.
Whereas investigations with designs that allow accurate causal inference are scarce, unidirectional
paths from 1 domain to the other have received little support. Indeed, the overlap of externalizing
problems with cognitive and readiness deficits early in development suggests the influence of
antecedent variables. Low socioeconomic status, family adversity, subaverage IQ, language deficits,
and neurodevelopmental delay are explored as possible underlying factors.

Links between academic underachievement and difficulties deserving of special education services (Professional Group for
in behavioral adjustment have long been noted (Sampson, ADD and Related Disorders, 1991).
1966; see also historical review of McGee, Share, Moffitt, Wil-
liams, & Silva, 1988). When several British epidemiologic in- Definition of Domains
vestigations in the 1960s and early 1970s yielded clear evidence
of overlap between reading deficits and behavioral problems of Externalizing Behavior
an acting-out or externalizing nature, interest in the phenome-
non was rekindled (see review of Rutter, 1974). This association Childhood behaviors marked by defiance, impulsivity,
continues to generate research and lively debate (e.g., McGee & disruptiveness, aggression, antisocial features, and overactivity
Share, 1988; Patterson, 1990; Rourke, 1988; Rutter, 1989; are called undercontrolled, or externalizing (see Achenbach &
Schonfeld, 1990; Silver, 1990) for several reasons: First, in terms Edelbrock, 1978). The distinctiveness of such features from be-
of prevalence rates, personal and societal suffering, and resis- havior patterns termed overcontrolled, or internalizingevi-
tance to most intervention strategies, both externalizing behav- denced by withdrawal, dysphoria, and anxietyhas been es-
ior problems and achievement difficulties constitute major tablished in numerous investigations (for a review, see Quay,
problems of childhood (Kazdin, 1987; Taylor, 1988). Second, 1986).1 Most notably, externalizing problems are more stable
each domain strongly predicts later maladjustment, in that ex- than internalizing behaviors, carrying (except in instances of
ternalizing problems often lead to antisocial behavior and sub- severe inhibition or depression) a worse prognosis as well as
stance abuse (Eron, 1987; Gittelman, Mannuzza, Shenker, & resistance to most forms of intervention (Robins, 1979).
Bonagura, 1985), and severe underachievement in reading not In child psychopathology, behavioral deviance can be char-
only persists but also carries a poor prognosis for other domains acterized in several ways. For example, to convey a dimension of
(Rutter & Yule, 1975; Spreen, 1988). Third, elucidation of un- behavior, investigators typically sum or average data gathered
derlying mechanisms may yield theoretical insights into behav- from rating scales or behavior observations, yielding a quantita-
ior-cognition links in both normal and atypical development, tive score. Through application of cutoff scores or multivariate-
an important tenet of the field of developmental psychopathol- clustering strategies, the same instruments may yield a category
ogy (Cicchetti, 1989). Fourth, the association has direct impli- that is, a subgroup of children with common characteristics.
cations for policy, as evidenced by recent efforts toward modify- When investigators use extensive clinical judgments, inclusion-
ing U.S. law to include attentional deficits as a distinct category ary markers, and clear exclusionary criteria, they obtain diag-
noses. Finally, alternative definitions, such as the legal entity of
juvenile delinquency, are sometimes invoked with respect to
Work on this article was supported by National Institute of Mental antisocial activities. This review includes reports that make use
Health Grant No. MH 45064.
I acknowledge the insightful comments of three anonymous re-
1
viewers, whose critiques facilitated the development of the arguments The domains are not completely independent, however. In fact,
herein. correlations between externalizing and internalizing behaviors are of-
Correspondence concerning this article should be addressed to Ste- ten moderate to high, especially when considerable psychopathology
phen P. Hinshaw, Department of Psychology, Tolman Hall, University exists and particularly in young children (Achenbach & Edelbrock,
of California, Berkeley, California 94720. 1983; Rose, Rose, & Feldman, 1989).

127
128 STEPHEN P. HINSHAW

of each of these means of defining externalizing behavior prob- Smallish, 1990; Weiss & Hechtman, 1986) and aggressive-con-
lems. duct-disordered behavior (e.g, Eron, 1987; Robins, 1970; Rut-
A major issue for the field is the validity of narrower dimen- ter, Tizard, Yule, Graham, & Whitmore, 1976). In all, given
sions or categories within the externalizing domain. Such valid- their prevalence, correlates, and persistence, externalizing be-
ity depends on the potential separability of dimensions or sub- havior problems constitute a major problem for society (Kaz-
groups, not only on the basis of defining criteria, butmore din, 1987).
importanton their degree of independence or divergent valid-
ity concerning such external variables as family history, patho- Academic Underachievement
physiology, course, and response to intervention. Although Below-average academic attainment can be classified in sev-
opinion was strong in the recent past that undercontrolled be- eral ways. In the first place, the achievement domain of interest
havior was unidimensional and that externalizing subtypes of must be specified. Because of the need for literacy in industri-
children were not distinct (e.g, Quay, 1979; Sandberg, Wiesel- alized societies, proficiency in reading is essential, and those
berg, & Shaffer, 1980), consensus has emerged that two major who investigate achievement difficulties focus nearly exclu-
types of externalizing behavior, inattention and hyperactivity sively on problems in reading decoding and reading compre-
on the one hand and aggression-conduct problems on the hension. Difficulties in spelling, particularly those of a pho-
other, show at least partial independence and some degree of netically inaccurate nature, are correlated strongly with reading
divergent validity (Hinshaw, 1987). Thus, despite their consider- deficiencies (Sweeney & Rourke, 1985); I do not consider them
able overlap (e.g., Offord, Alder, & Boyle, 1986; Szatmari, Boyle, separately. Although increasing interest has been shown in chil-
& Offord, 1989a), I consider these features separately in relation dren with arithmetic and mathematics disorders (e.g, Strang &
to academic underachievement.2 Rourke, 1985), I focus on deficient reading achievement herein.
Prevalence estimates for externalizing disorders vary with Next, the degree of underachievement and its relation to
the stringency of definitional criteria. For categories that are measured intelligence must be considered. In the classic study
based on quantitative instruments, cutoff scores of 1.5 or 2 of Rutter and Yule (1975), subaverage reading achievement was
standard deviations above the population mean are often used, subdivided into two categories: (a) reading achievement signifi-
yielding from about 2% to over 15% of the population, depend- cantly behind the level expected for the child's age (termed gen-
ing on the skewness of the distribution of scores. Because of the eral reading backwardness, or GRB) versus (b) reading achieve-
heterogeneity of categories defined solely on the basis of cutoff ment significantly behind the level predicted from intelligence
scores, their validity is often limited. Regarding formal diag- as well as age (specific reading retardation, or SRR). Both the
nostic criteria, the official psychiatric nosology in the US. nature of these definitional criteria and the sizable association
Diagnostic and Statistical Manual of Mental Disorders (DSM- between IQ and reading scores dictate that GRB children
III-R) lists several disruptive behavior disorders (American Psy- usually display somewhat subaverage intellectual performance
chiatric Association [APA], 1987). Conduct disorder, which in- (particularly in the verbal domain), which is invoked as the
volves persistent patterns of rule-breaking and violent behav- major explanation for their reading difficulties, whereas
ior, is estimated to have a prevalence of 9% for boys and 2% for youngsters with SRR typically have average or above-average
girls (see also Offord et al, 1986). Attention-deficit hyperactiv- IQ scores. Although considerable overlap between these groups
ity disorder (ADHD), signifying developmental^ inappropri- exists (i.e, nearly all children with SRR can also be classified as
ate levels of inattention, impulsivity, and overactivity, is believed GRB; about half of the youngsters with GRB also have specific
to have an overall prevalence of approximately 3%; boys out- reading deficits), meaningful comparisons can be made be-
number girls by a considerable margin, particularly in clinical tween "pure" GRB childrenthose without IQ-discrepant
samples (APA, 1987; see also Szatmari, Offord, & Boyle, 1989). reading scoresversus those with "specific" reading deficits.
Oppositional-defiant disorder, a controversial category that Rutter and \ule (1975) revealed a pattern of differential corre-
may constitute a precursor of conduct disorder, has an un- lates for these two groups. Compared with GRB children, chil-
known prevalence, in large measure because of its marginal dren with SRR overwhelmingly were boys, had fewer neurologi-
reliability and questionable validity (Key et al, 1988). Precise cal signs, and displayed a somewhat better outcome in arithme-
prevalence figures for these disorders await (a) investigations of tic but had a far worse prognosis in reading and spelling.
the validity of various definitions and (b) continued refinement Although replication efforts have not confirmed all such differ-
of sampling and instrumentation in epidemiologic studies. ences (e.g, Silva, McGee, & Williams, 1985), the separability of
The significance of externalizing behaviors is indicated by
their correlates and their persistent course. Attentional prob-
2
lems are associated with such variegated features as develop- Recent reports have explored the potential validity of further sub-
mental immaturity, language delays, and accidents (e.g., poison- divisions. For example, evidence exists that inattention and motoric
ing and bone fractures); conduct problems, on the other hand, overactivity constitute separable subdimensions of hyperactivity and
are correlated with both low family income and dysfunctional that antisocial behaviors may profitably be subdivided into those that
family systems (e.g., Offord et al, 1986; Szatmari, Offord, & are overtly aggressivefighting, defiance, tantrumsversus those
that are clandestine or covertfire setting, truancy, cheating, stealing
Boyle, 1989). Furthermore, both attention-disordered and con- (see Loeber & Lahey, 1989, for a review). In focusing on the two major
duct-disordered children have noteworthy difficulties in peer subdivisions, I use the terms attention deficits, attention-deficit hyperac-
relationships (Milich & Landau, 1989), and follow-up studies tivity disorder (ADHD), and hyperactivity interchangeably to refer to one
have documented the intransigence of both inattentive-hyper- type and aggression, antisocial behavior, and conduct disorder to de-
active behavior patterns (e.g, Barkley, Fischer, Edelbrock, & note the other dimension/category.
EXTERNALIZING BEHAVIOR AND UNDERACHIEVEMENT 129

these two groups is at least partially validated, mandating the emotional problems (see Rutter et al, 1970) dictates that I do
specification of IQ levels in children with reading difficulties not consider studies of children with clear mental retardation.
(see also Aman & Singh, 1983).3 Furthermore, to ensure that
measures of intellectual ability are not confounded with read- Conceptual, Developmental, and Measurement Issues
ing skill, individual (as opposed to group-administered) IQ
tests are necessary. Before reviewing relevant empirical research, I should high-
As with externalizing behavior, the prevalence of under- light several thorny issues regarding the domains under consid-
achievement varies with the stringency of the denning criteria. eration. Discussion of these conceptual and methodologic
With a relatively stringent cutoff of reading scores at least 28 points focuses attention on the many difficulties that arise in
months below age norms, the prevalence of GRB for 10-year- attempting to draw conclusions regarding causal precedence or
old boys was found to be approximately 7% in a rural setting explanatory mechanisms. In the first place, as noted above,
(Isle of Wight) and nearly 20% in inner London (Rutter et al., both domains under consideration are extremely heteroge-
1974). Regarding SRR, defined by a reading score at least 2 neous: Externalizing behavior problems comprise a wide vari-
standard errors of estimate below the level predicted from IQ, ety of constituent behaviors, and definitions of underachieve-
the respective rates were approximately 4% and 10%. Current ment also include a host of problems (e.g, subaverage IQ, spe-
American estimates for specific (i.e., IQ-discrepant) reading dis- cific achievement deficits, grade retention, poor marks in
abilities are similar, from 2% to 8% of the school-age population school). Because some of the components within each area (e.g,
(APA, 1987). Although boys are considered at higher risk to hyperactivity vs. aggression; specific vs. general reading defi-
display specific (but not necessarily general) reading delays, cits) have been validated as partially independent, global state-
recent data indicate that the preponderance of boys referred for ments concerning linkages between underachievement and ex-
learning difficulties may reflect biases in teacher identification ternalizing behavior should often be replaced by models involv-
procedures and in the greater amounts of problem behavior ing narrower variables. \fet much research in the field does not
displayed by reading disabled boys (McGee et al, 1988; Shay- make use of measurement strategies that can adequately distin-
witz, Shaywitz, Fletcher, & Escobar, 1990). guish among such subdivisions (see discussion of Loney & Mi-
The problems of children who display academic under- lich, 1982), limiting the specificity of causal models that can be
achievement are not limited to the academic domain. Indeed, formulated.5
self-esteem deficits, problems in language skills, and interper- Next, some of the components of each domain may be quite
sonal difficulties are common (e.g, Mann & Brady, 1988; Stone contaminated with aspects of the alternate domain. For exam-
& LaGreca, 1990). Also, as noted above, youngsters with IQ- ple, retention and poor marks may be as closely related to behav-
discrepant reading deficits tend to remain well behind their ior as to achievement deficits per se (e.g, Mantzicopoulos,
peers in reading skills during their educational careers (McGee Morrison, Hinshaw, & Carte, 1989). Thus, I focus primarily, but
et al., 1988; Rutter et al, 1976; see review of Spreen, 1988).
Thus, like externalizing behavior, underachievement in read- 3
Considerable controversy exists about the use of intelligence scores
ing has meaningful correlates and a persistent course.
as the criterion from which specific deficiencies in reading achieve-
Perhaps because of the rather inexplicable nature of the per- ment are judged. For example, intelligence does not not constitute a
sistent reading problems of children with SRRwho, with nor- fixed capacity for academic attainment, as attested by recent evidence
mal IQs and few neurological signs, would be expected to make that it does not place a rigid ceiling on reading levels (Share, McGee, &
adequate gainsconsiderable research interest has focused on Silva, 1989). Furthermore, the achievement problems of children with
this group. In the United States, the educational term learning marked reading deficits are real, regardless of their IQ scores. \et
disability and the psychiatric term specific developmental dis- knowledge of a child's intellectual level in relation to reading level may
order (APA, 1987) are used to describe the problems of these still be important.
4
children, who are the subjects of most investigations in the Indeed, the underlying deficits of this group often are presumed to
field.4 \fet I do not limit my coverage to specific (IQ-discrepant) be of a subtle neurological variety, given their adequate intelligence
and given the typical exclusion of such factors as poor schooling and
reading problems. In the first place, the majority of children
cultural deprivation in diagnosis. Yet such diagnostic labels as dys-
with achievement deficits do not display intelligence-attain- lexia, which connotes an underlying neurological deficit for reading
ment disparities (e.g, Rutter, Tizard, & Whitmore, 1970); thus, disabilities, have been severely criticized because of the heterogeneity
the exclusion of such problems as low grades, special class place- of children with severe reading deficits and the lack of a consistent
ment, or general reading problems could severely limit general- neurological or neuropsychological explanation for specific reading
izability. Second, the separation of specific from general read- deficiencies (e.g., Rourke, 1985; Rutter & Yule, 1975). I make neither
ing delay is often problematic, especially for young children: assumptions of homogeneity nor automatic etiological inferences
The categories overlap considerably and their distinctiveness is about either academic underachievement or externalizing behavior
unstable across development (e.g, McGee, Williams, Share, problems in this review.
5
Anderson, & Silva, 1986). Finally, because recent longitudinal Note that although components of a broadband construct may be
findings demonstrate provocative links between early verbal partially independent, these components may still coexist. For exam-
ple, regarding the externalizing domain, hyperactivity and aggression
deficits and externalizing behavior in adolescence (Schonfeld, often overlap (Hinshaw, 1987). Because achievement difficulties may
Shaffer, O'Connor, & Portnoy, 1988), low verbal intelligence pertain chiefly to those youngsters displaying both types of external-
may be an important domain in its own right. On the other izing difficulty (e.g., McGee, Williams, & Silva, 1984b), a focus on
hand, the relatively unequivocal causal link between globally narrowband conceptions or on differentiated subgroups should not
retarded intellectual performance and a host of behavioral and overlook the frequent comorbidity between them.
130 STEPHEN P. HINSHAW

