PIVEN,

BROADER
Am JPALMER,
Psychiatry
AUTISM JACOBI,
154:2,
PHENOTYPE
February
ET AL. 1997

Broader Autism Phenotype: Evidence From a Family History

Study of Multiple-Incidence Autism Families
Joseph Piven, M.D., Pat Palmer, Ph.D., Dinah Jacobi, B.A.,
Debra Childress, B.S., and Stephan Arndt, Ph.D.

Objective: Studies of families ascertained through a single autistic proband suggest that the
genetic liability for autism may be expressed in nonautistic relatives in a phenotype that is
milder but qualitatively similar to the defining features of autism. The objective of this study
was to examine behaviors that may define this broader phenotype in relatives ascertained
through two autistic siblings. Method: The authors used a semistructured family history in-
terview to compare the rates of social and communication deficits and stereotyped behaviors
in relatives ascertained through two autistic siblings (families with multiple-incidence au-
tism; 25 families) with the rates in relatives of Down syndrome probands (30 families).
Results: Higher rates of social and communication deficits and stereotyped behaviors were
found in the relatives in the families with multiple-incidence autism. Conclusions: These data
suggest that further studies should be undertaken to delineate the boundaries of the broader
autism phenotype and that this broader phenotype should be included in some future genetic
analyses of this disorder.
(Am J Psychiatry 1997; 154:185–190)

A utism is a behavioral syndrome defined by the

presence of social deficits, communication abnor-
malities, stereotyped or repetitive behaviors, and a
similar to but more broadly defined than that which
defines autism occurs more commonly in relatives of
autistic individuals than in the general population. In
characteristic course. There is now considerable evi- the first twin study of autism, in addition to finding a
dence from family and twin studies that, for a subgroup pairwise concordance rate for autism of 36% among 11
of autistic individuals, the etiology is mainly genetic (1). monozygotic pairs, Folstein and Rutter (3) reported an
The risk of recurrence of autism in families (i.e., the even higher concordance rate (82%) for a more broadly
frequency of autism in subsequently born siblings) is defined cognitive impairment that included autism,
estimated at 6%–8%, or up to 200 times the risk in the mental retardation, language delay, reading disorder,
general population (2). Three twin studies of geo- spelling disorder, or articulation disorder. August et al.
graphically defined populations (3–5) detected pairwise (7) reported further possible evidence to support the re-
concordance rates of approximately 65% (the average lationship of cognitive disorders to autism, showing the
over the three studies) and 0%, in monozygotic and familial aggregation of these disorders in the siblings of
dyzygotic pairs, respectively, producing a heritability autistic probands. However, these investigators ques-
estimate of over 90%. Further, there is no convincing tioned whether the aggregation of cognitive disorders in
evidence that perinatal factors play an important role in autism families was confounded by the co-occurrence
the etiology of most cases of autism (6). of mental retardation in the majority of the autistic pro-
Although the importance of genetic factors in autism bands. A subsequent study by Freeman et al. (8), di-
has been firmly established, the definition of the pheno- rectly testing the relatives of autistic probands and com-
type for use in genetic studies continues to be in ques- paring data on them to published norms, failed to
tion. A number of family and twin studies have sug- detect higher rates of cognitive disorders in the autism
gested that a behavioral phenotype that is qualitatively relatives. Similarly, Szatmari et al. (9) reported no dif-
ferences on direct testing in a comparison of the rela-
tives of probands with pervasive developmental disor-
Received Nov. 29, 1995; revision received July 15, 1996; accepted der and the relatives of Down syndrome probands.
Sept. 3, 1996. From the Department of Psychiatry, University of Iowa
College of Medicine. Address reprint requests to Dr. Piven, 1875 John
More recently, Szatmari et al. (10) used the family his-
Pappajohn Pavilion, Department of Psychiatry, University of Iowa tory method and also failed to find a higher rate of cog-
Hospitals and Clinics, Iowa City, IA 52242-1057; jpiven@blue.weeg. nitive deficits in relatives of probands with pervasive
uiowa.edu (e-mail). developmental disorder.
Supported by NIMH grants MH-51217 and MH-01028. Accompanying the reports of possible cognitive deficits
The authors thank Edwin Cook, M.D., Elizabeth Reeves, M.D.,
Debra Scott-Miller, Gerald Gritzmacher, and the Iowa Department in relatives of autistic individuals have been parallel re-
of Education for help with the ascertainment of families for this study. ports of the aggregation of social deficits in family mem-

