Journal of Autism and Developmental Disorders, Vol. 24, No.

5, 1994

Autism Diagnostic Interview-Revised: A

Revised Version of a Diagnostic Interview
for Caregivers of Individuals with Possible
Pervasive Developmental Disorders 1
Catherine Lord 2
The Universityof Chicago

Michael Rutter
Institute of Psychiatry

Ann Le Couteur
Royal Free Hospital

Describes the Autism Diagnostic Interview-Revised (ADI-R), a revision of the

Autism Diagnostic Interview, a semistructured, investigator-based interview for
caregivers of children and adults for whom autism or pervasive developmental
disorders is a possible diagnosis. The revised interview has been reorganized,
shortened, modified to be appropriate for children with mental ages from about
18 months into adulthood and linked to ICD-IO and D S M - I V criteria.
Psychometric data are presented for a sample of preschool children.

1We acknowledge the help of all the parents who participated in these interviews and of
Sharon Storoschuk, Joycr Magill, Cathy Mulloy, and Lyn Anderson-Cook who conducted
interviews. Support was provided by Public Health Service Grant #7 R01 MH46865-03,
Training in RESH Diagnoses of Autism and Spectrum Disorder 7-T35-MH 19726-02 and
the Alberta Heritage Fund for Medical Research, first author and the MRC Child Psychiatry
Unit.
2Address all correspondence to Catherine Lord, Department of Psychiatry, University of
Chicago, 5841 S. Maryland Avenue, MC 3077, Chicago, Illinois 60637. Data for all items,
including those not in the algorithm are available upon request.

659
0162-325719411000--0659507.00/09 1994PlenumPublishingCorporation
660 Lord, Rutter, and Le Couteur

The Autism Diagnostic Interview-Revised (ADI-R) is a newly modified

version of the Autism Diagnostic Interview (ADI; Le Couteur et al., 1989),
a standardized, semistructured, investigator-based interview for caregivers
of autistic individuals, which provides a diagnostic algorithm for the ICD-10
definition of autism (World Health Organization [WHO], 1992) and
DSM-IV (American Psychiatric Association [APA], 1993). In this report,
we describe the revision of the ADI and procedures for training and use
of the new instrument, the ADI-R.
Psychometric properties for the original ADI were provided for a
carefully selected, blindly interviewed and coded, sample of 16 autistic
and 16 mentally handicapped children and adults covering a range of
IQs and chronological ages (Le Couteur et al., 1989). The interview
yielded separate scores in the areas of communication, social skills, and
restricted, repetitive and stereotyped behaviors, as well as early history,
allowing the separate quantification of severity in each of these domains.
Since work on the ADI began in the early 1980s, a number of develop-
ments have suggested revisions that allow items to more accurately reflect
autism-specific deficits and current theoretical perspectives (see Lord &
Rutter, 1994). The original ADI was intended for research purposes and
provided behavioral assessment for subjects with a chronological age of
5 years or above and a mental age of at least 2 years. However, because
most autistic children are now diagnosed during the preschool years
(Short & Schopler, 1988), there was a need for an instrument that could
differentiate autism from other disorders as they present in very young
children. In addition, if the interview were to be used regularly for clinical
purposes, it was important to increase its efficiency and shorten its length,
so that it could be incorporated more easily into existing multidisciplinary
diagnostic assessments.

CHANGES IN ORGANIZATION

The original ADI began with a series of opening questions; followed

by questions concerning infancy and early (under age 5 years) development
in social skills, communication and play; followed by an overlapping, but
not identical, set of items convening social and communication skills at the
time of interview; a section on restricted and repetitive behaviors and in-
terests scored for both current behavior and their occurrence "ever"; and
a final section on general behavior difficulties. Because two goals of the
revision were to make the interview briefer and more appropriate for
younger children, questions concerning early development were consoli=
dated with those concerning careened behaviors. Thus, the ADI-R consists
Autism Diagnostic Interview-Revised 661

of five sections: opening questions; questions on communication (both early

and current); those on social development and play (again both early and
current); enquiries about repetitive and restricted behaviors (all scored for
both current and ever judgments); and a reduced number of questions con-
cerning general behavior problems. It is now possible, after substantial
practice, to give the interview to the parent of a 3- or 4-year-old suspected
of autism in approximately 11/2 hours; for older children, the interview may
take somewhat longer.