not exclusively, on appraisal of underachievement by individual disorders vary when different sources are used to identify dis-
test administration. In addition, substantial debate centers ordered behavior (Offord, Boyle, & Racine, 1989). Thus, associa-
around whether inattention, considered heretofore as a type of tions with underachievement might pertain to externalizing
externalizing behavior, is actually a cognitive deficit (for debate, behavior as defined by one set of informants but not another.
see Barkley, 1989; Douglas, 1983; Sergeant, 1989). An associa- This latter scenario would be particularly likely, for example,
tion between inattention (even if measured by adult ratings) and when teacher ratings were used to define externalizing behav-
underachievement may thus signify a link within the general ior and when grades (given by teachers) indicated under-
realm of cognitive deficits rather than a behavior-cognition achievement: Shared method variance might account for a
correlation. good deal of the cross-domain linkage.
Third, manifestations of both externalizing behavior and For all of the reasons highlighted in this section, measure-
underachievement change markedly over time. For example, ment issues play a key role in explanatory accounts of overlap
the same youngsters who display delinquency in adolescence between externalizing behavior and underachievement. Inves-
and severe antisocial behavior in adulthood often show comor- tigators must use both developmental ly sensitive measures and
bid aggression and hyperactivity during the grade school years designs that afford inference of causal precedence, including
(Magnusson, 1988; Moffitt, 1990). Furthermore, many of these assessment of antecedent variables, if cross-domain links are to
same youngsters are quite likely to have displayed difficult tem- be elucidated fully. I therefore highlight factors related to meth-
peraments in infancy and marked oppositionality in the pre- odology in this review.
school era. In short, there may be a heterotypic continuity
across seemingly different aspects of externalizing behavior. Evidence for the Association
Similarly, many grade school children with reading deficits evi-
dence language delays during the preschool years, suggesting Primarily to discover the specific types of underachievement
that early linguistic deficits become commensurate with under- that are linked to subcategories of externalizing behavior, I now
achievement when academic curricula are introduced (e.g, review empirical evidence regarding the strength of the linkage
Mann & Brady, 1988). It follows that (a) measures must be tai- between domains. An initial question pertains to the degree of
lored to children's developmental levels and (b) the status of a covariation that would be meaningful or significant. Assuming
variable as a cause of problems in the counterpart domain (as an upper bound estimate of 10% for the domain of IQ-discrep-
opposed to its status as a consequence of the alternate domain) ant underachievement and a similar upper bound estimate of
may be confused unless careful attention is paid to developmen- 10% for noteworthy externalizing behavior problems, the de-
tally sensitive measurement. gree of overlap would be 1%the product of the independent
A hypothetical example may help to illustrate this last point. probabilitiesif the domains were associated at chance levels.
Suppose it is discovered that in a given sample, acting-out behav- In fact, however, some contend that the overlap between atten-
ior in the first gradenot accompanied by underachievement tional disorders and underachievement exceeds 50% (McGee &
predicts IQ-discrepant reading deficits 3 years later. One Share, 1988). The wide range of comorbidity estimates
might be tempted to conclude that a unidirectional causal link- currently in the literature mandates careful appraisal of the
age between early externalizing behavior and subsequent learn- evidence.
ing disabilities is supported. For one thing, however, it may be Because of the well-documented problems associated with
nearly impossible to find IQ-discrepant reading failure in first inferring both prevalence and comorbidity rates from clinical
graders, because children cannot be 1 to 2 years behind pre- samples, which are typically biased toward extreme pathology
dicted reading levels when they have been exposed to less than and toward frequent co-occurence of a wide range of associated
one grade of formal schooling. In other words, methods of de- problems (see Rutter, 1989), the preferred database would com-
fining underachievement in later years may be notably insensi- prise epidemiologic investigations of unselected populations.
tive at earlier phases of development. Furthermore, these hypo- Inspection of the major child epidemiologic studies from the
thetical first graders might well display delays (preexisting as 1970s and 1980s, however, reveals that very few studies of this
well as concurrent) in receptive or expressive language abilities, scope used individualized tests of achievement or intelligence,
delays that often translate directly into difficulties with the reflecting the need for most epidemiologic investigators to
phonologic processing necessary for proficiency in reading. trade depth of assessment for breadth of coverage (for a review
Such language deficits could actually be causative of early exter- of recent population surveys of child disorders, see Branden-
nalizing behavior and of subsequent underachievement. With- burg, Friedman, & Silver, 1990). Thus, I include only a handful
out assessment of antecedent variables and without knowledge of relevant epidemiologic reports, beginning with ground-
of developmental trajectories, erroneous causal precedence breaking work by Rutter and colleagues (see Rutter, 1974, for a
may well be inferred. review of earlier studies).
An additional issue pertains to the source of assessment data
regarding the domains under consideration, particularly exter-
Key Epidemiologic Investigations
nalizing behavior. Although adult informants provide the pri-
mary information regarding behavioral symptomatology, key Isle of Wight and inner London studies. The epidemiologic
informants (e.g., parents vs. teachers) show rather modest corre- study that reawakened the field to the association of interest
spondence regarding their appraisals (see Achenbach, McCon- was the Isle of Wight investigation of Rutter et al. (1970). This
aughy, & Howell, 1987). Furthermore, recent evidence suggests pioneering study made use of a two-stage procedure for assess-
that not only the prevalence but also the correlates of childhood ing both underachievement and psychiatric disturbance. First,
EXTERNALIZING BEHAVIOR AND UNDERACHIEVEMENT 131

the general population of 9- to 11-year-olds on this rural island by a particular scale can incorporate only those domains ade-
was screened for (a) intellectual and academic retardation, by quately represented in the item pool. As for the Rutter Scale B,
means of group achievement tests, and (b) psychiatric distur- only two general, broadband dimensions (antisocial vs. neu-
bance, by means of parent and teacher behavioral question- rotic behavior) were available; the few items pertaining to inat-
naires. Children considered at risk for intellectual-scholastic tention and overactivity did not form a separate scale with early
impairment, plus a randomly selected group of children from use of this instrument (Rutter, 1967).6 Refinement of rating
the general population, then received individual testing; those instruments and interview protocols has led to respecification
thought to manifest psychiatric disorder received semistruc- of the linkage between externalizing behavior and under-
tured parent-and-child interviews and additional adult ratings. achievement for preadolescents.
So as not to confound appraisal of the overlap between do- Waltham Forest report. In a replication of the Isle of Wight
mains, the cognitive and psychiatric evaluations were per- and inner London methodologies with a younger cohort, Rich-
formed independently. man, Stevenson, and Graham (1982) studied linkages between
Relevant to the topic of interest, children with SRR were cognitive-reading difficulties and behavior problems in a mid-
over four times more likely than the general population to dis- dle-class outer borough of London. Data collection began with
play antisocial behavior, whether the latter was appraised from over 800 three-year-olds, with follow-up at ages 4 and 8 for both
teacher ratings, parent ratings, or child interviews. This overlap behavior-problem and nonrisk subsamples. I focus here on age
was significantly larger than the association that would be ex- 8 data; preschool precursors of both behavioral disturbance
pected by chance. For example, using data from the teacher and learning problems are addressed in a subsequent section.
rating instrument, Rutter's Scale B (see Rutter, 1967), 24% of To avoid "empty sets" of reading-delayed children, Richman
the youngsters with SRR displayed above-cutoff levels of anti- et al. (1982) eased the criteria for defining GRB and SRR from
social behavior, compared with approximately 5% of the gen- 28 to 18 months below age- or IQ-expected performance. Re-
eral population. Furthermore, the reverse conditional probabili- garding cross-domain associations, children with behavioral
ties were similar, in that over a third of the youngsters meeting deviance (defined by parent and teacher ratings in addition to
criteria for antisocial behavior had SRR, a rate far higher than global clinical severity ratings) were more likely than compari-
would be expected by chance (see Rutter & Yule, 1970). This son youngsters to display underachievement, although the find-
epidemiologic investigation was replicated in an inner London ing fell short of statistical significance. In addition, for boys
borough (see Berger, Yule, & Rutter, 1975), with similar but not only, GRB was associated with neurotic (but not with antiso-
identical screening methods. The base rates of both SRR and cial) behavior. As noted by Richman et al., the relatively small
antisocial behavior were higher than on the Isle of Wight, and numbers of reading-delayed or behaviorally deviant children
the degree of overlap was also greater: Nearly half of the chil- may have reduced the power to ascertain overlap.
dren with SRR scored above cutoffs for antisocial behavior on Dunedin study. A more recent source of epidemiologic data
teacher ratings (see Sturge, 1982). is the Dunedin Multidisciplinary Child Development Study, a
As regards specificity, several findings are noteworthy. First, major prospective study of a birth cohort in New Zealand.
significant comorbidity was not found between SRR and classi- Throughout this continuing investigation, which has followed
fications of internalizing (neurotic) behavior on the Isle of the cohort during odd-numbered years of life beginning at age
Wight. That is, SRR children did not have high levels of neu- 3, the association between cognitive-achievement delays and
rotic behavior, and children classified as neurotic did not have externalizing behavior has been explored.
elevated levels of SRR, arguing for a specific link between SRR When the cohort was 7 years old, the Dunedin team per-
and acting-out behavior problems. Second, although children formed factor analyses with the Rutter teacher scale, yielding a
with GRB also had greater levels of problem behavior than did three-item Hyperactivity factor (composed of the items restless/
the general population, this association was not specific to anti- overactive, squirmy, and poor concentration/short attention
social behavior; GRB overlapped significantly with neurotic as span) that was separate from the usual externalizing dimension
well as with antisocial behavior. It was therefore concluded that of aggressive and antisocial behavior (McGee, Williams, et al.,
antisocial behavior correlates more strongly and specifically 1985; see also Schachar, Rutter, & Smith, 1981). Each of the
with achievement failure than with subaverage intelligence. three dimensions of deviant behaviorAggressive-Antisocial,
Although the formal classification of antisocial disturbance Hyperactive, and Anxious-Fearful (corresponding to neurotic
in this investigation was made from interviews as well as rat- or internalizing behavior)was examined for partial correla-
ings, the figures presented above arose from the use of cutoffs tions with Verbal and Performance IQ and with the Burt Read-
on teacher rating scales. The common use of such ratings to ing Test, controlling for the effects of the other two behavioral
ascertain externalizing behavior mandates brief comment. dimensions. The partial correlations of the aggression and in-
First, although such ratings constitute an economical and gener- ternalizing factor scores with the cognitive variables were es-
ally accepted mechanism for obtaining information about chil- sentially zero, whereas the Hyperactivity dimension correlated
dren's behavior, they are subject to distortions and bias (Saal,
Downey, & Lahey, 1980). They may be less sensitive to internal- 6
izing than to externalizing aspects of behavior, especially when Also, because diagnostic interviews used in the Isle of Wight study
relied on British conceptions of hyperkinesis as a severe syndrome of
teachers are informants (e.g., Aman & Singh, 1983; Hinshaw, pervasively overactive behavior, usually accompanied by mental retar-
Han, Erhardt, & Dressier, 1991); for adolescents, they may be dation, the more formal diagnoses yielded by the Isle of Wight team
less sensitive to acting-out behavior than are self-reports (Of- did not include categories congruent with current conceptions of at-
ford et al., 1986). Next, the dimensions and categories yielded tentional deficits or hyperactivity.
132 STEPHEN P. HINSHAW

significantly (and negatively) with Verbal IQ, Performance IQ, relatively high IQ scores. In addition, even among somewhat
and reading level, even when the other behavioral dimensions older children, the distinction between SRR and GRB across
were controlled. The main conclusion is thus that hyperactiv- the span of 9 to 11 years is relatively unstable (Share, McGee,
ity-inattention is specifically associated with below-average in- McKenzie, Williams, & Silva, 1987). Given the importance of
telligence and with underachievement in childhood, helping to younger age groups for discerning underlying mechanisms that
establish the validity of this domain as a subtype of externaliz- could mediate the overlap between externalizing behavior and
ing behavior (see Hinshaw, 1987). underachievement, these difficulties in accurately categorizing
In a closely related report, McGee, Williams, and Silva reading problems are noteworthy.
(1984b) used cutoff scores from the two externalizing factors to Overall, the Dunedin reports challenge the specificity of the
classify the 7-year-olds into hyperactive, aggressive, hyperac- linkage between antisocial behavior and SRR found in the Isle
tive-aggressive, or non-behaviorally disturbed categories. The of Wight and inner London studies. Here, the externalizing
two hyperactive groups displayed deficits in IQ and achieve- features of hyperactivity and inattention were associated with
ment, but the aggressive-only children typically did not differ both subaverage IQ and reading delays.7 Note, however, that
from the control population. \fet aggression played an interac- youngsters with the combination of hyperactivity and aggres-
tive role regarding specific reading delays: At both 7 and 9 sion were at greatly elevated risk for SRR in the Dunedin re-
years, the hyperactive-aggressive subgroup had a rate of SRR ports.
(36%) twice as high as rates for the single-disorder groups and Other epidemiologic investigations. Several other studies us-
five times that of the control children. Such results again point ing epidemiologic methods or population sampling merit brief
to the value of differentiating dimensions and categories within mention. First, Lambert and Sandoval (1980) used various psy-
the externalizing domain. chometric criteria to classify a community sample as either (a)
When the cohort had reached the age of 9, the Rutter parent hyperactivethat is, either home, school, or physician criteria
and teacher scales were supplemented with additional items of or (b) learning disabledthat is, various discrepancy formu-
inattentive, impulsive, and hyperactive behavior (McGee, Wil- las involving Wechsler Intelligence Scale for Children-Revised
liams, & Silva, 1985). Not only did a more differentiated factor (WISC-R) IQ and Peabody Individual Achievement Test
structure emerge, with inattention forming an independent di- (PIAT) achievement scores. Under lenient criteria, 53.5% of
mension, but also only this factor correlated with IQ, reading, hyperactive subjects were defined as learning disabled, but so
spelling, and speech measures (negatively in each case). Thus, were 20% of controls: a greatly inflated estimation. With more
an even finer distinction among types of externalizing behav- conservative definitions, lesser percentages of both hyperactive
ior was partially validated, in that inattention was specifically and control youngsters met criteria for learning disability (e.g.,
associated with achievement and IQ. 16% of hyperactive children and 11% of controls meeting three
At age 11, the cohort's externalizing and internalizing fea- of five strict criteria).
tures were appraised by means of formal diagnostic interviews Second, although lacking individual cognitive assessments,
with the children, which were supplemented by parent and the Ontario Child Health Study provided demographic data for
teacher reports to yield categories of attention-deficit disorder, Canadian 4- to 16-year-olds regarding both poor school perfor-
conduct-oppositional disorders, and internalizing disorders. mance, defined as the child's having failed a grade or needing
Only the children with attentional deficits (whether paired with full-time remedial education, and the use of special education
the conduct-oppositional or internalizing categories or alone) services (Szatmari, Boyle, & Offord, 1989). Whereas children
displayed IQ, reading, and spelling performance scores below with carefully defined attention deficits displayed poorer
those of the nondiagnosed and internalizing children (Ander- school performance and had higher rates of special education
son, Williams, McGee, & Silva, 1989). Also, 62% of the 45 usage than did nondiagnosed agemates, these rates were not
children classified as pervasively ADHDthat is, categorized significantly different from those of children with other dis-
on the basis of at least two independent sourceshad severe orders (a global category of conduct disorders and emotional
SRR, defined by scores on the Burt Reading Test that were disorders). Given that retention and special education place-
below the entire sample's average score at age 9 (see also McGee ment may reflect general behavioral difficulties as well as
& Share, 1988). Thus, the earlier findings relating attentional
problems to underachievement were replicated with a more
formal diagnostic approach; such cross-method replicability
7
adds to the viability of the findings. The British and New Zealand epidemiologic results may not be as
Dimensional IQ or reading scores were more often used in discrepant as they appear. Rutter and Graham (1970) divided the Rut-
the Dunedin reports than were the categories of SRR or GRB. ter Scale B items into informal clusters, one of which comprised four
Indeed, these investigations raise the issue of the viability of "motoric" items (restless/overactive, fidgety, twitches, and poor con-
categorical classifications of reading deficits, particularly for centration/short attention span). Inspection of Table 7.3 (Rutter &
young children. First, as noted earlier, the typical requirement Graham, 1970, p. 109) reveals that for children with specific reading
retardation (SRR), teachers endorsed these items much more fre-
of rather severe (28-month) disparities between reading level quently than any from the antisocial cluster; indeed, the endorsement
and either intelligence or age is quite restrictive for young chil- of poor concentration was nearly 85% for SRR youngsters, almost
dren. Indeed, it is nearly impossible for children with even double the rate for any other item on the entire scale. Yet because such
slightly subaverage IQ scores to obtain reading scores lower items were not consolidated into a formal dimension with early use of
than 1.5 or 2 standard errors of estimate below prediction, leav- the Rutter scales, inattention-hyperactivity could not be indepen-
ing the SRR category viable only for those young children with dently examined for covariation with reading delays.
EXTERNALIZING BEHAVIOR AND UNDERACHIEVEMENT 133

achievement problems per se, the lack of divergent validity Clinical Reports
among clinical disorders is not surprising.
Although the nonrepresentativeness of clinical samples is po-
In an Australian epidemiologic report, Holborow and Berry
tentially problematic for inferring estimates of comorbidity,
(1986) used the Conners Abbreviated Symptom Questionnaire
most of the evidence for overlap between the domains of inter-
(Goyette, Conners, & Ulrich, 1978) to measure hyperactivity
est comes from such sources. Given increasing sophistication
and used four additional teacher-completed items to determine
in sampling and instrumentation, I highlight key clinical re-
the prevalence of learning difficulties. Among children meet-
ports from the last dozen years. Because of the volume of such
ing the typical cutoff score for hyperactivity on the Conners,
studies, I make no claims for exhaustive coverage.
26.5% met rating criteria for learning difficulties, compared
Links between externalizing disorders and underachievement
with 5% of the remainder of the sample. Reversing the condi-
in elementary grades. Most clinical reports that examine over-
tional probabilities, over 40% of the children rated as having
lap between externalizing problems and underachievement be-
learning problems were also considered hyperactive. The chief
fore adolescence focus on hyperactivity-attention deficits
limitation of this report is the ad hoc, teacher-rated definition
rather than aggression or conduct disorder. In an early report
of learning difficulties, which contributes to shared method
that made use of an IQ-achievement discrepancy formula for
variance for defining both underachievement and externaliz-
ascertaining specific achievement deficits, Cantwell and Satter-
ing problems.
field (1978) found that 39% of a clinical sample of hyperactive
A salient issue is the age at which associations between behav-
children (vs. 9% of controls) were two or more grades behind
ioral problems and achievement difficulties first emerge. Hin-
their predicted grade level in reading. The same percentage of
shaw, Morrison, Carte, and Cornsweet (1987) performed a popu-
overlap (39%) was found by August and Garfinkel (1990), who
lation study of kindergarten children in several suburban
used a DSM-III-R definition of ADHD and a non-regression-
school districts, with a participation rate of over 80%. Parent
based discrepancy formulaWide Range Achievement Test-
and teacher ratings from the Revised Behavior Problem Check-
Revised (WRAT-R) scores 1 standard deviation below Peabody
list (RBPC; Quay, 1983) were examined for associations with
Picture Vocabulary Test-Revised (PPVT-R) IQ scoresto de-
group tests of academic readiness and individual intellectual
fine reading disability (see also August & Garfinkel, 1989).
assessments. For teachers, whereas the Conduct Disorder factor
On the other hand, among a clinical sample of 241 hyperac-
displayed nearly zero correlations with cognitive and early
tive children, defined by cutoffs on the Conners Teacher Rating
achievement indexes, the Attention Problem and several inter-
Scale and clinical interviews with parents, only 22 (9%) met
nalizing scales showed negative correlations ranging from -.25
criteria for specific reading disability, which was indicated by a
to .35. No significant correlations were noted for parent-rated
1 standard deviation disparity between WRAT Reading and
factors, however, suggesting that situational specificity may ex-
WISC IQ (Halperin, Gittelman, Klein, & Rudel, 1984). Further-
ist for the correlates of behavioral problems (see also Szatmari,
more, in a recent investigation with rigorously defined samples
Offord, Siegel, Finlayson, & Tuff, 1990).
of ADHD and ADHD-aggressive boys, Forness, Youpa,
In a companion report with additional cohorts, Morrison,
Hanna, Cantwell, and Swanson (in press) found that only 6%
Mantzicopoulos, and Carte (1989) found that children at risk
for learning disabilitiescategorized on the basis of low per- met strict criteria for reading disability (i.e., a reading score at
ceptual and prereading skillswere rated as behaviorally de- least 1.5 standard deviations of the difference below IQ), with
no difference in prevalence of reading problems between the
viant across all RBPC teacher dimensions. Yet partial correla-
subgroups. With a slightly more lenient criterion of a 1 standard
tions (not calculated in published study) are noteworthy: The
deviation discrepancy, 10% (7 of the 71 boys) were reading dis-
correlation between attention problems and prereading achieve-
ment, partialing conduct problems, was a significant .25, but abled.8 In short, unless one uses estimates of underachievement
the correlation of conduct problems with prereading, partial- that also classify overly large numbers of comparison children
ing attention problems, was .00. Thus, even before children as underachieving (Lambert & Sandoval, 1980), the contention
experience a formal academic curriculum, hyperactivity is as- of McGee and Share (1988) that hyperactivity-attentional defi-
sociated with readiness problems. cits overlap with learning disabilities at rates above 50% does
Summary. The specific link between antisocial behavior not seem warranted.
and IQ-discrepant reading deficits from the seminal reports of Although these and other clinical reports have yielded
Rutter et al. (1970) and Berger et al. (1975) was not replicated in meager evidence for the differential relationship of hyperactiv-
the Dunedin studies, in which attentional deficits and hyperac- ity versus aggression with underachievement (e.g., McCon-
tivity comprised the externalizing scales or categories that were aughy, Achenbach, & Gent, 1988; Reeves, Werry, Elkind, & Za-
associated with both subaverage IQ and reading delay. A possi- metkin, 1987), the recent investigation of Frick et al. (1991) is
ble explanation is that the original Rutter Scale B used in En-
8
gland did not yield a separate inattention-hyperactivity factor; Because children with attentional disorders often score below
only through addition of pertinent items and subsequent factor norms on an empirically derived cluster of Wechsler Intelligence Scale
analyses did divergently valid dimensions (and distinct sub- for Children-Revised items (Arithmetic, Coding, and Digit Span; see
Kavale & Forness, 1984), their full-scale IQ scores may be somewhat
groups) of aggression versus hyperactivity emerge. Other epide-
depressed, meaning that they must attain extremely low achievement
miologic results essentially confirm the finding that among scores to meet stringent criteria for learning disability (e.g., a 1.5 stan-
kindergarten and grade school children, inattention and hyper- dard deviation discrepancy from IQ). Thus, a 1 standard deviation
activity are the most consistent correlates of underachieve- disparity is considered more valid by some investigators (Forness,
ment. Youpa, Hanna, Cantwell, & Swanson, in press).
134 STEPHEN P. HINSHAW