Am J Psychiatry 154:2, February 1997 185

BROADER AUTISM PHENOTYPE

bers. Wolff et al. (11), blind to proband diagnosis, inter- relatives in families with multiple-incidence autism pro-
viewed the parents of autistic children and the parents of vide a potentially important study group for exploring
nonautistic mentally retarded comparison subjects and the boundaries of the phenotype in autism.
found that the parents of the autistic children were more
often judged to lack emotional responsiveness and empa-
thy, show impaired rapport with the examiner, and have METHOD
histories of oversensitivity to experience, special interest
patterns, and oddities of social communication. Landa et Selection of Autism Families
al. (12, 13) reported significant differences between par-
Families with at least two autistic children were ascertained for this
ents of autistic and Down syndrome probands who were study through a systematic search for all such multiple-incidence
rated blindly on measures of social language use and autism families in Iowa and from families known to two tertiary
spontaneous narrative discourse. Gillberg (14), in a study evaluation centers for autism in the Midwest at the start of the study.
of the parents of 23 children with Asperger syndrome, The goal of this systematic ascertainment scheme was to reduce any
potential bias with respect to familial aggregation of possibly related
reported social deficits in 11 of the 23 fathers that were disorders, including social and communication deficits, stereotyped
similar to, but milder than, those seen in Asperger syn- behaviors, and psychiatric disorders. Families of autistic probands
drome. In a study using best-estimate ratings of subjects were eligible for this study if 1) two children (aged 4–30 years)
and informants directly interviewed with a semistruc- showed evidence of autism, either on the basis of a previous clinical
diagnosis or, in the case of public school screening, on the basis of an
tured personality interview, Piven et al. (15) detected sig- experienced teacher’s behavioral observations; and 2) a review of
nificantly higher rates of social deficits in the parents of medical records indicated that neither proband had evidence of a ma-
autistic children than in the parents of children with jor co-occurring medical condition thought possibly to be etiologi-
Down syndrome. cally related to autism, such as tuberous sclerosis, neurofibromatosis,
phenylketonuria, a chromosomal anomaly identified on karyotype or
Most recently, Bolton et al. (16) used a semistructured fragile X screening, or noteworthy CNS injury (1).
family history interview, designed to examine features Iowa families. Through a medical record review of patients seen over
hypothesized to be part of a more broadly defined autism the last 24 years in the child psychiatry clinic at the University of Iowa
phenotype, to examine the family histories of the first-de- and currently living in Iowa, 23 families were identified as possibly hav-
gree relatives of 99 autistic and 36 Down syndrome pro- ing two autistic children. These families were recontacted regarding par-
ticipation in the study. Four families refused to participate, and one
bands. The results indicated that the relatives of the could not be located, leaving 18 potential families for further screening.
autistic probands had significantly higher rates of com- Letters requesting referral of families with at least two children sus-
munication and social deficits and stereotyped behaviors pected of having autism were subsequently sent to all pediatricians and
than the relatives of the Down syndrome probands. Simi- family practitioners in Iowa (N=1,260). After two mailings, 79% of the
physicians responded, identifying 28 unique families with potential mul-
larly, in their twin study of autism Bailey et al. (5) found tiple-incidence autism sibships. Eleven families were already known
a 92% concordance rate for the presence of either a social through our medical record review at the University of Iowa. Six were
or cognitive abnormality in monozygotic twins, com- excluded on the basis of a telephone discussion with the referring phy-
pared to a 10% rate in dizygotic twins. In contrast to sician or a review of medical records that indicated that a diagnosis of
these results, however, is the report by Szatmari et al. autism was unlikely. One pair of adopted siblings met the diagnostic
criteria for the study but were excluded because the biological parents
(10), who used the same family history schedule used by were unavailable for participation. Four families refused to participate
Bolton et al. (16) and found no difference in the rate of or could not be located. Therefore, of the 28 potential multiple-inci-
social or communication deficits between comparison dence sibships identified by the Iowa pediatricians and family practitio-
subjects and relatives ascertained through a proband ners, six were left for further screening.
During the first six months of 1994, all public schools in Iowa were
with pervasive developmental disorder. systematically screened by area special education directors, who con-
In the present study we used the family history method tacted the schools in their districts to learn of sibships with potential
to examine rates of a more broadly defined autism phe- multiple-incidence autism. Sixteen sibships were identified. Fourteen
notype (as defined previously by Bolton et al. [16]) in a were already known to us, and one family refused further contact,
leaving one sibship with potential multiple-incidence autism.
sample of relatives ascertained through two autistic sib- At this level of screening, 25 sibships with potential multiple-inci-
lings (multiplex autism families) and comparison sub- dence autism were identified in Iowa. After direct evaluation (to be
jects. To our knowledge, rates of the broader autism phe- described), seven families were excluded as both probands did not
notype have not been previously reported for families meet the study criteria for autism, leaving 18 families. No proband
with multiple-incidence autism. These families offer sev- was found on physical examination to have evidence of a major co-
occurring medical condition. Of the 36 identified probands, at least
eral advantages over families ascertained through a single one in each of the 18 multiple-incidence sibships had been cytogeneti-
autistic proband. Probands in multiple-incidence autism cally tested for fragile X or was subsequently tested before entry into
families are less likely than those in single-incidence fami- our study. All subjects were negative for the fragile X anomaly.
lies to have autism as a result of nongenetic causes (see Non-Iowa families. At the start of this study five families with mul-
review by Piven and Folstein [1]) and are therefore likely tiple-incidence autism were known to Dr. Edwin Cook at the Univer-
sity of Chicago and were referred to our study. All agreed to partici-
to represent a more etiologically homogeneous sample pate. Four families known to Dr. Elizabeth Reeves at the St. Paul-
than probands from families ascertained through a single Ramsey County Hospital in Minnesota were also referred to our study;
autistic proband. In addition, relatives ascertained two refused to participate, leaving a total of seven families with mul-
through families with multiple-incidence autism may tiple-incidence autism ascertained from outside of Iowa. Direct ex-
amination (to be described) indicated that they all met the study cri-
have a higher genetic liability for autism than relatives teria for autism, and none was excluded because of a major medical
ascertained through families with single-incidence autism condition or cytogenetic evidence of fragile X syndrome.
or pervasive developmental disorder. For these reasons, Final sample. The final 25 multiplex autism families included 42