noteworthy because of its exemplary methodology. Here, exter- initial group differences remained. For instance, once equated
nalizing children were categorized into ADHD, conduct dis- for PPVT scores, the three clinical groups did not differ in
order, or clinic control diagnoses on the basis of combined reading achievement, arguing against a higher rate of specific
parent, child, and teacher interview data from the Diagnostic reading problems in the externalizing categories. Furthermore,
Interview Schedule for Children. Underachievement was deter- only the combined hyperactive plus conduct-disordered group
mined on the basis of regression-formula discrepancies be- differed significantly in reading from the comparison subjects
tween individual IQ and achievement scores that also con- when sex and PPVT were controlled. The conclusion was that
trolled for the age of the child. Depending on the stringency of the major deficit of youngsters with externalizing problems
the denning formula for underachievement, from about 6% to (particularly attention-deficit disorders) is in the area of verbal
20% of either conduct-disordered or attention-disordered chil- abilities. Indeed, long-standing verbal deficits could potentially
dren showed underachievement, reflecting a modest degree of be a powerful third variable, mediating both externalizing be-
association between each domain and learning problems. "Vet havior and underachievement, a point addressed subsequently.
multivariate logit analyses that controlled for the co-occurrence Second, Szatmari et al. (1990) measured neurocognitive
of ADHD and conduct disorder revealed that only attention- functioning in a clinic sample of externalizing and internaliz-
disorder status was uniquely associated with underachieve- ing youngsters. Their battery included measures of visual-per-
ment. Indeed, the apparent overlap between conduct disorder ceptual processing and problem solving. Carefully diagnosed
and underachievement was related to the presence of atten- groups of attention-disordered versus conduct-disordered chil-
tional problems in many conduct-disordered children. Thus, dren did not differ regarding rates of neurocognitive deficits,
mirroring the epidemiologic results of the Dunedin investiga- whereas both categories displayed more problems (primarily of
tion, Frick et al. (1991) concluded that attentional difficulties a visual-perceptual type) than did the internalizing young-
constitute the externalizing domain that is uniquely associated sters. When neurocognitive scores were used to predict dimen-
with underachievement during childhood. sional teacher ratings of problem behavior, teacher-rated school
To summarize, although rates of overlap between externaliz- performance mediated the relationship, suggesting that neuro-
ing disorders and psychometrically defined learning disabili- cognitive impairments may lead to disturbed behavior by
ties are above chance levels, recent investigations with rigorous means of underachievement. I discuss further the possible role
formulas for defining underachievement have yielded comor- of perceptual-neurodevelopmental delay in predicting both ex-
bidity estimates that are lower than earlier figures. These rates ternalizing behavior and underachievement in a later section.
are still higher, however, than those reported for internalizing Links between juvenile delinquency and cognitive-achieve-
youngsters. In addition, children with externalizing disorders ment problems. Because of interest in tracking the overlap
often have achievement-related difficulties that may not be between domains into adolescence, I briefly summarize find-
classifiable as formal learning disabilities, including retention ings regarding the cognitive-achievement status of delinquent
and school suspension (Barkley et al., 1990; Forness et al., in youngsters. The volume of empirical reports on links between
press). Furthermore, whereas many initial clinical reports did delinquency and underachievement requires that I rely on sev-
not find differences between hyperactive and aggressive chil- eral major reviews (Hawkins & Lishner, 1987; Hirschi & Hinde-
dren in rates of underachievement, the most methodologically lang, 1977; Loeber & Stouthamer-Loeber, 1987; Quay, 1987;
sophisticated investigation in the field (Frick et al., 1991) indi- Rutter&Giller,1983).
cates that hyperactivity-inattention is the externalizing domain First, delinquent adolescents have lower IQ scores than do
most clearly associated with academic failure during the child- their peers, particularly in the verbal domain. Although such
hood years (for cogent discussion and provocative longitudinal intelligence deficits are not typically sufficient to place delin-
findings, see Loney, Kramer, & Milich, 1981). Furthermore, quents in a mentally retarded range, they are robust (on the
the academic status of hyperactive children continues to be order of half a standard deviation) and are particularly evident
compromised during adolescent follow-up (Fischer, Barkley, in subgroups of delinquent youngsters with aggressive-psycho-
Edelbrock, & Smallish, 1990). pathic features (Quay, 1987). Such IQ deficits are not explain-
Links between externalizing disorders and cognitive-verbal able by socioeconomic status (SES) or racial differences be-
deficits. Although the focus of this review is on underachieve- tween delinquent and comparison groups (Hirschi & Hinde-
ment, more basic cognitive deficits might also be associated lang, 1977; Wilson & Herrnstein, 1985). Furthermore, from the
with externalizing behavior problems. Whereas space limita- reverse perspective, high IQ scores serve as a protective factor
tions preclude a comprehensive review of this issue, I consider against delinquent outcomes for aggressive boys (White, Mof-
two recent reports that address the link between verbal, cogni- fitt, & Silva, 1989). Note also that recent reports challenge the
tive, and perceptual deficits with externalizing behavior. differential detection hypothesis, which contends that it is not
First, with the intention of distinguishing among clinical delinquency per se but rather detection and incarceration that
diagnostic groups regarding mechanisms of psychopathology, are associated with lowered intelligence. Notably, Moffitt and
Werry, Elkind, and Reeves (1987) examined hyperactive, hyper- Silva (1988a) found that both adjudicated delinquents and a
active plus conduct-disordered, anxiety-disordered, and com- severity-matched group of self-reported delinquents had compa-
parison subjects on a host of laboratory measures of cognition, rably depressed IQ scores.
attention, motor coordination, impulsivity, and reading recogni- Second, delinquent adolescents have subaverage academic
tion (see also Reeves et al., 1987). The most striking finding was achievement, including an elevated rate of learning disabilities.
that when age, sex, and PPVT scores were partialed from the Indeed, the link between academic failure and delinquency is
analyses that compared categories, almost none of the many claimed to be stronger than the association between lowered
EXTERNALIZING BEHAVIOR AND UNDERACHIEVEMENT 135

verbal IQ and delinquency (Hawkins & Lishner, 1987). Further- school attachment, consequences of poor achievement that
more, among several skill deficits, achievement difficulties may mediate subsequent antisocial activity (e.g., Hirschi, 1969).
have the strongest concurrent associations with official and 2. Externalizing behavior leads to underachievement. Here,
self-reported delinquency (Dishion, Loeber, Stouthamer- behavioral disturbance predating school entry or appearing
Loeber, & Patterson, 1984); poor academic performance by the during early schooling would be viewed as primary; its interfer-
end of elementary school significantly enhances the prediction ence with proper classroom behavior might be the key media-
of adolescent delinquency (Loeber & Dishion, 1983). tor of underachievement. For this model to be viable, the early
These findings suggest that early underachievement is externalizing features should predict subsequent underachieve-
causally related to subsequent antisocial activity. Yet because ment independently of poor readiness skills, which might ac-
underachievement and hyperactivity are often linked during company the behavioral features.
grade school and because hyperactivity predicts later delin- 3. Both domains lead to the other. This bidirectional model
quency (Gittelman et al, 1985), support for a direct path from acknowledges that both of the previous unidirectional models
low achievement to delinquent behavior is weakened (see occur simultaneously (see Olweus, 1983).
Loeber, 1990, for methodologic discussion). On the other hand, 4. Underlying variables result in both problem domains.
delinquency rates decline when adolescents are not in school Such antecedent variables could be intraindividual (e.g, temper-
for example, during school vacations or after school dropout ament, language difficulties) or environmental (e.g., discordant
(see review of Phillips & Kelly, 1979). Furthermore, as is dis- homes, large family size). Because this model requires that they
cussed subsequently, some reading disabled children without causally precede the association, preliminary evidence for third
externalizing features develop delinquency by late adolescence variables would include the joint presence of externalizing be-
(Maughan, Gray, & Rutter, 1985). Both of these findings sug- haviors and cognitive difficulties in early years. More compre-
gest that at least in some instances, school failure may predis- hensive investigations would require prospective, longitudinal
pose to acting-out behavior. As is evident, causal paths between evaluations that include sensitive measures of the hypothesized
domains may be variegated and complex. causal variables and their statistical control in analyses of ex-
To summarize, delinquency is associated with low verbal IQ planatory factors.
and both general and specific achievement problems, but The heterogeneity of the domains under review and the prob-
achievement difficulties are more strongly predictive of delin- able complexity of the links between them may render linear
quent behavior than is low intellectual ability per se. Thus, by causal models, even those with bidirectional implications,
adolescence, there is a clear link between aggressive-antisocial overly simplistic. Indeed, investigators of adult psychopathol-
acts and underachievement, whereas during childhood the ogy are increasingly cognizant of the need to seek reciprocally
more specific relationship pertains to hyperactivity-inatten- deterministic, multifactorial models of causation (Ohman &
tion. Such findings mandate examination of the causal nature Magnusson, 1987). Because of the rapid development of chil-
of this complex association across development, the topic area dren and because of the plethora of individual, familial, and
to which I now turn. school variables that could enter into causal equations, such
complex models may be especially pertinent for developmental
psychopathology.
Causal Pathways and Underlying Mechanisms Much of the early literature regarding the current topic
Hypothetical Causal Models comes from cross-sectional or retrospective research designs.
For example, Rutter and Yule (1970) compared three subgroups
When two variables are associated, it is often assumed that of boys from the Isle of Wight investigation: (a) antisocial only
the first caused the second or that the second caused the first. (ASB), (b) SRR only, and (c) jointly ASB and SRR (see discus-
Within such unidirectional models, additional factors can be sion in Yule & Rutter, 1985). Because the latter, mixed subgroup
hypothesized to mediate paths from cause to effect, yielding had a later onset of antisocial tendencies than did the exclu-
both direct and indirect causal linkages. Even simple causal sively antisocial boys and because they resembled the SRR-only
models can therefore quickly increase in complexity. Also, each youngsters regarding key background variables, Rutter and
factor might cause the other, suggesting bidirectional influ- Yule (1970) concluded that either antisocial behavior results
ence. Furthermore, a third factor (or set of factors) may cause from reading deficits or that prior variables must underlie the
both variables of interest, or distinctbut correlatedback- association. In similar analyses with the inner-London sample,
ground factors may cause each domain, leading to a largely however, Sturge (1982) did not find such a clear-cut pattern of
spurious association between the main variables (Olweus, 1983; results. Because these designs have clear limitations in detect-
Sturge, 1982). ing causal relationships, I turn to prospective, longitudinal stud-
In basic terms, several causal models might explain the co- ies, which are potentially more informative about causal infer-
variation between externalizing behavior and underachieve- ences (Loeber, 1990).
ment (see Huesmann, Eron, & Yarmel, 1987; Olweus, 1983, for
cogent discussion):
1. Underachievement leads to externalizing behavior. This Prospective, Longitudinal Investigations
model requires a history of learning failure that precedes (or First, because of the aforementioned trend for inattention-
exacerbates) the emergence of externalizing features. This hyperactivity to be correlated with underachievement during
causal relationship might include such additional variables as grade school and for antisocial behavior-delinquency to show
frustration, lowered self-image, demoralization, or lack of association with academic failure in adolescence, I subgroup
136 STEPHEN P. HINSHAW

the following reports on the basis of the age range (elementary more, in Palfrey, Levine, Walker, and Sullivan (1985), early
vs. secondary grades) in which follow-up data were collected. measures of inattention correlated with poor reading readiness
Next, I consider only those reports with an interval of at least 1 skills in kindergarten. Finally, both Jorm, Share, Matthews,
year between initial and follow-up assessment periods, seem- and Maclean (1986) and McGee et al. (1986), who made classifi-
ingly a minimum time period for appraising the relationships cations of GRB or SRR in grade school, showed externalizing
of interest. Third, whereas consensus is growing that meta-ana- behavior by one or both reading-delayed groups in kindergar-
lytic methods are the preferred means for amalgamating find- ten, before exposure to academic curricula.
ings across independent investigations, the extreme differences Such early associations between the domains of interest
in measures, methods, and analytic plans across the relatively strongly suggest the influence of prior antecedent variables.
few studies to be reviewed preclude their meaningful combina- Furthermore, these associations mandate statistical control
tion for meta-analysis. Fourth, I limit my discussion to reports when investigators attempt to establish causal precedence. That
that have appeared since the review of Rutter (1974). is, without such techniques as multiple regression, which allow
To yield data that would shed light on causal relationships, for examination of the unique contribution of a predictor vari-
investigations should include (a) assessments of both achieve- able once its correlates are partialed, the status of a predictor as
ment-related and behavioral variables at initial as well as follow- an independent cause is suspect. Unfortunately, several reports
up periods, (b) measures of relevant antecedent variables, and in Table 1 failed to make use of such analyses but still went on to
(c) statistical analyses that allow for causal inference, through posit a unidirectional influence from one domain to the other.
control for the effects of antecedent factors or correlated predic- In those reports in which such analyses were performed, it was
tors. Extremely few reports in the literature include all such found in one instance that internalizing rather than externaliz-
features; these receive the strongest weight in my discussion of ing behavior uniquely predicted reading (Lambert & Nicoll,
causal mechanisms. Nevertheless, because of their heuristic 1977) and in another that early readiness skills predicted subse-
value, I also include additional studies without such fully infor- quent underachievement more strongly than did behavioral
mative designs, realizing their limitations regarding explana- measures (McMichael, 1979; see also Jorm, Share, Maclean, &
tions of causal mechanisms. Matthews, 1986). In short, associations in elementary school
Elementary grades. Table 1 contains details of the investiga- between underachievement and externalizing behavior are typi-
tions in which follow-up evaluations were made during elemen- cally predated by correlated precursors, mandating a search for
tary school. All reports used community samples of moderate- underlying variables that may be responsible for the early associ-
to-large size, a necessary feature to obtain enough children with ation and requiring adequate statistical controls in making
either deviant behavior patterns or underachievement. Even so, causal inference. When such controls were performed in the
those studies with categorical definitions of either poor aca- studies reviewed, support for unidirectional causation was
demic performance or behavior problems often had a relatively sharply mitigated.
small number of children who were classified (e.g., McGee et al, Several additional conclusions are suggested from Table 1. In
1986). Examination of the table reveals a diversity of methods the first place, despite the caveats just raised, some suggestive
for assessing cognitive deficits during initial assessment periods evidence exists that the experience of reading failure may exac-
and for evaluating underachievement during follow-up periods; erbate initial externalizing behavior. For example, Jorm, Share,
dimensional as well as categorical indexes are used in many Matthews, and Maclean (1986) found that whereas hyperac-
reports. Indeed, only Richman et al. (1982), Jorm, Share, tive-inattentive behavior was elevated at school entry for chil-
Matthews, and Maclean (1986), and McGee et al. (1986) used dren later classified as GRB, antisocial behavior in these
categories of GRB and SRR, partly because of the aforemen- youngsters increased after first and second grade. Also, the
tioned difficulties in obtaining stable classifications of these careful analyses of McGee et al. (1986) demonstrated that al-
categories during early and middle childhood. Teacher ratings though teacher-rated externalizing behavior was elevated at
predominate as the chief means of assessing behavior. school entry for boys later categorized as reading delayed, boys
Perhaps the most salient overall finding from these reports is with GRB showed a relative increase in hyperactivity from age 5
that in all six instances in which relevant analyses were per- to age 7, and boys with SRR displayed a relative increase in
formed, patterns of problem behavior during the initial assess- hyperactivity from age 7 to age 9.
ment were associated with early cognitive-readiness deficien- Second, at least in the preschool and early elementary years,
cies (Lambert & Nicoll, 1977, did not perform relevant analy- links between behavior problems and underachievement are
ses; Stott, 1981, did not include measures of cognitive not always specific to the externalizing domain. For instance,
functioning during the initial assessment). Specifically, Kellam, in Richman et al. (1982), behavior-problem status at 3 years was
Branch, Agrawal, and Ensminger (1975) found that teacher rat- relatively nonspecific: No consistent internalizing or external-
ings of internalizing and externalizing items, which predicted izing dimensions were yielded from the rating scale and inter-
lowered IQ and poor reading status 2 years later, were asso- view that were used. This generic behavior-problem category
ciated with low readiness performance in first grade. Next, was predictive of GRB 5 years later. Also, in Kellam et al.
antisocial behavior in kindergarten, which predicted poor (1975), the associations between each of six diverse behavioral
reading 1.5 years later in the report of McMichael (1979), was items and either concurrent or subsequent achievement status
also correlated with deficient readiness skills during the initial were of nearly the same magnitude (see Tables 6 and 8; see also
assessmenU-'Such associations emerged even earlier in the re- Stott, 1981, for similar findings). In addition, Lambert and Ni-
port of Richman et al. (1982): At 3 years, general behavioral coll (1977) found that internalizing rather than externalizing
deviance overlapped with cognitive and language delay. Further- ratings made independent predictions of later reading scores.
EXTERNALIZING BEHAVIOR AND UNDERACHIEVEMENT 137

On the other hand, the investigations of Jorm, Share, Mat- third-variable models, Olweus (1983) obtained peer assess-
thews, and Maclean (1986) and McGee et al. (1986) revealed ments of aggression as well as averaged school grades at both
that teacher-rated dimensions or categories of externalizing 6th and 9th grades in a Swedish sample of boys; he also ob-
(hyperactive, aggressive) problems were correlated with delayed tained measures of a number of familial and environmental
reading status but that internalizing behaviors were not. Al- background variables. Cross-lagged correlations provided no
though those reports with better-validated measures of prob- evidence for the low-grades-predict-aggression unidirectional
lem behavior tend to converge on the primacy of externalizing model, but a near-significant effect was found for the aggres-
behavior in relation to underachievement, complete specificity sion-predicts-low-grades path. However, given critiques of
cannot be presumed. cross-lagged correlations as an analytic tool, Olweus also per-
Third, the contention that behavior problems are associated formed structural equation modeling. Once the background
specifically with IQ-discrepant reading delay (SRR; see Berger variables of social class, parents' ages, divorce, birth out of wed-
et al., 1975; Rutter et al., 1970) is not supported by these investi- lock, and birth order were entered into the causal models, no
gations. Indeed, in many of the reports, indexes of behavioral unidirectional or reciprocal causation paths remained signifi-
maladjustment were correlated with either IQ measures or with cant, leading to the conclusion that antecedent variables ac-
reading scores/categories that were not adjusted for intelli- counted for the association between poor grades and aggressive
gence. Also, in direct opposition to the early British investiga- status. As mentioned earlier, however, poor grades constitute
tions, those studies with categorizations of GRB and SRR an index of underachievement that may well be directly con-
(Jorm, Share, Matthews, & Maclean, 1986; McGee et al., 1986; taminated by externalizing behavior.
Richman et al, 1982) typically indicated that behavior prob- Second, the report of Schonfeld et al. (1988), which measured
lems were related more strongly to the former than to the latter relevant background variables and used path analyses, pro-
category. Thus, little support exists for a preferential association vided some evidence for a link between cognitive deficits in
of specific, IQ-discrepant reading deficits to behavioral prob- childhood and conduct disorder 10 years later. Here, an indi-
lems. rect path between age 7 WISC Verbal IQ and severity ratings of
Fourth, the magnitude of association or prediction across conduct disorder at age 17mediated by adolescent Wechsler
domains is typically modest to moderate. One perspective on Adult Intelligence Scale (WAIS) IQ scoresheld up even when
this issue is provided by McMichael (1979), who found that (a) early aggressive behavior and (b) the background variables of
although kindergarteners with antisocial problems tended to environmental disadvantage, neurological soft signs, and paren-
develop reading problems by the end of the first grade, over half tal psychopathology were controlled. Yet measurement issues
of these youngsters failed to show subsequent underachieve- cloud the viability of these findings, in that initial aggression
ment. Thus, the relationship of interest may pertain to only a was indexed by behavior ratings from the psychologists who
subgroup of children with externalizing behavior.9 performed the cognitive testing. Such individualized appraisals
A final point is that in addition to failing to control for corre- of externalizing behavior in the laboratory, doctor's office, or
lated measures at preassessment, most reports did not analyze testing room are not sensitive indicators (see, for example, Slea-
for bidirectional relationships. That is, few investigators exam- tor & Ullmann, 1981). Because more valid indexes might have
ined whether both early behavior-problem status predicted sub- altered the paths, the unidirectional link of Schonfeld et al.
sequent underachievement and whether early cognitive-readi- from verbal deficits to conduct disorder should be viewed
ness problems predicted later misbehavior. Unless both predic- cautiously.
tive relationships are examined, however, the full set of causal Several additional investigations, without such stringent anal-
possibilities cannot be adequately appraised. yses, provided apparent support for the unidirectional model
Secondary grades. Table 2 contains highlights of investiga- that underachievement leads to externalizing behavior. For ex-
tions that examined follow-up status during adolescence. Note ample, Farrington (1979) showed that both low IQ and low
that none of these reports made use of categories of SRR or vocabulary status at 8-10 years of age predicted self-reported
GRB to define underachievement during the follow-up assess- and official delinquency 6 years later. Because, however,
ment; instead, investigators used such indexes as school grades, teacher ratings of externalizing behavior at preassessment also
special classroom placement, early school withdrawal, and la- predicted delinquent outcomes, and because analyses control-
tent variables measured by multiple indicators. The aforemen- ling for behavior measures were not performed, the causal role
tioned caveats about the overlap of such variables with behavior (text continues on page 140)
per se pertain here (see also Shaywitz et al, 1990). Externalizing
behavior was also assessed in diverse ways: Peer sociometric 9
Another issue here has to do with the antisocial nature of the chil-
evaluations, diagnostic interviews, examiner appraisal of ag- dren studied by McMichael (1979), who used Rutter's (1967) method
gressive behavior during testing sessions, and self-reports of for classifying behavior-problem subgroups. In this procedure, an
delinquency were used in addition to the more typical teacher overall behavior-problem score on the total scale is used to classify a
ratings. child as deviant; antisocial versus neurotic categorizations are then
Two of the investigations depicted in Table 2 included (a) made on the basis of simple numeric comparisons. Thus, if a child
measures of each domain at multiple time points, (b) assess- received many endorsements for items pertaining to attention deficit-
hyperactivity, he or she could still be classified as antisocial if the anti-
ments of relevant antecedent variables, and (c) structural equa- social score were just 1 point higher than the neurotic score (e.g, 1 vs. 0;
tion modeling to analyze datafeatures that allow more confi- see Jorm, Share, Matthews, & Maclean (1986). Thus, McMichael's find-
dence in causal inferences. First, in an explicit attempt to estab- ings might pertain more to hyperactivity than to antisocial tendencies
lish the credibility of unidirectional versus reciprocal versus per se.
138 STEPHEN P. HINSHAW

Table 1
Longitudinal Investigations with Follow-Up Assessments in Elementary Grades
Initial assessment
Cognitive measure Behavioral measure
Grade
Study Location N (age) Content Scale type Content Scale type
Kellam, Branch, Chicago 649 1st IQ (Kuhlman- Categorical Teacher ratings on Categorical
Agrawal, & Anderson) and 6 items,
Ensminger readiness including
(1975) (Metropolitan) concentration
tests and authority
acceptance
(externalizing),
social contact
(internalizing)
Lambert & Nicoll San Francisco b 1st Readiness battery Dimensional Teacher ratings on Dimensional
(1977) Bay Area (Let's Look at items including
Children) interpersonal
(externalizing)
and
intrapersonal
adjustment

McMichael Edinburgh, 198 Kc Readiness battery Dimensional, Teacher ratings on Dimensional,


(1979) Scotland (boys only) (Thackray) categorical Rutter Scale B categorical

Stott(1981) Guelph, 1,292 K Teacher ratings on Categorical


Ontario, 6 items: timid,
Canada distant,
lethargic
(internalizing);
hyperactive,
impulsive,
hostile
(externalizing)
Richman, Waltham 185r 3yr Language Dimensional and Parent ratings and Categorical
Stevenson, & Forest, development categorical semistructured
Graham (1982) England (Reynell), parent 8
picture interview
vocabulary,
general ability
test (Griffith)

Palfrey, Levine, Greater 285 2 wk-5 yr' Varied Parent ratings, Categorical
Walker, & Boston depending on teacher ratings,
Sullivan (1985) child's age1 psychologist
impressions on
items and on
scales
measuring
attentionj
EXTERNALIZING BEHAVIOR AND UNDERACHIEVEMENT 139

Follow-up assessment

Achievement measure Behavioral measure


Grade
N Content Scale type Content Scale type Key findings*
365 3rd IQ (Kuhlman-Anderson) Categorical Although teacher ratings in 1st grade
and reading predicted both IQ and reading
(Metropolitan) tests categories in 3rd grade, the ratings
were also correlated with IQ and
readiness categories in 1st grade;
further analyses were not
performed

2nd Group-administered Dimensional Teacher ratings of intrapersonal


reading test adjustment in 1st grade
(Comprehensive significantly predicted reading
Primary) scores in 2nd grade, with sex,
ethnicity, SES, and readiness level
controlled (incremental /J2 = .064);
ratings of interpersonal adjustment
did not make significant prediction
167 lstc Group-administered Dimensional, Same as initial assessment Although teacher ratings of antisocial
(boys only) reading tests categorical behavior in K predicted poor
(Southgate) reading status at the end of 1st
grade, these ratings were also
associated with poor readiness in
K. The latter were a stronger
predictor of Ist-grade reading
(incremental R2 of antisocial
scores = .024)"
1,100 2nd Group-administered Categorical Same as initial assessment All 6 behavioral items in K predicted
reading and arithmetic reading and arithmetic status in
tests 2nd grade, with largest associations
for lethargic (fi = .21 for reading)
and hyperactive (<' = .20 for
reading).' Yet no cognitive-
readiness measure was available at
initial assessment
185f 8 yr IQ (Wechsler Categorical Parent ratings on Dimensional, Behavior problem status at 3 yr
Intelligence Scale for Rutter Scale A categorical significantly predicted, for boys,
Children) and reading and teacher mild and severe GRB as well as
(Neale) tests, yielding ratings on low IQ at 8 yr (see Table 8.7). Also,
classifications of GRB Rutter Scale B low general cognitive ability at 3
and SRR yr predicted, for boys only,
behavior-problem status (antisocial
and neurotic) at 8 yr (see Tables
11.4 and 11.6)"
174 2nd Group-administered Dimensional Teacher ratings on Categorical Although attentional-problem status
reading and general and Kindergarten at or before 5 yr significantly
achievement test categorical Performance predicted low reading status in 2nd
Profile; observer grade, with 45% of the persistent
ratings of attentional concerns group vs. 12%
overall of controls in the low-reading
competence category, attention problems were
also associated with low verbal and
perceptual-motor skills in K

(table continues)
140 STEPHEN P. HINSHAW

Table 1 (continued)

Initial assessment

Cognitive measure Behavioral measure


Urade
Study Location N (age) Content Scale type Content Scale type

Jorm, Share, Geelong, 543 K Peabody Picture Dimensional Teacher ratings on Dimensional,
Matthews, & Australia Vocabulary Rutter Scale B categorical
Maclean (1986) Test plus additional
item of
inattention

McGee, Dunedin, 476 5yr Parent ratings on Dimensional


Williams, New (boys only) Rutter Scale A
Share, Zealand and teacher
Anderson, & ratings on
Silva(1986) Rutter Scale B

Note. SES = socioeconomic status; K = kindergarten; yr = year; wk = week; GRB = general reading backwardness; SRR = specific reading
' I attempt to provide quantitative indexes of the magnitude of association between domains, when they are reported or when available data allow
b
Assessments were performed in 17 classrooms in four schools; N not reported.
Primary One in Scotland is equivalent to American kindergarten; Primary Two is equivalent to American first grade.
d
Additional analyses revealed that early-readiness classifications failed to predict increments in antisocial behavior over the 1.5-year interval.
e
I calculated $ statistics from the formula in Hays (1981). Also, additional analyses revealed that overall behavioral maladjustment in kindergarten
however, reported), even when social class was controlled.
f
From a randomly selected birth cohort of 705 nonimmigrant 3-year-old children, 94 were designated as behaviorally deviant. This group, along
8
Both the rating scale (Behavior Check-list) and the interview (Behavioral Screening Questionnaire) were designed to tap general behavioral and
h
Also, at the initial assessment (3 yr), behavior-problem status was correlated with both general cognitive-developmental delay and language delay
1
Eight assessments were performed over the 5-year period constituting the initial assessment. Age-appropriate tests of cognitive functioning were
1
Four groups were distinguished: No attentional concerns at any age; early concerns that abated by age 5; late-onset concerns, which did not appear
defined.
k
Classifications of SRR and GRB were not stable from first to second grade, in part because of floor effects for reading scores in first grade.
1
These items were the three from the Hyperactivity factor of Schachar, Rutter, and Smith (1981). It was also reported that the separate teacher rating
Peabody Picture Vocabulary Scores in kindergarten were reported.
m
This report included intermediate assessments at ages 7 and 9 years as well.
Only children with stable classifications of GRB or SRR from ages 9 and 11 were included in these categories.
0
With regard to changes at the intermediate assessments, by age 7, the GRB boys had shown an increase in hyperactivity scores over the levels they
apparent for the reading-delayed and comparison boys during the initial assessment (age 5), and only the SRR group showed any subsequent
problems, defined by stability of behavior-problem status across two or more age levels.

of the cognitive variables is indeterminate. Also, Maughan et al. Other reports seemed to favor the externalizing-behavior-
(1985), who followed subjects from the Sturge (1982) investiga- predicts-underachievement unidirectional model, but problem-
tion, found that over 40% of the 10-year-old boys with pure atic measures and insufficient analyses mandate caution. Lam-
SRRthat is, without concurrent antisocial behaviorhad be- bert et al. (1976) found that behavioral ratings from second- and
come delinquent by age 17, a proportion nearly as high as that fifth-grade teachers predicted grade point average (GPA) and
for initially antisocial boys. Maughan et al. added the qualifica- achievement test scores 7 years later, even when initial IQ was
tions, however, (a) that by age 17, a substantial number of sub- controlled. Yet the composite ratings included both internaliz-
jects had been lost to attrition and (b) that such results were not ing and externalizing items, limiting any specificity of the pre-
apparent for the prediction of age 14 antisocial behavior, for diction; as well, the criterion measures of grades and group
which initial antisocial behavior was by far the stronger predic- achievement test scores may not have been free of behavioral
tor. Furthermore, antecedent variables were not controlled in contamination. Also, Kupersmidt and Coie (1990) found that
this report. peer-nominated aggression in fifth grade predicted school drop-
EXTERNALIZING BEHAVIOR AND UNDERACHIEVEMENT 141

Follow-up assessment

Achievement measure Behavioral measure


Grade
N (age) Content Scale type Content Scale type Key findings*

453 1st, IQ (Columbia) and Categorical Teacher ratings on Dimensional, GRB children had higher antisocial
2nd reading (Neale) tests, Rutter Scale B categorical behavior scores than did SRR or
yielding classifications control children in 1st and 2nd
ofGRBandSRR" grades (magnitude of score
differences was about 3:1, p < .01),
but no differences were apparent
in K.. Yet, for items tapping
inattention-hyperactivity, GRB
children had higher scores than
did SRR or control children in K
and 1st grade (magnitude not
reported, p<. 01)'
437 I I yr" IQ (Wechsler Categorical" Same as initial Categorical With teacher ratings, both the GRB
(boys only) Intelligence Scale for assessment; in and SRR children had higher
Children) and reading addition, antisocial and hyperactive (but not
(Burt) tests, yielding interviews were internalizing) factor scores than
classifications of GRB performed, did controls at 5 yr (magnitude of
and SRR yielding score differences not reported, p
diagnoses < .05). By 11 yr, both the GRB
and SRR children had more
diagnoses of attention-deficit
disorder and oppositional-conduct
disorder than did controls (<' =
.22 for overall diagnoses)

retardation,
their calculation.

was somewhat more predictive of low achievement in 2nd grade than was concurrent 2nd-grade maladjustment (significance tests were not,

with 91 matched comparison children, was studied more intensively.


developmental concerns; no dimensions or clusters of internalizing or externalizing behavior were yielded.
(for boys only; see Table 4.7). Thus, correlated "causes" existed for boys.
administered at each assessment (e.g., McCarthy Scales were given in kindergarten).
until kindergarten; and persistent concerns, which were evident at multiple ages. See Palfrey et al. for description of how attentional concerns were

of attentiveness, made in kindergarten, was correlated with 2nd-grade IQ (r = .52) and with 2nd-grade reading (r = .34). No analyses involving

displayed at age 5; the SRR group showed an increase in hyperactivity from age 7 to age 9. As for parent ratings, no differences in behavior were
increases (for antisocial behavior, from age 5 to age 7). Finally, both the GRB and SRR boys were quite likely to exhibit persistent behavior

out, whereas school absences in fifth grade did not predict sub- tion-deficit disorder at this early age all independently pre-
sequent police contacts. Absences, however, are far from a per- dicted subsequent antisocial behavior. In addition, however, the
fect cognitive-achievement measure. Finally, Ledingham and significant interactions of attention-deficit disorder with intel-
Schwartzman (1984) and DeBaryshe, Patterson, and Capaldi ligence and family problems suggested to Moffitt (1990) that
(1991) did not include measures of cognitive functioning at "family adversity and verbal ability may modify the risk for
preassessment, precluding adequate tests of the alternative behavioral outcome that is posed by early ADD-H" (p. 904).
causal models. Also, at age 13, only the subgroup with delinquency plus a his-
In an extremely heuristic investigation, Moffitt (1990) pre- tory of attentional deficits had early problems in motor develop-
dicted antisocial behavior at age 11 from early childhood mea- ment, suggesting the addition of a neurodevelopmental risk
sures, through use of multiple regression analyses that included factor to the pattern of predictors. Furthermore, only this delin-
interaction terms. Specifically, once age 5 aggression had been
controlled, verbal IQ, family adversity, and diagnosis of atten- (text continues on page 146)
142 STEPHEN P. HINSHAW