186 Am J Psychiatry 154:2, February 1997

 PIVEN, PALMER, JACOBI, ET AL.

male and eight female autistic probands ranging in age from 4 to 28 Family history interview. The parents were interviewed through
years. Adequate estimates of performance IQ were available from the use of a standardized family history interview—the Family History
medical records of 45 of the 50 probands. The following IQ meas- Interview for Developmental Disorders of Cognition and Social Func-
ures, if administered by a psychologist, were considered adequate for tioning—to query for the presence of a range of abnormalities that
estimation of performance IQ: the Wechsler Intelligence Scale for have been previously reported to occur at high rates in members of
Children—Revised (17), the Wechsler Intelligence Scale for Chil- families of autistic probands or have been hypothesized to possibly
dren—III (18), the Wechsler Adult Intelligence Scale—Revised (19), be genetically related to autism. This instrument was jointly devel-
the Leiter International Performance Scale (20), and the Merrill-Pal- oped by researchers at the Institute of Psychiatry in London (under
mer Scale of Mental Tests (21). When results of multiple tests were the direction of Professor Michael Rutter) and the Johns Hopkins
available, the test (with priority as indicated in the preceding list) per- University (under the direction of Dr. Susan Folstein) for use in a
formed closest to 12 years of age was used for estimating performance cross-national collaborative family study of autism. This interview
IQ. The performance IQs of 51% of the subjects were 70 or higher, includes additional probes for a range of developmental characteris-
22% were 50–69, 27% were 30–49, and none was less than 30. Five tics (including motor development, IQ, communication skills, aca-
probands were felt to not have had adequate testing at the time this demic skills, and social behaviors) and psychiatric disorders and
study was undertaken, either because the test used was inappropriate symptoms. Questions about adult and childhood functioning are also
or the proband had difficulty taking the test. Resources were not asked separately. Characteristics are generally rated as absent (rat-
available to attempt further testing of these five individuals. ing=0), mild or probably present (rating=1), or severe or definitely
After complete description of the study to the subjects, written in- present (rating=2). A more detailed description of this interview
formed consent was obtained from the adults and written informed schedule has been provided by Bolton et al. (16).
assent was obtained from the minors. The items from the family history interview to be examined in this
study were those that were contrasted in the autistic and comparison
Selection of Comparison Families subjects in the autism family study of Bolton et al. (16). In the study
by Bolton et al., items were selected for comparison on the basis of a
confirmatory factor analysis demonstrating loading of items onto
Thirty families each having a child with Down syndrome secon- three principal factors that parallel the defining features of autism:
dary to a nondysjunction of chromosome 21 constituted the compari- deficits in communication, deficits in reciprocal social interaction,
son group in this study. The rationale for choosing this group was our and the presence of selected stereotyped or repetitive behaviors. With
need to control for the effect of caring for a handicapped child on the guidance from the factor analysis and conceptual notions about a
emotional and social functioning of parents and siblings. Also, rela- latent autism phenotype, relevant items on the family history inter-
tives of a child with Down syndrome would not be expected to have view were combined to produce operational definitions for the three
a greater genetic liability, over that of the general population, for so- domains of behavior examined. Except for some minor changes, the
cial or communication deficits or stereotyped behaviors—the behav- definitions used in the present study for communication and social
ioral variables of interest in this study. deficits and stereotyped behaviors are identical to those used in the
An attempt was made to obtain equal numbers of families in each of Bolton et al. study; these are listed in table 1. In general, a subject was
three proband age groups: 4–12, 13–18, and >18 years. Initially, a letter considered affected in a particular behavioral domain if at least two
to parents was sent home with children in the public schools in eastern behaviors were present to a mild or probable degree (i.e., had a rating
Iowa. Nine families, all in the lowest proband age group, were recruited of 1) or if at least one behavior in the domain was rated as severe or
by this means. Using a second strategy, we randomly recruited the re- definitely present (i.e., had a rating of 2). Several items included in the
maining 21 families (70%) from a list of the families of newborns diag- Bolton et al. study (16) were not included in the version of the inter-
nosed with Down syndrome at the University of Iowa who lived within view used in this study (including the items “childhood social dys-
150 miles of the university. To obtain comparable numbers of families function,” “adult social dysfunction and isolation,” and “adult re-
in the three proband age groups, preference was given first to families petitive behaviors”) and therefore are not included in our analyses. In
who had probands in the middle and oldest age groups. The final group addition, the item “obsessions and compulsions” was not included in
of Down syndrome families included 13 male and 17 female Down syn- our definitions of stereotyped behaviors on the basis of a previous
drome probands. The probands’ ages ranged from 2 to 27 years. study (25) in which we failed to detect high rates of obsessive-com-
After description of the study, written informed consent was ob- pulsive disorder in the parents of autistic individuals.
tained from the adults, and written informed assent was obtained The parents in each family were interviewed about themselves and
from the minors. about the siblings, grandparents, aunts and uncles, and first cousins
of the probands. Because of the extensive involvement with families
Assessment of Autistic Probands required in this study and the limited resources available, we could
not remain blind to the case-control status of the subjects. To limit
Diagnosis. Parents of all subjects were interviewed regarding the sub- any bias resulting from this practice, case vignettes based on the in-
jects’ diagnosis with a standardized interview, the Autism Diagnostic formation gathered with the family history interview were composed
Interview (22). An algorithm constructed for use with the Autism Diag- for all parents and rated blindly by one of us (J.P.).
nostic Interview (which uses the ICD-10 criteria for autism [23]) has
been shown to adequately discriminate autistic and nonautistic IQ- Analysis
matched subjects (22). For 10 videotaped interviews, adequate inter-
rater agreement (kappa <0.90) on diagnoses of autism made by using Sample characteristics for the autistic and comparison subjects
the algorithm was established by all raters before the start of data col- (e.g., parental age, education level) were compared by using simple
lection. In addition, the probands were directly assessed by using the statistics. The analysis of family data required that we account for the
Autism Diagnostic Observation Schedule (24), a structured observation fact that relatives within the same family (aunts/uncles, grandparents,
and interview schedule developed to aid in the diagnosis and assessment and parents) are not statistically independent. To take this into ac-
of autistic individuals. The information from the Autism Diagnostic Ob- count in the analysis, we treated the family as the unit of analysis. For
servation Schedule functioned as a check on the proband’s current be- example, the presence of a history of social deficits in one parent or
havior as reported by the parents on the Autism Diagnostic Interview. the other was treated as the independent variable in a logistic regres-
Physical examination. All subjects were evaluated in a screening sion predicting family diagnosis (autism versus Down syndrome).
neurodevelopmental examination for evidence of major neurological Similarly, logistic regression was used in case-control comparisons by
impairment or medical conditions thought to be etiologically related family of aunts and uncles and by family of grandparents. There was
to autism (listed earlier). Almost all subjects had been previously no significant difference between the autism families and the Down
screened through a medical evaluation at a tertiary care center and syndrome families in either the number of aunts and uncles (t=0.26,
not found to have evidence of any exclusionary criteria for this study. df=53, p=0.78) or the number of grandparents (t=1.41, df=53,
No subject was excluded on the basis of our additional neurodevelop- p=0.16). Results were considered significant if they passed the p<0.05
mental screening examination. level of significance (two-tailed).