Table 2
Longitudinal Investigations With Follow-Up Assessments in Secondary Grades
Initial assessment
Cognitive measure Behavioral measure
Grade
Study Location N Content Scale type Content Scale type
Lambert, Hartsough, San Francisco 300" 2nd,5thc IQ (Wechsler Intelligence Dimensional Teacher ratings on Dimensional
& Zimmerman Bay Area Scale for Children) 7 items tapping
(1976) and group- aggression,
administered immaturity,
achievement tests depression,
obtained from school dependency*1
records
Farrington (1979) London 411 8-10yr IQ (Progressive Matrices) Categorical Teacher and peer Categorical
(boys only) and vocabulary (Mill ratings of
Hill synonyms) tests troublesomeness,
daring,
dishonesty

Olweus(1983) Sweden 444 6th Averaged grades in (a) Dimensional Peer ratings of Dimensional
(boys only) Swedish and English, verbal and
and (b) math and other physical
subjects aggression
against peers and
verbal aggression
against teachers'

Ledingham & Montreal, 1,756 1st, 4th, Peer ratings on Categorical


Schwartzman Canda 7thk Pupil Evaluation
(1984) Inventory,
yielding
aggressive,
withdrawn, and
aggressive-
withdrawn
groups
Maughan, Gray, & Inner London 129 10 yr IQ (Wechsler Intelligence Categorical Teacher ratings on Categorical
Rutter(1985) (boys only) Scale for Children) Rutter Scale B,
and reading (Neale) yielding
tests, yielding antisocial
classification of SRR classification
EXTERNALIZING BEHAVIOR AND UNDERACHIEVEMENT 143

Follow-up assessment
Achievement measure Behavioral measure
Grade
N (age) Content Scale type Content Scale type Key findings*

168 9th, 12th Scholarship factor, Dimensional Adjustment, Peer Dimensional Initial teacher ratings (composite of 7
including GPA Involvement, items) predicted scholarship even when
and achievement Guidance IQ was controlled (incremental R2 =
test scores Problem factors' .31 for 9th grade and .12 for!2th grade).
Yet few significant predictions of Peer
Involvement and Guidance Problems
emerged'
409 14-16 yr Self-ratings on 38 Categorical Low IQ status at 8-10 yr predicted self-
(boys only) items assessing reported (p < .05) and official (p <
delinquency; .01) delinquency; low vocabulary status
official also predicted both types of
convictions for delinquency (p values .001; its
delinquent prediction of self-reported delinquency
offenses was significant even when SES, family
history, and family-marital conflict
were controlled). Teacher and peer
ratings at 8-10 yr also predicted
official delinquency (p < .001), even
when background factors were
controlled11
444 6th Same as initial assessment Same as initial assessment Although cross-lagged correlation
(boys only) analysis revealed a near-significant
tendency for initial aggression to
predict low grades at follow-up to a
greater degree than initial low grades
predicted subsequent aggression,
structural equation modeling indicated
only a small (standardized coefficient
= .252) link between initial aggression
and subsequent low grades.
Furthermore, the addition of
antecendent variables eliminated all
unidirectional and reciprocal causation
paths, suggesting a spurious
relationship model for the association
between domains'
753 4th, 7th, Classroom Categorical Aggressive and aggressive-withdrawn
10th placement, coded children were more likely than
as regular vs. withdrawn or comparison children to
special or below be in special classes or behind their
expected grade expected grade levels at 3-year follow-
level up (<# = .245 for omnibus chi-squared
statistic; no specific cell contrasts were
performed)
113 14 yr, 17 yr Group- Categorical Teacher ratings on Categorical Children with SRR at 10 yr were
(boys only) (only 84 administered Rutter Scale B extremely likely to display poor reading
boys at reading test at 14 yr; official status at 14 yr; antisocial status made
17yr) delinquency no incremental prediction. Also,
status at 17 yr whereas antisocial status at 10 yr was
predictive of similar status at 14 yr, with
SRR making no incremental
prediction, 41% of boys with SRR
alone at 10 yr were delinquent at 17 yr,
a rate comparable to that for boys with
initial antisocial status

(table continues)
144 STEPHEN P. HINSHAW

Table 2 (continued)
Initial assessment
Cognitive measure Behavioral measure
Grade
Study Location N (age) Content Scale type Content Scale type
Schonfeld, Shaffer, New York 126' 4yr, 7 yr IQ (Stanford-Binet at 4 Dimensional Psychologist Categorical
O'Connor, & City (boys only) yr; Wechsler ratings of
Portnoy(1988) Intelligence Scale for negativism,
Children at 7 yr) tests aggression, and
hostility, from
observations
during test
administration

Kupersmidt & Coie North 112 5th Averaged grades in Categorical Peer nominations Categorical
(1990) Carolina reading, language, and of "starting
math; excessive fights"
absences

Moffitt(1990) Dunedin, Varies" 3 yr-9 yr IQ (Wechsler Intelligence Dimensional Parent ratings on Dimensional
New Scale for Children Rutter Scale A,
Zealand Revised) and reading and teacher
(Hurt) tests ratings on Rutter
Scale B, yielding
antisocial scores"

DeBaryshe, Oregon 205 6th Latent variable of Dimensional


Patterson, & (boys only) antisocial
Capaldi(1991) behavior,
measured
through parent
and teacher
ratings, child
interviews, and
behavior
observations'
EXTERNALIZING BEHAVIOR AND UNDERACHIEVEMENT 145

Follow-up assessment
Achievement measure Behavioral measure
Grade
N (age) Content Scale type Content Scale type Key findings"
115 17yr IQ (Wechsler Adult Dimensional Psychiatric Categorical; First, verbal and full-scale IQ scores at
(boys only) Intelligence Scale) diagnoses Dimensional age 7 of children who later developed
and achievement (including conduct disorder were significantly
(Peabody conduct lower than those of the children who
Individual disorder) based did not develop conduct disorder. Next,
Achievement on structured path analyses revealed that effects of
Test) tests interviews with early aggression on IQ at 17 were
adolescents and eliminated when childhood IQ was also
parents;0 also, entered. In addition, childhood IQ
global severity indirectly influenced global severity
ratings of ratings at 17 through WAIS scores
psychopathology (standardized path coefficient of .50
from childhood to adolescent IQ and
.32 from adolescent IQ to global
severity rating); yet early aggression
also influenced global severity ratings
(coefficient = -.27). Finally, such third
variables as neurological soft signs,
parental psychopathology, and
environmental disadvantage did not
contribute to the concurrent
relationship between adolescent IQ and
conduct disorder0
99 12th Index of negative Categorical Police-juvenile Categorical Peer-nominated aggression in 5th grade
school outcome court contacts predicted school drop out in a stepwise
(suspensions, logistic regression, as did 5th-grade
retentions, school absences. 5th-grade school
truancy); early grades did not, however, predict police
school contact at follow-up
withdrawal (drop
outs)
435 llyr-15yr Parent ratings on Dimensional and Verbal IQ scores, family adversity, ADD
(boys only) Revised categorical classification, and the interaction of
Behavior ADD classification with verbal IQ and
Problem with family adversity predicted
Checklist; antisocial behavior at 11 yr, even after
teacher ratings antisocial behavior at 5 yr had been
on Rutter Scale controlled. Furthermore, only boys
B; self-reported with ADD histories and delinquency
delinquency; classifications at age 13 had a history
official police of motor delay at 3-5 yr. Finally, only
records the ADD plus delinquent subgroup had
persistently high levels of aggressive
acts at 15 yr
156' 7th Latent variables of Dimensional Structural equation modeling revealed
(boys only) (a) academic that the optimal fit was provided by
engagement, a model in which antisocial behavior
measured in 6th grade had a direct influence on
through parent achievement in 7th grade (/8 = -.21).
and child In this model, however, the indirect
interviews and links were stronger: Coercive discipline
teacher ratings of practices were strongly correlated with
items reflecting antisocial behavior (r = .83), and these
homework and practices predicted academic
schoolwork; and engagement (0 = -.73), which in turn
(b) academic affected academic achievement (/3 =
achievement, .61)
measured
through parent
and teacher
ratings of school
performance and
standardized
achievement tests
in four subject
areas
(table continues)
146 STEPHEN P. HINSHAW

Table 2 (continued)
Note. GPA = grade point average; yr = year; SES = socioeconomic status; SRR = specific reading retardation; ADD = attention-deficit disorder.
* I attempt to provide quantitative indexes of the magnitude of association between domains, when they are reported or when available data allow
their
b
calculation.
The 300 subjects were a random sample of 2nd and 5th graders from a population of 2,400 students in 11 schools in Northern California.
c
Initial assessments were performed in two different grade levels; 7-year follow-up assessments were performed for both groups.
d
Peer sociometric nominations and self-ratings were also included in the initial assessment.
' The factors of 9th and 12th grade cognitive and behavioral status emerged from factor analyses of archival indexes of both domains. For example,
Adjustment included counselor ratings of positive attributes, and Guidance Problems included frequency of disciplinary referrals and absences,
but Peer Involvement was not specified. Overall Status measures were composites of behavioral (delinquency) and achievement-related (e.g., honor
roll) measures; they are therefore not considered.
' Analyses of the ability of separate behavioral items to predict achievement were not performed.
*h Some predictor measures were obtained at 8 yr, but others were not collected until 10 yr.
Only the p values (and no actual <t> values) were reported. Also, no correlations between the cognitive and the behavioral predictors at 8-10 yr were
reported, nor were longitudinal predictions corrected for the influence of the alternate domain.
' The overall goal was to determine whether (a) poor grades in 6th grade predicted later aggression, (b) aggression in 6th grade predicted later poor
grades, or (c) both (a) and (b) were true (i.e., reciprocal causation). In addition, such antecedent variables as SES, parental ages, divorce, birth order,
and birth out of wedlock were added to the model, to test whether (d) common background factors could account for the association (i.e., spurious
relationship model).
' SES, parental age, and birth out of wedlock were the antecedent variables that accounted for the association between poor grades and aggression.
k
1
Initial assessments were performed at three different grade levels; 3-year follow-up assessments were performed for all groups.
Two groups of inner-city, Black boys were investigated: a group with neurological soft signs at the initial assessment and a matched comparison
group.
m
Only 94 subjects were available for path analyses.
"0 The parent interviews also yielded data regarding parental psychopatholpgy and environmental disadvantage.
Logistic regression analyses were also performed, to ascertain the predictions from early cognitive and behavioral status to the presence vs.
absence of conduct disorder. Low IQ scores scores were related only to conduct disorder; no other psychiatric diagnoses were associated. Schonfeld
et al. (1988) noted that only acquired (i.e., Verbal) and not fluid (i.e. Performance) IQ was associated with conduct disorder.
"For biennial assessments of the Dunedin cohort, Ns were 1,037 (3 yr), 991 (5 yr), 954 (7 yr), and 955 (9 yr). Analyses were restricted to boys,
however, because of the small number of girls classified as delinquent at follow-up.
" Also assessed, as background variables, were family adversity (a composite of SES, maternal mental health, maternal IQ, and family social
environment), motoric functioning (Bayley scales at age 3 and McCarthy scales at age 5), history of perinatal problems, and diagnosis of ADD.
' Also measured were parental academic achievement (e.g., IQ, years of education) and coercive discipline at home.
* Listwise deletion procedures resulted in a final sample of 156 for the structural equations. Most missing data resulted from the unavailability of
school test scores for the academic achievement variable in 7th grade.

quent plus attention-deficit disorder subgroup had significant tion of interest. That is, such variables would need to correlate
neuropsychological deficits (see Mofl&tt & Sirva, 1988b). Causal with each domain and to exert their influence early in develop-
paths were therefore found to span cognitive, behavioral, psy- ment; furthermore, their effects should not be explainable by
chosocial, and neurodevelopmental variables, interacting in other background factors. Also, in comprehensive tests of alter-
complex fashion.10 native models, the variables would need to supersede any unidi-
Overall, like many reports conducted during childhood (see rectional or reciprocal causal paths when entered into predic-
Table 1), investigations with adolescent follow-up often failed to tion equations (e.g., Olweus, 1983). I begin with the macro vari-
control for the alternate domain during initial assessment, and able of SES, move next to familial factors, and then cover the
several also failed to include adequate measures of antecedent intraindividual variables of IQ, language skills, and neurode-
variables. Such issues severely constrain inferences regarding velopmental delay. Considering such variables as separate enti-
causal precedence. Concerning the two reports with designs ties ignores their probable interactions: I thus provide several
that could optimally test alternative causal models, one yielded illustrations of combined effects.
no support for unidirectional or reciprocal causation models
once antecedent variables were partialed (Olweus, 1983).
Whereas the other appeared to support a unidirectional link
between early verbal deficits and later conduct disorder, the
domain of early aggression was inadequately measured (Schon- 10
Two additional investigations deserve mention. The study of
feld et al., 1988). Finally, because such antecedent factors as Huesmann, Eron, and Yarmel (1987) was not included in Table 2 be-
coercive parenting (DeBaryshe et al., 1991), family adversity, cause the follow-up analyses were performed in the subjects' adult-
verbal skills, and motor delay (Moffitt, 1990), as well as low hood, 22 years after the initial assessments at age 8. This report re-
SES, birth out of wedlock, and parental age (Olweus, 1983) have vealed that (a) antisocial outcomes (e.g., arrest records, self-reports)
been found to influence outcomes, I examine more closely the were not significantly predicted by early IQ, once initial aggression
potential influence of third variables on the association be- (measured through peer sociometric methods) was controlled, but that
tween underachievement and externalizing behavior. (b) early aggression predicted age 30 educational attainment, even
when early IQ was controlled. Next, because the report of McGee,
Antecedent Variables Share, Moflitt, Williams, & Silva (1988) was an extension into adoles-
cence of the investigation of McGee, Williams, Share, Anderson, &
My chief goal here is to ascertain, in brief fashion, whether Silva (1986; see Table 1), it did not constitute an independent investiga-
potential antecedent factors could plausibly explain the associa- tion.
EXTERNALIZING BEHAVIOR AND UNDERACHIEVEMENT 147

SES Sirva, 1984a; Rutter et al., 1976). Siblings of learning disabled


and of purely hyperactive youngsters also tend to have achieve-
In combination with other psychosocial and familial vari- ment and language difficulties (e.g, August & Stewart, 1983).
ables, SES has been implicated as a factor that could account for Thus, on an aggregate level, aggressive disorders appear to have
the association between underachievement and externalizing a family configuration different from that of either under-
behavior (see Olweus, 1983; Sturge, 1982). Other evidence, how- achievement or hyperactivity; the latter two difficulties show
ever, suggests that SES does not alone explain the overlap. First, intriguing similarities (see also Schachar & Wachsmuth, 1990).
regarding academic performance, although impoverished liv- Regarding intrafamilial processes, severe marital discord
ing conditions and overcrowding are associated with such vari- and coercive parent-child exchanges are clearly related to anti-
ables as lowered intellectual ability, globally low achievement, social behavior patterns (Patterson, 1982); they may also asso-
and poor grades, there is contradictory evidence regarding the ciate with some aspects of underachievement (see Stevenson &
effects of socioeconomic factors on specific or IQ-discrepant Fredman, 1990). Data are sparse, however, regarding the effects
reading delays (see Rutter & Yule, 1975; Stevenson & Fredman, of familial functioning on the association between domains.
1990). A safe conclusion is that the effects of social class on The seminal report of Richman et al. (1982)in which unse-
IQ-discrepant reading failure are considerably smaller than lected 3-year-olds from a birth cohort were followed for 5 years
those on unadjusted achievement or on intelligence per se. Fur- provides a rare opportunity to examine the interactive role of
thermore, more specific aspects of family interactionfor ex- early familial variables with other risk factors in the develop-
ample, parental attitudes toward literacy or parental listening ment of underachievement and behavioral difficulties. In this
to the child's readingappear to mediate the relationship that investigation, age 3 social deprivation was measured by low
exists between SES and reading attainment (Stevenson & Fred- SES, use of social services, and large family size; family dys-
man, 1990). function was indexed by a composite variable that included
On the behavioral side, although inattention and hyperactiv- maternal depression-anxiety, marital dysfunction, overcrowd-
ity show little or no correlation with social class (e.g., Szatmari, ing, and life stressors; and parental criticism-hostility toward
Boyle, & Offord, 1989), aggressive and conduct-disordered be- the child was appraised by research interviewers.
havior is associated with SES (eg., Paternite, Loney, & Lang- Amidst a series of complex relationships, language delay at 3
horne, 1976; Szatmari, Boyle, & Offord, 1989). Yet, in adoles- years of agea variable correlated with social deprivation
cence, links between externalizing behavior and poor academic predicted later learning problems and behavior problems (see
performance hold regardless of social strata (Hawkins & subsequent section); adverse family environment at 3 years pre-
Lishner, 1987). Furthermore, the effects of social class on anti- dicted the onset of new behavior problems by age 8 but did not
social behavior are outweighed by more specific family vari- correlate with the persistence of extant behavior problems. In-
ables, like parental hostility and parent-child conflict (Pater- triguingly, early adverse family climate predicted specific (i.e.,
nite et al, 1976; see also Howlin & Rutter, 1987). IQ-adjusted) reading deficits at age 8, even when early behavior-
In summary, although lowered standards of living may set problem status and early language delay were controlled. The
the stage for subaverage achievement and aggression, SES is most important aspect of familial adversity in this relationship
unlikely to explain the early relationship between hyperactivity was maternal depression-anxiety, leading Richman et al.
and academic failure, particularly if the latter is separated from (1982) to speculate that a particular type of maternal unavail-
IQ. Also, in older children and in adolescents, links between ability may be crucial, in formative years, to subsequent read-
delinquent behavior and underachievement are maintained ing difficulties that are not explainable by low IQ. The presence
across diverse social strata. Measures of family functioning of early language delays enhanced this relationship, pointing to
bear examination as the specific mechanisms and processes by the combination of intraindividual and familial variables in
which effects of social disadvantage on underachievement and precipitating reading problems. In addition, the prediction was
externalizing behavior may be mediated. stronger for boys than for girls and for children with age 8
behavior problems (internalizing and externalizing) than for
those without. Familial factors are thus likely to be embedded
in a network of other antecedent variables in predisposing to
Familial Enables
underachievement and behavioral disturbance.
Such findings, which bear replication and extension (see also
The relationship of family factors to child psychopathology Estrada, Arsenio, Hess, & Holloway, 1987), suggest the com-
and to underachievement constitutes an enormous topic (see plexity of interrelationships among child-related, linguistic, so-
Hetherington & Martin, 1986; Stevenson & Fredman, 1990). cial, and early familial variables in the development of both
Concerning family history of psychopathology, there appears underachievement and behavioral difficulties. A key methodo-
to be a different aggregation of disorders in families of aggres- logic concern in this area involves ascertaining whether family
sive as opposed to underachieving children. Whereas child ag- variables are antecedents or consequences of child problems;
gression is correlated with paternal psychopathy and substance longitudinal investigations are essential to help infer causal re-
abuse (see Lahey et al., 1988) and with maternal antisocial fea- lationships. Finally, although family factors are often consid-
tures and somatization (Lahey, Russo, Walker, & Piacentini, ered to be psychosocial, such variables contain both genetic
1989), children with learning difficulties (or with pure hyperac- and environmental influences, except in adoptive families (see
tivity) often have family histories marked by underachievement Plomin, Loehlin, & DeFries, 1985). Thus, biological and envi-
as well as speech and language delay (e^ McGee, Williams, & ronmental factors are typically confounded with respect to fa-
148 STEPHEN P. HINSHAW

mi Hal causes of cormorbidity between underachievement and well & Baker, 1980). Thus, speech and language problems ap-
behavior problems. parently are excellent candidates as underlying variables: They
correlate with both domains of interest, and they predate the
Intelligence emergence of the association.
In the first place, however, whereas early language delays
As noted earlier, the initial findings that antisocial behavior predict later achievement problems (Stevenson, 1984), most in-
was correlated specifically with IQ-discrepant underachieve- dexes of delayed language are associated with lowered IQ. Spe-
ment have not been confirmed. Indeed, in the investigations cific (i.e., IQ-discrepant) language deficits are a weaker predic-
reviewed herein, lowered intelligence and globally low achieve- tor of later achievement problems than are general (non-IQ-ad-
ment are more frequent correlates of externalizing behavior justed) deficits, leading to the conclusion that subaverage IQ
than are SRR or IQ-discrepant achievement. The possibility may, in most cases, be the actual causal factor (Silva, 1987;
remains, then, that low intellectual capacity is causally responsi- Stevenson & Richman, 1978). Furthermore, in grade school
ble for both underachievement and externalizing behavior. \fet children, some types of reading problems do not involve lan-
the relationship is not likely to be so straightforward. guage-processing difficulties but are more visual-perceptual in
First, externalizing behavior and adverse family climate cor- nature (e.g., Rourke, 1985). Thus, IQ-independent language
relate with IQ-discrepant achievement in some investigations problems are unlikely to be strong or sensitive predictors of
(e.g., Richman et al, 1982; Rutter et al, 1970), and IQ cannot underachievement.
account for such associations. Second, measurement issues are Regarding relationships with behavior problems, mild-to-
salient. That is, because of the aforementioned tendency for moderate language problems are associated with several types
children with attentional problems to score poorly on several of difficulties, and attention-deficit disorder is at or near the
subtests of the WISC-R despite otherwise "normal" intelli- top of the list (see Beitchman, Hood, Rochon, & Peterson,
gence, IQ scores may be spuriously depressed, leading to inva- 1989; Cantwell, Baker, & Mattison, 1979; Cohen, McDonald,
lid presumptions of links between externalizing behavior and Horodezky, & Davine, 1989). (The most prevalent behavioral
subaverage intelligence. Third, on the basis of their longitu- correlates of marked early language delay are severe conditions
dinal data, Schonfeld et al. (1988) argued that verbal, or "crys- such as autism and pervasive developmental disorders.) Al-
tallized," intelligencewhich is heavily influenced by the per- though replication is sorely needed, there is some evidence that
son's learning environmentis more predictive of adolescent expressive, as opposed to receptive, language disorders early in
conduct problems than is performance IQ, which taps "fluid" life are more specifically correlated with externalizing prob-
intelligence and relates to hereditary factors (see also Wilson & lems (see Beitchman, Tuckett, & Batth, 1987). Viewed as a
Herrnstein, 1985). Fourth, more specific types of verbal media- group, however, early language problems do not predict exter-
tion deficienciestermed social cognitive information process- nalizing behavioral problems with specificity; in fact, the most
ing deficits by Dodge (1985)have been shown to relate to defi- common behavioral outcome of early language delay involves
cient self-regulation, peer adjustment problems, and a pattern internalizing problems (Richman et al., 1982; Stevenson, Rich-
of hyperactive plus aggressive behavior (Milich & Dodge, 1984). man, & Graham, 1985). Also, in contrast to the usual pattern
Thus, externalizing behavior may be more predictable from for externalizing behavior disorders, girls are at higher risk than
deficient learning environments or deficits in verbal mediation boys for behavioral consequences of language delay (Beitch-
of social stimuli than from lowered intelligence per se. man, Hood, & Inglis, 1990).
In short, intelligence is sufficiently heterogeneous as a con- Overall, whereas predictive relationships are not highly spe-
struct that it may mask more specific aspects of cognitive func- cific and whereas IQ is a confounding factor, some proportion
tioning that hold stronger relationships with externalizing be- of children with early language delays develop externalizing
havior. In subsequent research, it will be critical to include in- problems, particularly attention deficits; and a sizable percent-
dexes of social cognitive information processing as well as age of children with early language problems manifest subse-
measures of general intelligence and achievement as antecedent quent underachievement. Additional longitudinal investiga-
variables (Schonfeld et al., 1988). Yet because of its clear status tions are needed to document and explain the comorbidity of
as a correlate of both aggressive-delinquent behavior and of behavioral and achievement-related difficulties following early
poor achievement, subaverage IQ may predispose toward the language delay.
association between these two domains, particularly when fol- Concerning putative causal chains, there are a host of ways in
low-up into adolescence is performed. which speech-language problems could predispose toward the
association of interest (see Silva, 1987). For one thing, verbal
mediation is a key factor in self-regulation, and acting-out be-
Speech and Language Difficulties
havior is negatively associated with a variety of relevant verbal
A sizable literature has emerged on the relationship between skills (Hogan & Quay, 1984). Also, the young child with diffi-
early language problems and the development of both behav- culties in producing (and processing) language might be at risk
ioral difficulties and achievement deficits (for reviews, see for considerable frustration and for negative parent-child and
Cantwell & Baker, 1977; Howlin & Rutter, 1987; Silva, 1987; peer interchanges, all of which could precipitate or maintain
Stevenson, 1984). Furthermore, current conceptions of reading externalizing behavior and achievement difficulties (for ex-
failure emphasize the key role of phonologic- and linguistic- tended discussion, see Howlin & Rutter, 1987). Furthermore,
processing difficulties (Wagner & Torgeson, 1987), and lan- because language delay is associated with lowered IQ and with
guage-disordered children are prone to academic failure (Cant- perceptual-motor deficitsfactors associated with neurode-
EXTERNALIZING BEHAVIOR AND UNDERACHIEVEMENT 149

velopmental immaturitygeneral neurodevelopmental delays marked, IQ-discrepant achievement deficits with either hyper-
might predispose both to language problems and to the subse- activity or conduct disorder is less than 20%, well below esti-
quent conjoint presence of externalizing behavior and under- mates that are often reported in the literature." This degree of
achievement, a contention deserving of brief elaboration. comorbidity is still far in excess of chance rates, however. Fur-
thermore, the school-related difficulties of children with exter-
Neurodevelopmental Immaturity nalizing problems are not limited to formal learning disabili-
ties: Such indexes as poor grades, retention, suspension, and
Problems in the maturation of basic perceptual and motoric academic deficits that fall short of full criteria for under-
functions, presumably reflecting neurodevelopmental delay, achievement are common in hyperactive and aggressive chil-
have been invoked as a prior causal factor for language deficits dren (e.g., Barkley et al., 1990; Fischer et al, 1990; Forness et al,
by Beitchman (1985). Whether, in fact, language difficulties in press). Thus, the overlap between externalizing difficulties
are preceded by more basic perceptual, precognitive deficits is a and academic failure clearly is sizable and important.
controversial topic. In contrast to the more accepted view that Second, there is a crucial developmental progression with
language processing proceeds from a top-down, semantically regard to the association. In early and middle childhood, the
driven model (for cogent discussion, see Yule & Rutter, 1985), specific link is between hyperactivity-inattention and under-
Tallal (1980) produced evidence that language dysfunction is achievement; aggression shows overlap with learning problems
explainable on the basis of subtle but real perceptual dysfunc- during these years chiefly through its comorbidity with inat-
tions that underlie the processing of spoken sounds. Further- tention-hyperactivity (Frick et al, 1991). \et, by adolescence,
more, in a provocative report with language-impaired pre- clear links have emerged between frankly antisocial behavior
schoolers, Tallal, Dukette, and Curtiss (1989) demonstrated and variables related to verbal deficits and underachievement.
that the children's behavior profiles differed from those of con- Such a progression prompts a detailed examination of the devel-
trols, not on the basis of patterns conforming to current con- opmental course of attentional deficits and of the aggression
ceptions of emotional or behavior disorders, but solely regard- that frequently co-occurs with them (cf. Barkley et al, 1990).
ing items that tap attentional, motoric, and perceptual process-
ing. This finding fuels the contention that neurodevelopmental Underlying Mechanisms
delay rather than behavioral disturbance distinguishes children
with early language impairments from their peers. The overriding conclusion from the investigations reviewed
Neurodevelopmental immaturity-delay, however, is a fuzzy above is sobering: Despite avid theoretical and empirical inter-
construct, badly in need of sharpening. Note, in addition, that est in revealing the underpinnings of overlap between exter-
psychosocial variables appear critical for mediating the out- nalizing behavior and underachievement, alternative causal
comes of children who, with clear perinatal complications, are models have rarely been tested with sufficient rigor. I have
at risk for neurodevelopmental delay (e.g., Werner & Smith, highlighted several plausible reasons for this state of affairs,
1982). Despite the importance of examining neurodevelop- including (a) the genuinely difficult methodologic and concep-
mental risk, it would be quite premature to contend that in- tual issues regarding the domains under consideration, (b) inad-
traindividual/biological neurodevelopmental factors are solely equate assessment of antecedent variables, (c) insufficient use
responsible for producing the association between under- of statistical analyses with adequate control of correlated pre-
achievement and externalizing behavior. dictors and of antecedents, and (d) the need for costly longitu-
dinal investigations. Measurement issues are also paramount;
in fact, the two most sophisticated investigations to datein
Summary
which antecedent variables were assessed and causal models
Although none of the antecedent factors discussed herein is were explicitly testedboth had limitations regarding measure-
likely to be the sole explanatory variable for the association of ment of key domains during initial assessment (Olweus, 1983;
interest, each factor may predispose at least some children to- Schonfeld et al, 1988). Although definitive investigations re-
ward overlapping patterns of externalizing behavior problems garding the processes underlying the association of interest still
and underachievement. Furthermore, increasing evidence remain to be performed, the field is beginning to appreciate the
points to the potential interaction of several antecedent factors quality of research efforts needed to elucidate causal mecha-
in shaping comorbidity (see Moffitt, 1990; Olweus, 1983; Rich- nisms (e.g, Loeber, 1990; Moffitt, 1990).
man et al, 1982). Overall, to elucidate the mechanisms that The absence of adequate designs and statistical analyses of-
underlie relationships between externalizing behavior and un-
derachievement, investigators must reliably assess pertinent an- 11
The partial independence of underachievement and externalizing
tecedent variables and must statistically control for their influ- behavior is further demonstrated by investigations of Porter and
ence in analyses of causal models. Rourke (1985) and Fuerst, Fisk, and Rourke (1989), who used cluster
analysis to identify behavioral subtypes of learning disabled children.
Conclusions and Issues The largest subgroup in each study had essentially normal behavioral
adjustment, and the biggest "problem" subgroups had internalizing
Overlap Between Domains difficulties. Only 17%-18% of the total samples of learning disabled
youngsters had a behavioral profile marked by hyperactivity-aggres-
First, whereas externalizing behavior and formally defined sion. Thus, attention deficits or conduct problems, on the one hand,
underachievement are clearly associated, recent investigations and learning disabilities, on the other, are not simply alternate names
with stringent diagnostic criteria reveal that the overlap of for the same set of problems (Rourke, 1988; Silver, 1990).
150 STEPHEN P. HINSHAW

ten precludes inferences regarding unidirectional links. Yet sev- which to categorize children as well as sizable samples. None-
eral reports suggest the possibility of at least some causation theless, creatively designed interventions could begin to shed
from one domain to the other. For example, although asso- light on the causal nature of the association under discussion.
ciated with hyperactivity and aggression in early grades, read-
ing delay-retardation may predispose to subsequent increases Inattention
in externalizing behavior (Jorm, Share, Matthews, & Mclean, I have emphasized that in elementary grades, inattention-hy-
1986; McGee et al, 1986; McGee et al, 1988). Furthermore, a peractivity is a more consistent correlate of underachievement
subgroup of children with reading failure unaccompanied by than is aggression, a finding that emphasizes the importance of
externalizing behavior may go on to develop formal delin- defining the nature of attentional deficits. Opinion varies dra-
quency by late adolescence (Maughan et al., 198S). Although matically as to the essence of such problems; alternative con-
both findings are consistent with theoretical notions regarding ceptions emphasize their behavioral, information-processing,
the role of school failure in shaping acting-out behavior (e.g, motoric response output, or motivational characteristics (e.g.,
Hawkins & Lishner, 1987; Hirschi, 1969; Rutter& Ciller, 1983), Barkley, 1989; Douglas, 1983; Sergeant & van der Meere, 1990).
measures of potential antecedent variables that could have ac- Whatever their precise nature, early attentional problems inter-
counted for the increase in antisocial behavior over time were act in crucial ways with neurodevelopmental delay and with
not provided, mandating caution in inferring unidirectional later neuropsychological dysfunction to predispose toward aca-
causal relationships. demic failure and aggressive delinquency (Moffitt, 1990; Mof-
Advocating the reverse unidirectional model, Patterson and fitt & Silva, 1988b). Clarifying the construct of dysregulated
colleagues have argued strongly that early antisocial behavior- attention should have a major impact on understanding causal
shaped by coercive family interactionscauses academic fail- mechanisms linking externalizing behavior with under-
ure once schooling begins (Patterson, DeBaryshe, & Ramsey, achievement.
1989). 'Vet the absence of measures of cognitive-readiness prob-
lems during initial assessment periods mitigates support for Final Issues
this unidirectional hypothesis (DeBaryshe et al, 1991). On the
other hand, in a more rigorous test of unidirectional hypothe- First, in several of the reports that were reviewed, the rela-
ses, Huesmann et al. (1987) demonstrated that age 8 aggression tionships between domains pertained not to extreme groups of
predicted low achievement in adulthood, even when early IQ children, but rather to continuously distributed variables. For
was controlled, whereas age 8 IQ did not predict adult antiso- example, Richman et al. (1982) found that early adverse family
cial behavior when early aggression was partialed. environment was an independent predictor of IQ-discrepant
The focus on overarching developmental pathways may mask reading scores at age 8, across the entire distribution of such
the operation of distinct causal mechanisms for different sub- scores. This family index did not, however, predict to the ex-
groups of children. Indeed, only some children with notewor- treme group of children with SRR. Also, in the 22-year longitu-
thy externalizing behavior problems in kindergarten go on to dinal data of Huesmann et al. (1987), early aggression was an
develop underachievement (e.g., McMichael, 1979). I highly rec- independent predictor of a continuous measure of age 30 educa-
ommend emphasis on assessment strategies that can uncover tional attainment. Such findings have relevance for the entire
subgroups with different developmental trajectories as an anti- population of children, arguing for dimensional as well as cate-
dote to overly broad claims for models that do not apply to all gorical approaches in psychopathology research (e.g., Blash-
children. Of course, research involving such subgroups will re- field, 1984).
quire large initial samples.
Regarding the inference of causal relationships, it is striking 12
that the entire database for this review comprises nonexperi- Although not explicitly related to the association of interest, the
mental investigations. Even with careful measurement strate- study of Bradley and Bryant (1985) provides an example of heuristic
experimental research. In investigating the relationship between pre-
gies, appropriate designs, and assiduous statistical analyses, in-
school phonologic abilityexemplified by recognition of rhymes and
ferring causation on the basis of correlational models is prob- alliterationand subsequent reading proficiency, these investigators
lematic. Experimental intervention studies would clearly be first showed a predictive longitudinal relationship, suggesting a causal
preferable.12 \fet the developmental and methodologic issues link. During an experimental intervention, instruction in rhyming and
that I have repeatedly mentioned render appropriate experi- alliteration caused subsequent improvement in reading-spelling (but
ments problematic. For example, given that for many children, not arithmetic) ability, yielding far stronger evidence for the causal
the intertwining of externalizing behavior and underachieve- hypothesis that phonologic ability is necessary for reading. For the
ment may emanate from complex interplays of antecedent fac- problem area under review, experiments would require (a) clinical sam-
tors that operate early in development, stratightforward inter- ples of either underachieving or externalizing youngsters, or groups
ventions in behavioral versus achievement domains may be with documented risk for such conditions, (b) randomly assigned treat-
impossible. "Vet from a clinical perspective, intervention efforts ment conditions directed toward either academic achievement or ex-
ternalizing behavior but not both, and (c) systematic assessment of
should be delivered before problems become entrenched (see both achievement-related and behavioral outcomes, preferably over
Loeber, 1990). Another difficulty is that the probable operation lengthy time periods. Interventions related to one domain would hypo-
of different causal mechanisms within different subgroups thetically induce improvements in the alternate problem area. Few
mandates examination of Subgroup X Treatment interactions extant reports come close to meeting these criteria, and space does not
rather than main effects of treatment conditions; yet formation permit discussion of those relevant intervention studies that have been
of relevant subgroups requires an adequate conceptual basis on performed.
EXTERNALIZING BEHAVIOR AND UNDERACHIEVEMENT 151