Am J Psychiatry 154:2, February 1997 187

BROADER AUTISM PHENOTYPE

TABLE 1. Definitions of Behavioral Deficits Used in Family History Study of Multiple-Incidence Autism groups separately on social
Familiesa deficits, communication defi-
Behavioral Deficit Definition cits, and stereotyped behav-
Communication deficits
iors are presented in table 2.
Language delay No single words at age 2 years; no phrases by 33 months The autism parents showed
Reading retardation (8 years or older) Educational assessment for reading problems; remedial significantly higher rates of
help social deficits and stereo-
Articulation disorder (5 years or Speech therapy; strangers unable to understand at age 5 typed behaviors but not com-
older)
Spelling difficulties (8 years or older) Frequent errors with common words munication deficits than did
Social deficits the Down syndrome parents.
Lack of affection (4 and 5 years) Little or no spontaneous affection One autism father met the cri-
Impaired social play (1 to 6 years) Little to-and-fro social play teria for autism on the basis
Impaired friendships (6 to 16 years) Limited or no friendships
Lack of conversation (4 to 15 years) Limited or no to-and-fro conversation
of current behavior, but no
Lack of friends (adulthood) Few or no friends informants could be obtained
Impaired conversation (adulthood) Limited or no conversation to verify the presence of autis-
Social inappropriateness (adulthood) Striking, frequent inappropriate behavior tic behavior in childhood.
Stereotyped behaviors To examine these differ-
Circumscribed interests (16 years or All-encompassing interest that is unusual in its intensity
older) and exclusion of other activities but is not odd or
ences in more distantly related
socially inappropriate; all-encompassing odd or socially relatives and attempt replica-
inappropriate interest that is unusual in its intensity and tion of the findings for the
exclusion of other activities parents, we similarly com-
Rigidity (16 years or older) Rigid or perfectionistic style that is commented on by
others outside the family but not associated with
pared the grandparents and the
impairment; rigid or perfectionistic style of behavior that aunts and uncles in the autism
is associated with impairment and Down syndrome families
aItems were selected from the Family History Interview for Developmental Disorders of Cognition and (autism families: 96 grandpar-
Social Functioning. Communication deficits and stereotyped behaviors were considered present if one ents, 145 aunts and uncles;
item was given a rating of 2 or two items were given ratings of 1. Social deficits were considered present Down syndrome families: 120
if one item was rated as “definitely present” or two items were rated as “probably present.” grandparents, 168 aunts and
uncles). The results, presented
in table 2, demonstrate that a
RESULTS pattern of group differences similar to those found in the
parents was detected in both the grandparent and aunt/
Twenty-five mothers and 23 fathers from the 25 mul- uncle groups. As it was felt that childhood communication
tiple-incidence autism families and 30 mothers and 30 behaviors could not be reliably assessed in grandparents,
fathers from the 30 Down syndrome families partici- rates of communication deficits in the grandparents were
pated in this study. A parent in an autism family was not analyzed. Differences in communication between the
included in the analysis only if he or she was the parent autism and Down syndrome aunt/uncle groups were also
of two autistic children. Two mothers had autistic chil- detected. One maternal uncle in the autism group had
dren with two different fathers, and so only 23 autism received a previous clinical diagnosis of autism.
fathers were included in this analysis. There was no sig- The sample of multiple-incidence autism families in-
nificant difference between the autism and comparison cluded only 12 siblings in addition to the probands (one
families in father’s age (t=0.76, df=51, p>0.45), father’s of whom also met the criteria for autistic disorder of the
level of education (χ2=1.39, df=4, p=0.85), mother’s age Autism Diagnostic Interview); the Down syndrome group
(t=0.29, df= 53, p=0.77), or mother’s level of education contained 53 siblings. The small number of siblings in
(χ2=6.95, df=4, p=0.14). The father’s occupational level, the autism group precluded meaningful comparisons;
as specified by the British Manual of the Classification of however, even within this small sample of siblings there
Occupations (26), also did not differ significantly be- was evidence of a higher rate of social deficits—33%
tween the two groups (χ2=6.2, df=4). Although the parent (N=4) versus 0% among the siblings in the Down syn-
groups did not differ on these demographic variables, it drome families. No differences were suggested from
is notable that seven (30%) of the 23 autism fathers ver- comparison of the rates of communication deficits—0%
sus one (3%) of the 30 Down syndrome fathers (χ2=5.50, versus 8% (N=4)—or stereotyped behaviors—0% ver-
df=1, p=0.02) each reported two or more episodes of sus 2% (N=1)—in siblings. Information on first cousins
either resigning or being fired from his job because of was not uniformly available to the parents, and there-
inability to get along with a colleague or supervisor. fore data on first cousins were not analyzed.
Comparison of the autism and Down syndrome par- To take into account possible bias of ascertainment
ents by means of multivariate logistic regression, with so- due to inclusion of families recruited from outside of
cial and communication deficits and stereotyped behav- Iowa (i.e., families not ascertained epidemiologically), we
iors included in the model, revealed significant differences examined rates of social and communication deficits and
between the two groups (–2 log likelihood χ2=19.9, df=3, stereotyped behaviors in the parents from the 18 families
p<0.0002). The results of comparisons of the parental ascertained epidemiologically in Iowa. Differences simi-