Next, I have noted that reading achievement and intelligence atric risk: Toward a model of neurodevelopmental immaturity. Psy-
are complex composites of a variety of component skills. Part chiatric Clinics of North America, 8, 721-735.
of the rationale for my focus on such broad measures of cogni- Beitchman, J. H., Hood, J., & Inglis, A. (1990). Psychiatric risk in chil-
tive functioning is their direct clinical relevance for children's dren with speech and language disorders. Journal of Abnormal Child
success. Yet more specific cognitive variables, from subtypes of Psychology, 18, 283-296.
verbal abilities to social-cognitive measures that might directly Beitchman, J. H., Hood, J., Rochon, J, & Peterson, M. (1989). Empiri-
cal classification of speech/language impairment in children: II.
mediate aggressive responding (e.g., Dodge, 1985), should re-
Behavioral characteristics. Journal of the American Academy of
ceive increased emphasis in discussions and investigations of Child and Adolescent Psychiatry, 28,118-123.
cognition-behavior relationships (see Schonfeld et al, 1988). Beitchman, J., Tuckett, M., & Batth, S. (1987). Language delay and
Although important in its own right, the broad variable of un- hyperactivity in preschoolers: Evidence for a distinct subgroup of
derachievement requires further elucidation if behavior-cogni- hyperactives. Canadian Journal of Psychiatry, 32, 683-687.
tion relationships are to be fully understood. Berger, M., Yule, W, & Rutter, M. (1975). Attainment and adjustment
Finally, as I have emphasized, the material covered herein in two geographical areas: II. The prevalence of specific reading
underscores the potential complexity of causal mechanisms retardation. British Journal of Psychiatry, 126, 510-519.
underlying the association of interest. Given the interactions Blashfield, R. K. (1984). The classification of psychopathology: Neo-
and transactions among social, familial, linguistic, and neuro- Kraepelinian and quantitative approaches. New \brk: Plenum Press.
behavioral variables that may culminate in the overlap between Bradley, L., & Bryant, P. (1985). Rhyme and reason in reading and spell-
underachievement and externalizing behavior, teasing apart ing. Ann Arbor: University of Michigan Press.
the effects of any single background factor is likely to be quite Brandenburg, N. A., Friedman, R. M, & Silver, S. E. (1990). The epide-
difficult or even misguided. The challenge for the field is to miology of childhood psychiatric disorders: Prevalence findings
derive explanatory models with sufficient rigor and complexity from recent studies. Journal of the American Academy of Child and
Adolescent Psychiatry, 29, 76-83.
to handle the diversity of causal factors.
Cantwell, D. P., & Baker, L. (1977). Psychiatric disorder in children
with speech and language retardation. Archives of General Psychia-
References try, 34,583-591.
Cantwell, D. P., & Baker, L. (1980). Academic failures in children with
Achenbach, T. M., & Edelbrock, C. S. (1978). The classification of child communication disorders. Journal of the American Academy of
psychopathology: A review and analysis of empirical efforts. Psycho- Child Psychiatry, 19, 579-591.
logical Bulletin, 85,1275-1301. Cantwell, D. P., Baker, L., & Mattison, R. E. (1979). The prevalence of
Achenbach, T. M., & Edelbrock, C. (1983). Manual for the Child Behav- psychiatric disorder in children with speech and language disorder:
ior Checklist and Revised Child Behavior Profile. Burlington, VT: An epidemiologic study. Journal of the American Academy of Child
Author. Psychiatry, 18, 450-461.
Achenbach, T. M., McConaughy, S. H., & Howell, C. T. (1987). Child/ Cantwell, D. P., & Satterfield, J. H. (1978). The prevalence of academic
adolescent behavioral and emotional problems: Implications of underachievement in hyperactive children. Journal of Pediatric Psy-
cross-informant correlations for situational specificity. Psychologi- chology, J, 168-171.
cal Bulletin, 707,213-232. Cicchetti, D. (1989). Developmental psychopathology: Some thoughts
Aman, M. G, & Singh, N. N. (1983). Specific reading disorders: Con- on its evolution. Development and Psychopathology, 1,1-4.
cepts of etiology reconsidered. In K. D. Gadow & I. Bialer (Eds.), Cohen, N. J., McDonald, I, Horodezky, N., & Davine, M. (1989, Oc-
Advances in learning and behavioral disabilities (Vol. 2., pp. 1-47). tober). Psychiatrically disturbed children with unsuspected language
Greenwich, CT: JAI Press. disorders: Psychiatric and language characteristics. Paper presented
American Psychiatric Association. (1987). Diagnostic and statistical at the annual meeting of the American Academy of Child and Ado-
manual of mental disorders (Sided., rev). Washington, DC: Author. lescent Psychiatry, New York.
Anderson, J., Williams, S., McGee, R., & Silva, P. (1989). Cognitive and
DeBaryshe, B., Patterson, G. R, & Capaldi, D. M. (1991). A perfor-
social correlates of DSM-IH disorders in preadolescent children.
mance model for academic achievement. Manuscript submitted for
Journal of the American Academy of Child and Adolescent Psychiatry,
publication.
28, 842-846.
August, G. I, & Garfinkel, B. D. (1989). Behavioral and cognitive sub- Dishion, T. 1, Loeber, R., Stouthamer-Loeber, M., & Patterson, G. R.
types of ADHD. Journal of the American Academy of Child and Ado- (1984). Skill deficits and male adolescent delinquency. Journal of
lescent Psychiatry, 28, 739-748. Abnormal Child Psychology, 12, 37-54.
August, G. J., & Garfinkel, B. D. (1990). Comorbidity of ADHD and Dodge, K. A. (1985). Attributional bias in aggressive children. In P. C.
reading disability among clinic-referred children. Journal of Abnor- Kendall (Ed.), Advances in cognitive-behavioral research and therapy
mal Child Psychology, 18, 29-45. (Vol. 4, pp. 73-110). San Diego, CA: Academic Press.
August, G. J., & Stewart, M. A. (1983). Familial subtypes of childhood Douglas, V I. (1983). Attentional and cognitive problems. In M. Rutter
hyperactivity. Journal of Nervous and Mental Disease, 171, 362-368. (Ed.), Developmental neuropsychiatry (pp. 280-329). New "Vbrk:
Barkley, R. A. (1989). The problem of stimulus control and rule-gov- Guilford Press.
erned behavior in attention deficit disorder with hyperactivity. In Eron, L. D. (1987). The development of aggressive behavior from the
L. M. Bloomingdale & J. M. Swanson (Eds.), Attention deficit dis- perspective of a developing behaviorism. American Psychologist, 42,
order (\o\. 4, pp. 203-228). Oxford, England: Pergamon Press. 435-442.
Barkley, R. A., Fischer, M., Edelbrock, C. S, & Smallish, L. (1990). The Estrada, P., Arsenic, W F, Hess, R. D., & Holloway, S. D. (1987). Affec-
adolescent outcome of hyperactive children diagnosed by research tive quality of the mother-child relationship: Longitudinal conse-
criteria: I. An 8-year prospective follow-up study. Journal of the Amer- quences for children's school-relevant cognitive functioning. Develop-
ican Academy of Child and Adolescent Psychiatry, 29, 546-557. mental Psychology, 23, 210-215.
Beitchman, J. H. (1985). Speech and language impairment and psychi- Farrington, D. P. (1979). Environmental stress, delinquent behavior,
152 STEPHEN P. HINSHAW

and convictions. In I. G. Sarason & C. D. Spielberger (Eds.), Stress Jorm, A. F, Share, D. L., Maclean, R., & Matthews, R. (1986). Cogni-
and anxiety (Vol. 6, pp. 93-107). Washington, DC: Hemisphere. tive factors at school entry predictive of specific reading retardation
Fischer, M., Barkley, R. A., Edelbrock, C. S., & Smallish, L. (1990). The and general reading backwardness: A research note. Journal of Child
adolescent outcome of hyperactive children diagnosed by research Psychology and Psychiatry, 27, 45-54.
criteria: II. Academic, attentional, and neuropsychological status. Jorm, A. F, Share, D. L, Matthews, R., & Maclean, R. (1986). Be-
Journal of Consulting and Clinical Psychology, 58, 580-588. haviour problems in specific reading retarded and general reading
Forness, S. R., Youpa, D., Hanna, G., Cantwell, D. P., & Swanson, J. M. backward children: A longitudinal study. Journal of Child Psychol-
(in press). Classroom instructional characteristics in attention defi- ogy and Psychiatry, 27, 33-43.
cit hyperactivity disorder: Comparison of pure and mixed sub- Kavale, K. A., & Forness, S. R. (1984). A meta-analysis assessing the
groups. Behavioral Disorders. validity of Wechsler Scale profiles and recategorization: Patterns or
Frick, P., Kamphaus, R. W, Lahey, B. B., Loeber, R., Christ, M. G., parodies? Learning Disability Quarterly, 7,136-156.
Hart, E., & Tannenbaum, L. E. (1991). Academic underachievement Kazdin, A. E. (1987). Treatment of antisocial behavior in children:
and the disruptive behavior disorders. Journal of Consulting and Current status and future directions. Psychological Bulletin, 102,
Clinical Psychology, 59, 289-294. 187-203.
Fuerst, D. R., Fisk, J. L., & Rourke, B. P. (1989). Psychosocial function- Kellam, S. G., Branch, J. D, Agrawal, D. C, & Ensminger, M. E. (1975).
ing of learning-disabled children: Replicability of statistically de- Mental health and going to school. Chicago: University of Chicago
rived subtypes. Journal of Consulting and Clinical Psychology, 57, Press.
275-280. Kupersmidt, J. B., & Coie, J. D. (1990). Preadolescent peer status, ag-
Gittelman, R., Mannuzza, S, Shenker, R., & Bonagura, N. (1985). Hy- gression, and school adjustment as predictors of externalizing prob-
peractive boys almost grown up: I. Psychiatric status. Archives of lems in adolescence. Child Development, 61,1350-1362.
General Psychiatry, 42, 937-947. Lahey, B. B., Piacentini, J. C, McBurnett, K., Stone, P., Hartdagen, S.,
Goyette, C. H., Conners, C. K., & Ulrich, R. F. (1978). Normative data & Hynd, G. (1988). Psychopathology in the parents of children with
on revised Conners Parent and Teacher Rating Scales. Journal of conduct disorder and hyperactivity. Journal of the American Acad-
Abnormal Child Psychology, 6, 221-236. emy of Child and Adolescent Psychiatry, 27,163-170.
Halperin, J. M., Gittelman, R., Klein, D. F., & Rudel, R. G. (1984). Lahey, B. B., Russo, M. F, Walker, J. L., & Piacentini, J. C. (1989).
Reading-disabled hyperactive children: A distinct subgroup of at- Personality characteristics of the mothers of children with disrup-
tention deficit disorder with hyperactivity? Journal of Abnormal tive behavior disorders. Journal of Consulting and Clinical Psychol-
Child Psychology, 12,1-14. ogy, 57,512-515.
Hawkins, J. D., & Lishner, D. M. (1987). Schooling and delinquency. In Lambert, N. M., Hartsough, C. S., & Zimmerman, I. L. (1976). The
E. H. Johnson (Ed.), Handbook of crime and delinquency prevention comparative predictive efficiency of intellectual and nonintellectual
(pp. 179-221). New York: Guilford Press. components of high school functioning. American Journal of Ortho-
Hays, W L. (1981). Statistics (3rd ed.). New York: Holt, Rinehart & psychiatry, 46,109-122.
Winston. Lambert, N. M., & Nicoll, R. C. (1977). Conceptual model for nonin-
Hetherington, E. M., & Martin, B. (1986). Family interaction patterns. tellectual behavior and its relationship to early reading achievement.
In H. C. Quay & J. S. Werry (Eds.), Psychopathological disorders of Journal of Educational Psychology, 69, 481-490.
childhood (ltd ed., pp. 332-390). New York: Wiley. Lambert, N. M., & Sandoval, J. (1980). The prevalence of learning
Hinshaw, S. P. (1987). On the distinction between attentional deficits/ disabilities in a sample of children considered hyperactive. Journal
hyperactivity and conduct problems/aggression in child psychopa- of Abnormal Child Psychology, 8, 33-50.
thology. Psychological Bulletin, 101, 443-463. Ledingham, J. E., & Schwartzman, A. E. (1984). A 3-year follow-up of
Hinshaw, S. P., Han, S., Erhardt, D., & Dressier, A. H. (1991). Internaliz- aggressive and withdrawn behavior in childhood: Preliminary find-
ing and externalizing behavior problems in preschool children: Con- ings. Journal of Abnormal Child Psychology, 12,157-168.
vergence between parent and teacher ratings and behavior observa- Loeber, R. (1990). Development and risk factors of juvenile antisocial
tions. Manuscript submitted for publication. behavior and delinquency. Clinical Psychology Review, 10,1-41.
Hinshaw, S. P., Morrison, D. C, Carte, E. T, & Cornsweet, C. (1987). Loeber, R., & Dishion, T. (1983). Early predictors of male delinquency:
Factorial dimensions of the Revised Behavior Problem Checklist: A review. Psychological Bulletin, 94, 68-99.
Replication and validation within a kindergarten sample. Journal of Loeber, R., & Lahey, B. B. (1989). Recommendations for research on
Abnormal Child Psychology, 15, 309-327. disruptive behavior disorders of childhood and adolescence. In B. B.
Hirschi, T. (1969). Causes of delinquency. Berkeley: University of Cali- Lahey & A. E. Kazdin (Eds.), Advances in clinical child psychology
fornia Press. (Vol. 12, pp. 221-251). New York: Plenum Press.
Hirschi, T., & Hindelang, M. J. (1977). Intelligence and delinquency: A Loeber, R., & Stouthamer-Loeber, M. (1987). Prediction. In H. C.
revisionist review. American Sociological Review, 42, 571-587. Quay (Ed.), Handbook of juvenile delinquency (pp. 325-382). New
Hogan, A. E., & Quay, H. C. (1984). Cognition in child and adolescent York: Wiley.
behavior disorders. In B. B. Lahey & A. E. Kazdin (Eds.), Advances Loney, J., Kramer, J., & Milich, R. S. (1981). The hyperactive child
in clinical child psychology (Vol. 7, pp. 1-34). New York: Plenum grows up: Predictors of symptoms, delinquency, and achievement at
Press. follow-up. In K. D. Gadow & J. Loney (Eds.), Psychosocial aspects of
Holborow, P. L., & Berry, P. S. (1986). Hyperactivity and learning dis- drug treatment for hyperactivity (pp. 381-415). Boulder, CO: West-
abilities. Journal of Learning Disabilities, 19, 426-431. view Press.
Howlin, P., & Rutter, M. (1987). The consequences of language delay Loney, J., & Milich, R. (1982). Hyperactivity, inattention, and aggres-
for other aspects of development. In W Yule & M. Rutter (Eds.), sion in clinical practice. In M. Wolraich & D. Routh (Eds.), Advances
Language development and disorders (pp. 271-294). Oxford, En- in developmental and behavioral pediatrics (Vol. 3, pp. 113-147).
gland: MacKeith. Greenwich, CT: JAI Press.
Huesmann, L. R., Eron, L. D., & Yarmel, P. W (1987). Intellectual Magnusson, D. (1988). Individual development from an interactional
functioning and aggression. Journal of Personality and Social Psy- perspective: A longitudinal study. Hillsdale, NJ: Erlbaum.
chology, 52, 232-240. Mann, V A., & Brady, S. (1988). Reading disability: The role of Ian-
EXTERNALIZING BEHAVIOR AND UNDERACHIEVEMENT 153