188 Am J Psychiatry 154:2, February 1997

 PIVEN, PALMER, JACOBI, ET AL.

TABLE 2. Logistic Regression Comparison of 25 Families With Mul- TABLE 3. Comparison of 25 Families With Multiple-Incidence
tiple-Incidence Autism and 30 Families With One Down Syndrome Autism and 30 Families With One Down Syndrome Proband on Fre-
Proband on Three Types of Behavioral Deficits in Parents, Grandpar- quency of Three Types of Behavioral Deficits in Fathers and Mothers
ents, and Aunts and Uncles
Down –2 Log
–2 Log Likelihood χ2 (df=1) Syn- Likelihood χ2
Autism drome (df=1)
Social Communication Stereotyped Type of Relative
Type of Relative Deficits Deficits Behaviors and Behavioral Deficit N % N % χ2 p
Parents 10.1*** 0.9 5.1** Fathersa
Grandparents 2.8* — 4.1** Social deficits 13 57 4 13 11.46 0.0007
Aunts and uncles 6.6** 6.3** 4.1** Communication deficits 5 22 6 20 0.02 n.s.
Stereotyped behaviors 6 26 1 3 6.22 0.01
*p<0.10. **p<0.05. ***p<0.01. Mothers
Social deficits 9 36 4 13 3.92 0.05
Communication deficits 5 20 0 0 8.49 0.004
lar to those found in the total sample were again detected; Stereotyped behaviors 3 12 0 0 4.91 0.03
i.e., the autism parents showed significantly higher rates aOnly 23 fathers from the autism families participated.
of social deficits and stereotyped behaviors, but not com-
munication deficits, than the parents in the Down syn-
drome families. parents and in the aunts and uncles in the autism families.
To explore possible gender effects in the features of the These findings suggest that the social and communication
broader autism phenotype, we compared the rates of so- deficits and stereotyped behaviors examined in this study
cial and communication deficits and stereotyped behav- may be expressions of the genetic liability for autism.
iors in the autism and Down syndrome fathers and in the To our knowledge, the present study is the first to
autism and Down syndrome mothers. The results are examine these characteristics in relatives ascertained
presented in table 3. Significantly higher rates of social through sibships with multiple-incidence autism. Fami-
deficits and stereotyped behaviors, but not communica- lies ascertained in this manner are likely to be more etio-
tion deficits, were detected in the autism fathers, whereas logically homogeneous than those ascertained through
higher rates of communication deficits, social deficits, probands from single-incidence families and perhaps
and stereotyped behaviors were detected in the autism may also have a higher genetic liability for autism. An
mothers. The significant difference in stereotyped behav- additional strength of this study is the inclusion of an
iors in the mothers was based on a low rate of occurrence epidemiologically ascertained sample in Iowa. A limita-
(12%) and reached significance, in part, because of the tion of this study is the use of the family history method.
absence of any Down syndrome mothers with these be- While this method has been reported to underestimate
haviors. Similarly, high specificity was demonstrated the rate of psychiatric disorders in families (27), the va-
for communication deficits in the autism mothers; com- lidity of this method for examining the characteristics
munication deficits occurred in 20% of the autism moth- measured in this study has been less well established. In
ers and none of the Down syndrome mothers. addition, it should be noted that while the parents of
To avoid performing multiple comparisons and poten- the probands with Down syndrome were selected to
tially inflating our rate of type I errors, we did not under- take into account the possible effects of having a handi-
take statistical comparisons of the rates of individual capped child, the stress of having two autistic children
items from the family history interview in the parents in is likely to be substantially greater than that of raising
the autism and Down syndrome families. However, it is a child with Down syndrome.
notable that the highest rates of any individual items were While our findings are in general agreement with
found for the autism fathers; 52% (N=12) were found to those of Bolton et al. (16), the rate of deficits detected
have few or no adult friendships (versus 3% of the Down in the first-degree relatives of autism and Down syn-
syndrome fathers; N=1), and 48% (N=11) were found to drome probands by Bolton et al. was substantially
be rigid (versus 7% of the Down syndrome fathers; N=2). lower than the rate detected in our study, precluding
The item “socially inappropriate” demonstrated the any comparison of absolute rates across studies. The
highest specificity for autism of any individual item ex- findings of this study, however, differ from the results
amined in the parents, occurring in 26% of the autism obtained in the family history study of Szatmari et al.
fathers (N=6) and 0% of the Down syndrome fathers. (10), in which a high rate of either social or communi-
cation deficits was not found in the families of pro-
bands with pervasive developmental disorder. The criti-
DISCUSSION cal difference between our results and those of Szatmari
et al. (10) may be that we ascertained families that con-
In this study we have replicated and extended the find- tained two autistic siblings, whereas Szatmari et al. as-
ing of others (3, 5, 7, 11, 13, 15, 16) that the relatives of certained families that each contained a single proband
autistic probands show familial aggregation of behaviors with pervasive developmental disorder. Pervasive de-
that are milder than but qualitatively similar to the defin- velopmental disorder, a more broadly defined condi-
ing features of autism. Within this study, the findings for the tion than autism, is more common in the population
parents in the autism families were replicated in the grand- and likely to be a more etiologically heterogeneous dis-