guage deficiencies. Journal of 'Consulting and Clinical Psychology, 56, Morrison, D, Mantzicopoulos, P., & Carte, E. (1989). Preacademic
811-816. screening for learning and behavior problems. Journal of the Ameri-
Mantzicopoulos, P., Morrison, D. C, Hinshaw, S. P., & Carte, E. T. can Academy of Child and Adolescent Psychiatry, 28,101-106.
(1989). Nonpromotion in kindergarten: The role of cognitive, percep- Offord, D. R., Alder, R. 1, & Boyle, M. H. (1986). Prevalence and
tual, visual-motor, behavioral, achievement, socioeconomic, and sociodemographic correlates of conduct disorder. American Journal
demographic characteristics. American Educational Research Jour- of Social Psychiatry, 6, 272-278.
nal, 26,107-121. Offord, D. R., Boyle, M. H., & Racine, Y. (1989). Ontario Child Health
Maughan, B., Gray, G., & Rutter, M. (1985). Reading retardation and Study: Correlates of disorder. Journal of the American Academy of
antisocial behavior: A follow-up into employment. Journal of Child Child and Adolescent Psychiatry, 28, 856-860.
Psychology and Psychiatry, 26, 741-758. Ohman, A., & Magnusson, D. (1987). An interactional paradigm for
McConaughy, S. H., Achenbach, T. M., &Gent, C. L. (1988). Multiaxial research in psychopathology. In D. Magnusson & A. Ohman (Eds.),
empirically based assessment: Parent, teacher, observational, cog- Psychopathology: An interactional perspective (pp. 3-21). San Diego,
nitive, and personality correlates of Child Behavior Profile types for CA: Academic Press.
6- to 11-year-old boys. Journal of Abnormal Child Psychology, 16, Olweus, D. (1983). Low school achievement and aggressive behavior in
485-509. adolescent boys. In D. Magnusson (Ed.), Human development: An
McGee, R., & Share, D. L. (1988). Attention deficit disorder-hyperac- interactional perspective (pp. 353-365). San Diego, CA: Academic
tivity and academic failure: Which comes first and what should be Press.
treated? Journal of the American Academy of Child and Adolescent Palfrey, J. S., Levine, M. D., Walker, D. K., & Sullivan M. (1985). The
Psychiatry, 27, 318-325. emergence of attention deficits in early childhood: A prospective
McGee, R., Share, D., Moffitt, T. E., Williams, S., & Silva, P. A. (1988). study. Developmental and Behavioral Pediatrics, 6, 339-348.
Reading disability, behaviour problems, and juvenile delinquency. Paternite, C. E., Loney, J., & Langhorne, J. E. (1976). Relationships
In D. H. Saklofske & S. B. G. Eysenck (Eds.), Individual differences in between symptomatology and SES-related factors in hyperkinetic/
children and adolescents: International perspectives (pp. 158-172). MBD boys. American Journal of Orthopsychiatry, 46, 291-301.
London: Hodder & Stoughton. Patterson, G. R. (1982). Coercive family process. Eugene, OR: Castalia.
McGee, R., Williams, S., Bradshaw, J., Chapel, J. L., Robins, A., & Patterson, G. R. (1990). Some comments about cognitions as causal
Silva, P. A. (1985). The Rutter scale for completion by teachers: Fac- variables. American Psychologist, 45, 984-985.
tor structure and relationships with cognitive abilities and family
Patterson, G. R., DeBaryshe, B. D, & Ramsey, E. (1989). A developmen-
adversity for a sample of New Zealand children. Journal of Child
tal perspective on antisocial behavior. American Psychologist, 44,
Psychology and Psychiatry, 26, 727-739.
329-335.
McGee, R., Williams, S., Share, D. L., Anderson, J., & Silva, P. A.
(1986). The relationship between specific reading retardation, gen- Phillips, J. C., & Kelly, D. H. (1979). School failure and delinquency:
eral reading backwardness, and behavioural problems in a large sam- Which causes which? Criminology, 17,194-207.
ple of Dunedin boys: A longitudinal study from five to eleven years. Plomin, R., Loehlin, J. C, & DeFries, J. C. (1985). Genetic and environ-
Journal of Child Psychology and Psychiatry, 27, 597-610. mental components of "environmental" influences. Developmental
McGee, R., Williams, S., & Silva, P. A. (1984a). Background character- Psychology, 21, 391-402.
istics of aggressive, hyperactive, and aggressive-hyperactive boys. Porter, J., & Rourke, B. P. (1985). Socio-emotional functioning of learn-
Journal of the American Academy of Child Psychiatry, 23, 280-284. ing disabled children: A subtypal analysis of personality patterns. In
McGee, R., Williams, S., & Silva, P. A. (1984b). Behavioral and develop- B. P. Rourke (Ed.), Neuropsychology of learning disabilities: Essen-
mental characteristics of aggressive, hyperactive, and aggressive-hy- tials of subtype analysis (pp. 257-279). New York: Guilford Press.
peractive boys. Journal of the American Academy of Child Psychiatry, Professional Group for ADD and Related Disorders. (1991). Response
23, 270-279. to notice of inquiry regarding special education for children with ADD.
McGee, R., Williams, S., & Silva, P. (1985). Factor structure and corre- Unpublished manuscript, Department of Pediatrics, University of
lates of ratings of inattention, hyperactivity, and antisocial behavior California, Irvine.
in a large sample of 9-year-old children from the general population. Quay, H. C. (1979). Classification. In H. C. Quay & J. S. Werry (Eds.),
Journal of Consulting and Clinical Psychology, 53, 480-490. Psychopathological disorders of childhood (2nd ed., pp. 1-42). New
McMichael, P. (1979). The hen or the egg? Which comes firstantiso- York: Wiley.
cial emotional disorders or reading disability? British Journal of Edu- Quay, H. C. (1983). A dimensional approach to behavior disorder: The
cational Psychology, 49, 226-238. Revised Behavior Problem Checklist. School Psychology Review, 12,
Milich, R., & Dodge, K. A. (1984). Social information processing in 244-249.
child psychiatry populations. Journal of Abnormal Child Psychology, Quay, H. C. (1986). Classification. In H. C. Quay & J. S. Werry (Eds.),
72,471-489. Psychopathological disorders of childhood (3rd ed., pp. 1-34). New
Milich, R., & Landau, S. (1989). The role of social status variables in York: Wiley.
differentiating subgroups of hyperactive children. In L. M. Bloom- Quay, H. C. (1987). Intelligence. In H. C. Quay (Ed.), Handbook of
ingdale & J. M. Swanson (Eds.), Attention deficit disorder (Vol. 4, pp. juvenile delinquency (pp. 106-117). New \brk: Wiley.
1-16). Oxford, England: Pergamon Press. Reeves, J. C, Werry, J. S., Elkind, G. S., & Zametkin, A. (1987). Atten-
Moffitt, T. E. (1990). Juvenile delinquency and attention deficit dis- tion deficit, conduct, oppositional, and anxiety disorders in chil-
order: Boys' developmental trajectories from age 3 to age 15. Child dren: II. Clinical characteristics. Journal of the American Academy of
Development, 61, 893-910. Child and Adolescent Psychiatry, 26,144-155.
Moffitt, T. E., & Silva, P. A. (1988a). IQ and delinquency: A direct test Rey, J. M., Bashir, M. R., Schwarz, M., Richards, I. N, Plapp, J. M., &
of the differential detection hypothesis. Journal of Abnormal Psy- Stewart, G. W (1988). Oppositional disorder: Fact or fiction? Journal
chology, 97, 330-333. of the American Academy of Child and Adolescent Psychiatry, 27,
Moffitt, T. E., & Silva, P. A. (1988b). Self-reported delinquency, neuro- 157-162.
psychological deficit, and history of attention deficit disorder. Jour- Richman, N., Stevenson, J., & Graham, P. (1982). Preschool to school: A
nal of Abnormal Child Psychology, 16, 553-569. behavioural study. San Diego, CA: Academic Press.
154 STEPHEN P. HINSHAW

Robins, L. N. (1970). The adult development of the antisocial child. Sergeant, J. A. (1989). In search of processing deficits of attention in
Seminars in Psychiatry, 6, 420-434. ADD-H children. In L. M. Bloomingdale & J. M. Swanson (Eds.),
Robins, L. N. (1979). Follow-up studies. In H. C. Quay & J. S. Werry Attention deficit disorder (Vol. 4, pp. 255-272). Oxford, England:
(Eds.), Psychopathological disorders of childhood (2nd ed., pp. 483- Pergamon Press.
513). New York: Wiley. Sergeant, J., & van der Meere, J. (1990). Convergence of approaches in
Rose, S. L., Rose, S. A., & Feldman, J. F. (1989). Stability of behavior localizing the hyperactivity deficit. In B. B. Lahey & A. E. Kazdin
problems in very young children. Development and Psychopathol- (Eds.), Advances in clinical child psychology (Vol. 13, pp. 207-246).
ogy, 1, 5-19. New \ork: Plenum Press.
Rourke, B. P. (1985). Overview of learning disability subtypes. In B. P. Share, D. L., McGee, R., McKenzie, D., Williams, S., & Silva, P A.
Rourke (Ed.), Neuropsychology of learning disabilities: Essentials of (1987). Further evidence relating to the distinction between specific
subtype analysis (pp. 3-14). New \brk: Guilford Press. reading retardation and general reading backwardness. British Jour-
Rourke, B. P. (1988). Socioemotional disturbances of learning disabled nal of Developmental Psychology, 5, 35-44.
children. Journal of Consulting and Clinical Psychology, 56,801-810. Share, D. L., McGee, R., & Silva, P. A. (1989). IQ and reading progress:
Rutter, M. (1967). A children's behaviour questionnaire for completion A test of the capacity notion of IQ. Journal of the American Academy
by teachers: Preliminary findings. Journal of Child Psychology and of Child and Adolescent Psychiatry, 28, 97-100.
Psychiatry, 8,1-11. Shaywitz, S. E., Shaywitz, B. A., Fletcher, J. M., & Escobar, M. D.
Rutter, M. (1974). Emotional disorder and educational underachieve- (1990). Prevalence of reading disability in boys and girls: Results of
ment. Archives of Disease in Childhood, 49, 249-256. the Connecticut Longitudinal Study. Journal of the American Medi-
cal Association, 264, 998-1002.
Rutter, M. (1989). Isle of Wight revisited: Twenty-five years of child
Silva, P. A. (1987). Epidemiology, longitudinal course, and some asso-
psychiatric epidemiology. Journal of the American Academy of Child ciated factors: An update. In W Yule & M. Rutter (Eds.), Language
and Adolescent Psychiatry, 28, 633-653. development and disorders (pp. 1-15). Oxford, England: MacKeith.
Rutter, M., & Giller, H. (1983). Juvenile delinquency: Trends and per- Silva, P. A., McGee, R, & Williams, S. (1985). Some characteristics of
spectives. New \brk: Guilford Press. 9-year-old boys with general reading backwardness or specific read-
Rutter, M., & Graham, P. (1970). Psychiatric aspects of intellectual and ing retardation. Journal of Child Psychology and Psychiatry, 26,407-
educational retardation. In M. Rutter, J. Tizard, & K. Whitmore 421.
(Eds.), Education, health, and behaviour (pp. 102-112). London: Silver, L. B. (1990). Attention-deficit hyperactivity disorder: Is it a
Longmans. learning disability or a related disorder? Journal of Learning Disabili-
Rutter, M., Tizard, J, & Whitmore, K. (Eds.) (1970). Education, health, ties, 23, 394-397.
and behaviour. London: Longmans. Sleator, E. K., & Ullmann, R. K. (1981). Can the physician diagnose
Rutter, M., Tizard, J., Yule, W, Graham, P., & Whitmore, K. (1976). hyperactivity in the office? Pediatrics, 67,13-17.
Research report: Isle of Wight studies, 1964-1974. Psychological Spreen, O. (1988). Prognosis of learning disability. Journal of Consult-
Medicine, 6, 313-332. ing and Clinical Psychology, 56, 836-842.
Rutter, M., & Yule, W (1970). Reading retardation and antisocial be- Stevenson, J. (1984). Predictive value of speech and language screen-
haviourThe nature of the association. In M. Rutter, J. Tizard, & K. ing. Developmental Medicine & Child Neurology, 26, 528-538.
Whitmore (Eds.), Education, health, and behaviour (pp. 240-255). Stevenson, J., & Fredman, G. (1990). The social environmental corre-
London: Longmans. lates of reading ability. Journal of Child Psychology and Psychiatry,
Rutter, M., & Yule, W (1975). The concept of specific reading retarda- 57,681-698.
tion. Journal of Child Psychology and Psychiatry, 16,181-197. Stevenson, J., & Richman, N. (1978). Behavior, language, and develop-
Rutter, M., Yule, B., Quinton, D., Rowlands, O., Yule, W, & Berger, M. ment in three-year-old children. Journal of Autism and Childhood
(1974). Attainment and adjustment in two geographical areas: III Schizophrenia, 8, 299-313.
Some factors accounting for area differences. British Journal of Psy- Stevenson, J., Richman, N., & Graham, P. (1985). Behaviour problems
chiatry, 125, 520-533. and language abilities at three years and behavioural deviance at
Saal, F. E., Downey, R. G, & Lahey, M. A. (1980). Rating the ratings: eight years. Journal of Child Psychology and Psychiatry, 26,215-230.
Assessing the psychometric quality of rating data. Psychological Stone, W L., & LaGreca, A. M. (1990). The social status of children
with learning disabilities: A reexamination. Journal of Learning Dis-
Bulletin, 55,413-428.
abilities, 23, 32-37.
Sampson, O. C. (1966). Reading and adjustment: A review of the litera-
Stott, D. H. (1981). Behaviour disturbance and failure to learn: A study
ture. Educational Research, 8,184-190. of cause and effect. Educational Research, 23,163-172.
Sandberg, S. T, Wieselberg, M., & Shaffer, D. (1980). Hyperkinetic and Strang, J. D., & Rourke, B. P. (1985). Arithmetic disability subtypes:
conduct problem children in a primary school population: Some The neuropsychological significance of specific arithmetic impair-
epidemiological considerations. Journalof"ChildPsychology andPsy- ment in childhood. In B. P. Rourke (Ed.), Neuropsychology of learn-
chiatry, 2.7,293-311. ing disabilities: Essentials of subtype analysis (pp. 167-183). New
Schachar, R., Rutter, M., & Smith, A. (1981). The characteristics of York: Guilford Press.
situationally and pervasively hyperactive children: Implications for Sturge, C. (1982). Reading retardation and antisocial behaviour. Jour-
syndrome definition. Journal of Child Psychology and Psychiatry, 22, nal of Child Psychology and Psychiatry, 23,21-31.
375-392. Sweeney, J. E., & Rourke, B. P. (1985). Spelling disability subtypes. In
Schachar, R., & Wachsmuth, R. (1990). Hyperactivity and parental B. P. Rourke (Ed.), Neuropsychology of learning disabilities: Essen-
psychopathology. Journal of Child Psychology and Psychiatry, 31, tials of subtype analysis (pp. 147-183). New York: Guilford Press.
381-392. Szatmari, P., Boyle, M., & Offord, D. R. (1989). ADDH and conduct
Schonfeld, I. S. (1990). A developmental perspective and antisocial disorder: Degree of diagnostic overlap and difference among corre-
behavior: Cognitive functioning. American Psychologist, 45, 983- lates. Journal of the American Academy of Child and Adolescent Psy-
984. chiatry, 28, 865-872.
Schonfeld, I. S., Shaffer, D., O'Connor, P., & Portnoy, S. (1988). Conduct Szatmari, P., Offord, D. R., & Boyle, M. H. (1989). Ontario Child
disorder and cognitive functioning: Testing three causal hypotheses. Health Study: Prevalence of attention deficit disorder with hyperac-
Child Development, 59, 993-1007. tivity. Journal of Child Psychology and Psychiatry, 30, 219-230.
EXTERNALIZING BEHAVIOR AND UNDERACHIEVEMENT 155

Szatmari, P., Offord, D. R., Siegel, L. S., Finlayson, M. A. J., & Tuff, L. Werner, E., & Smith, P. (1982). Vulnerable but invincible: A study of
(1990). The clinical significance of neurocognitive impairments resilient children. New "Vbrk: McGraw-Hill.
among children with psychiatric disorders: Diagnosis and situa- Werry, J. S., Elkind, G. S., & Reeves, J. C. (1987). Attention deficit,
tional specificity. Journal of Child Psychology and Psychiatry, 31, conduct, opposit ional, and anxiety disorders in children: III. Labora-
287-299. tory differences. Journal of Abnormal Child Psychology, 15, 409-
Tallal, P. (1980). Language disabilities in children: A perceptual or 428.
linguistic deficit? Journal ofPediatric Psychology, 5, 127-140. White, J. L., Moffitt, T. E., & Silva, P. A. (1989). A prospective replica-
Tallal, P., Dukette, D., & Curtiss, S. (1989). Behavioral/emotional pro- tion of the protective effects of IQ in subjects at high risk for juvenile
files of preschool language-impaired children. Development andPsy- delinquency. Journal of Consulting and Clinical Psychology, 57, 719-
chopathology, 1, 51-67. 724.
Taylor, H. G. (1988). Learning disabilities. In E. J. Mash & L. G. Terdal Wilson, J. Q., & Herrnstein, R. J. (1985). Crime and human nature. New
(Eds.), Behavioral assessment of childhood disorders (2nd ed., pp. York: Simon & Schuster.
402-450). New York: Guilford Press. Yule, W, & Rutter, M. (1985). Reading and other learning difficulties.
Wagner, R. K., & Torgeson, J. (1987). The nature of phonological pro- In M. Rutter & L. Hersov (Eds.), Child and adolescent psychiatry:
cessing and its causal role in the acquisition of reading skills. Psycho- Modern approaches (pp. 444-464). Oxford, England: Blackwell Sci-
logical Bulletin, 101,192-212. entific.
Weiss, G., & Hechtman, L. T. (1986). Hyperactive children grown up:
Empirical findings and theoretical considerations. New York: Guil- Received September 25,1990
ford Press. Accepted May 9,1991

Instructions to Authors
Authors should prepare manuscripts according to the Publication Manual of the American
Psychological Association (3rd ed.). All manuscripts must include an abstract containing a
maximum of 960 characters and spaces (which is approximately 120 words) typed on a separate
sheet of paper. Typing instructions (all copy must be double-spaced) and instructions on pre-
paring tables, figures, references, metrics, and abstracts appear in the Manual. Also, all manu-
scripts are subject to editing for sexist language.
Authors reporting original data will be required to state in writing that they have complied
with APA ethical standards in the treatment of their sample, human or animal, or to describe
the details of treatment. (A copy of the APA Ethical Principles may be obtained from the Ethics
Office, American Psychological Association, 750 First Street, N.E., Washington, DC 20002.)
Masked review will be first an author's option. If masked review is not requested in a cover
letter, it will become the prerogative of the processing editor. Authors requesting masked
review are requested to include with each copy of the manuscript a cover sheet, which shows the
title of the manuscript, the authors' names and institutional affiliations, and the date the
manuscript is submitted. The first page of the manuscript should omit the author's name and
affiliation but should include the title of the manuscript and the date it is submitted. Footnotes
containing information pertaining to the authors' identity or affiliations should be on separate
pages. Every effort should be made to see that the manuscript itself contains no clues to the
authors' identity.
APA policy prohibits an author from submitting the same manuscript for concurrent consid-
eration by two or more publications. APA policy also prohibits duplicate publication, that is,
publication of a manuscript that has already been published in whole or in substantial part
elsewhere. Authors have an obligation to consult journal editors if there is any chance or
question that the paper might not be suitable for publication in an APA journal. Authors who
do not inform the editor of the prior publication history of previously published material will
be engaging in unethical behavior. Also, authors of manuscripts submitted to APA journals are
expected to have available their raw data throughout the editorial review process and for at least
5 years after the date of publication.
Submit five copies of each manuscript. All copies should be clear, readable, and on paper of
good quality. A dot matrix or unusual typeface is acceptable only if it is clear and legible. In
addition to addresses and phone numbers, authors should supply electronic mail addresses and
fax numbers, if available, for potential use by the editorial office and later by the production
office. Authors should keep a copy of the manuscript to guard against loss. Mail manuscripts to
the Editor, Robert J. Sternberg, Psychological Bulletin, Department of Psychology, Yale Univer-
sity, P.O. Box 11A Yale Station, New Haven, CT 06520-7447.

You might also like