Am J Psychiatry 154:2, February 1997 189

BROADER AUTISM PHENOTYPE

order than autism. Selecting families through multiple- 4. Steffenburg S, Gillberg C, Holmgren L: A twin study of autism
in Denmark, Finland, Iceland, Norway, and Sweden. J Child Psy-
incidence autism sibships is likely to result in a more chol Psychiatry 1989; 30:405–416
homogeneous group of families with a higher genetic 5. Bailey A, Le Couteur A, Gottesman I, Bolton P, Simonoff E, Yuz-
liability for the behaviors we examined. da E, Rutter M: Autism as a strongly genetic disorder: evidence
Our results suggest possible sex-specific differences in from a British twin study. Psychol Med 1995; 25:63–77
6. Piven J, Simon J, Chase GA, Wzorek M, Landa R, Gayle J, Fol-
the expression of the broader autism phenotype. The stein S: The etiology of autism: pre-, peri- and neonatal factors.
absence of a significant difference between the rates of J Am Acad Child Adolesc Psychiatry 1993; 32:1256–1263
communication deficits in the autism and Down syn- 7. August GJ, Stewart MA, Tsai L: The incidence of cognitive dis-
drome fathers can not be taken to mean that these abilities in the siblings of autistic children. Br J Psychiatry 1981;
138:416–422
deficits do not exist, given the limitations of the family 8. Freeman BJ, Ritvo ER, Mason-Brothers A, Pingree C, Yokota A,
history method. The lack of significance in this com- Jensen WR, McMahon WM, Petersen PB, Mo A, Schroth P: Psy-
parison may reflect the small size of the samples being chometric assessment of first-degree relatives of 62 autistic pro-
compared. However, there may be gender differences in bands in Utah. Am J Psychiatry 1989; 146:361–364
9. Szatmari P, Jones MB, Tuff L, Bartolucci G, Fisman S, Mahoney
the degree to which parents exhibit communication W: Lack of cognitive impairment in first-degree relatives of per-
deficits. The high rate of communication deficits in the vasive developmental disorder probands. J Am Acad Child Ado-
autism mothers (20%) was striking, whereas these defi- lesc Psychiatry 1993; 32:1264–1273
cits were completely absent in the Down syndrome 10. Szatmari P, Jones MB, Fisman S, Tuff L, Bartolucci G, Mahoney
WJ, Bryson SE: Parents and collateral relatives of children with
mothers. These results warrant examination in an inde- pervasive developmental disorders: a family history study. Am J
pendent sample. Med Genet 1995; 60:282–289
The findings in this study have implications for future 11. Wolff S, Narayan S, Moyes B: Personality characteristics of par-
genetic studies of autism. Although the genetic basis for ents of autistic children. J Child Psychol Psychiatry 1988; 29:
143–153
autism is well established, the mechanism of genetic 12. Landa R, Folstein SE, Isaacs C: Spontaneous narrative-discourse
transmission has not yet been determined. Future ge- performance of parents of autistic individuals. J Speech Hear Res
netic analyses aiming to delineate genetic mechanisms 1991; 34:1339–1345
in autism (e.g., segregation analyses) need to consider 13. Landa R, Piven J, Wzorek M, Gayle J, Chase G, Folstein S: Social
the inclusion of the broader autism phenotype in their language use in parents of autistic individuals. Psychol Med
1992; 22:245–254
definitions of affected individuals. Similarly, inclusion 14. Gillberg C: Asperger syndrome in 23 Swedish children. Dev Med
of the broader autism phenotype in lod score analysis Child Neurol 1989; 31:520–531
could substantially improve the power to detect genes 15. Piven J, Wzorek M, Landa R, Lainhart J, Bolton P, Chase GA,
for autism, over the power of nonparametric methods, Folstein S: Personality characteristics of parents of autistic indi-
viduals. Psychol Med 1994; 24:783–795
by increasing the number of affected individuals avail- 16. Bolton P, Macdonald H, Pickles A, Rios P, Goode S, Crowson
able for study and by improving the specificity of diag- M, Bailey A, Rutter M: A case-control family study of autism. J
nosing nonaffected subjects. The issue of power is par- Child Psychol Psychiatry 1994; 35:877–900
ticularly relevant in genetic linkage studies of autism, 17. Wechsler D: Wechsler Intelligence Scale for Children—Revised.
given the absence of vertical transmission of the narrow New York, Psychological Corp, 1974
18. Wechsler D: Wechsler Intelligence Scale for Children, 3rd ed. San
phenotype (i.e., autism) and the often small size of fami- Antonio, Tex, Psychological Corp, 1991
lies with autistic children (28). The results of this study 19. Wechsler D: Wechsler Adult Intelligence Scale—Revised. New
and of others suggesting the existence of a genetically York, Harcourt Brace Jovanovich, 1981
related broader autism phenotype indicate the need for 20. Arthur G: The Arthur Adaptation of the Leiter International Per-
formance Scale. Chicago, Psychological Services Press, 1952
further detailed studies, using direct assessment of rela- 21. Stutsman R: Guide for administering the Merrill-Palmer Scale of
tives, to clarify the range of characteristics that define Mental Tests, in Mental Measurement of Preschool Children.
the boundaries of this phenotype. Further, the high rate Edited by Terman LM. New York, Harcourt, Brace and World, 1952
of deficits found in parents in families with multiple-in- 22. Le Couteur A, Rutter M, Lord C, Rios P, Robertson S, Hold-
cidence autism suggests the potential importance of grafer M, McLennan J: Autism Diagnostic Interview: a stand-
ardized investigator-based instrument. J Autism Dev Disord
comparing the rate of the broader phenotype in rela- 1989; 19:363–387
tives ascertained through two or more autistic pro- 23. World Health Organization: ICD-10 Categories F00-F99 Men-
bands with the rate for relatives ascertained through a tal and Behavioral Disorders (Including Disorders of Psychologi-
single autistic proband. cal Development): Clinical Descriptions and Guidelines. Geneva,
WHO, 1992
24. Lord C, Rutter M, Goode S, Heemsbergen J, Mawhood L,
Schopler E: Autism Diagnostic Observation Schedule: a stand-
REFERENCES
ardized observation of communicative and social behavior. J Au-
tism Dev Disord 1989; 19:185–212
1. Piven J, Folstein S: The genetics of autism, in The Neurobiology 25. Piven J, Landa R, Gayle J, Cloud D, Chase G, Folstein SE: Psy-
of Autism. Edited by Bauman ML, Kemper TL. Baltimore, Johns chiatric disorders in the parents of autistic individuals. J Am
Hopkins University Press, 1994 Acad Child Adolesc Psychiatry 1991; 30:471–478
2. Ritvo ER, Jorde LB, Mason-Brothers A, Freeman BJ, Pingree C, 26. British Manual of the Classification of Occupations. London,
Jones MB, McMahon WM, Petersen PB, Jensen WR, Mo A: The Office of Population Censuses and Surveys, 1980
UCLA-University of Utah epidemiologic survey of autism: recur- 27. Andreasen NC, Endicott J, Spitzer RL, Winokur G: The family
rence risk estimates and genetic counseling. Am J Psychiatry history method using diagnostic criteria: reliability and validity.
1989; 146:1032–1036 Arch Gen Psychiatry 1977; 34:1229–1235
3. Folstein SE, Rutter ML: Infantile autism: a genetic study of 21 28. Szatmari P, Jones MB: IQ and the genetics of autism. J Child
twin pairs. J Child Psychol Psychiatry 1977; 18:297–321 Psychol Psychiatry 1991; 32:897–908

190 Am J Psychiatry 154:2, February 1997