Modification of Items

Research in the last 10 years has provided much useful information

clarifying the nature of autism-specific deficits in social and communication
skills (Baron-Cohen, Tager-Flusberg, & Cohen, 1993; Lord & Rutter,
1994). Better discrimination between mentally handicapped autistic chil-
dren and nonautistic severely mentally handicapped children is possible
through more focused descriptions of contexts in which normally develop-
ing and mentally handicapped children behave in consistent ways from early
in development. However, it is important to acknowledge that there is dif-
ficulty differentiating autism when it is accompanied by profound mental
retardation; this difficulty applies to all diagnostic instruments and not just
the ADI-R. Revisions in the area of communication were aimed at iden-
tifying autism-specific aspects of each behavior. For example, an item on
pointing in the original ADI was modified so that the behavior of interest
is now pointing at a distance to express interest, thus excluding pointing
as an instrumental gesture. Several other items were modified to become
more general in focus. A question concerning sensitivity to noise was broad-
ened to include sensitivity to most or all loud noises; in the ADI-R, idi-
osyncratic sensitivity to highly specific, not necessarily loud, noises is now
scored as part of another item concerning abnormal, idiosyncratic negative
reactions to any kind of sensory stimuli. Changes in social items occurred
along the same lines. The quality of social overtures is now scored with
particular reference to the child's attempts to secure help; other items code
the extent to which the child makes social overtures and the range of in-
tentions manifested. The emphasis for social smiling was shifted from the
age of first social smile to whether and if so, when, the child smiles recip-
rocally with others. Range of facial expression is scored only during com-
municative use.
Repetitive use of objects, such as wheel spinning, is now distinguished
from compulsions or rituals that involve an end point and a sequence of
actions toward that end. Difficulty with changes in the subject's own routines,
662 Lord, Rutter, and Le Couteur

such as that of a child who is upset if he does not drink from the same cup
each night, is now scored separately from resistance to minor changes in the
environment, such as seen in a child who becomes distressed if the furniture
in his parents' bedroom is moved. Because of the addition of these new items,
the scoring in this area has also been changed in the algorithm. Some pairs
of items that fall under similar ICD-10/DSM-IV headings (e.g., stereotyped
and repetitive motor mannerisms) are grouped together and the highest score
from either item within the group is used in the algorithm. The purpose of
this approach is to allow finer distinctions among behaviors, without giving
undue credit to possibly overlapping behaviors in the algorithm.

Additional Items in the ADI-R

Several new social and communication items were added to identify

social behaviors expected in normally developing or nonautistic mentally
handicapped children under age 5 that would be abnormal or absent in
young autistic children and items expected to be associated with autism in
young children. These behaviors include use of another's body to commu-
nicate, showing and directing attention, and interest in and response to other
children. Items were also added to provide better coverage of abnormalities
as they appear in older high-functioning children and adults. For example,
there is a question about talk expressing interest in others and a question
about relatively appropriate, but unusually intense, circumscribed interests.
An item concerning behaviors associated with social avoidance related to
anxiety was added in order to tap behaviors that differentiate children and
adolescents with fragile X from those with autism (Wolff, Gardner, Paccia,
& Lappen, 1989). Similarly, there are questions about midline hand-wringing
movements and about hyperventilation in order to cover behavior that is
characteristic of Rett syndrome (Hagberg, Aicardi, Dias, & Rasmos, 1983;
Olson & Rett, 1991; Tsai, 1992). More specific questions about the age when
abnormalities were first manifest were added, as well as questions about
loss of skills and progressive deterioration, again aimed at providing more
accurate information for differential diagnosis between autism and syn-
dromes such as Rett syndrome or disintegrative disorders.

Elimination of Items from the Original ADI

Items judged to be redundant, limited in reliability, or applicable to

only a small proportion of children with pervasive developmental disorders
Autism Diagnostic Interview-Revised 663

(PDD) were eliminated. Such items included questions about sense of hu-
mor, pica, sharing activities, and understanding plots and instructions.

Definition of Items

Definitions of all behavioral items have now been provided in the

ADI-R; these complement and clarify the details given in each coding. It
is the interviewer's responsibility to determine whether, for each item, the
behavior described truly meets the specified criteria.

Administration of Interview

The interview continues to focus on caregiver descriptions of actual

behavior as it has occurred in the subject's daily life. Scoring is made on
the basis of the interviewer's judgment of the code that best fits the be-
haviors described by the caregiver, rather than on judgments made by the
informant. Although the administration of the ADI-R requires a substantial
amount of time, most parents and caregivers find it a relatively comfortable
experience, because they are allowed to describe important aspects of their
child's behavior in their own words. Parents often report that the process
left them with a sense that the interviewer valued their impressions and
opinions and wanted to know more about their child than could be deter-
mined by observation in a clinic. The experience of the interview, as part
of a multidisciplinary intake assessment, also helps some parents have a
better understanding of the factors, particularly social behaviors, that are
being evaluated in order to reach a diagnosis. Questions in the interview
are ordered by content area, but within these, they are deliberately written
and sequenced to provide caregivers with opportunities to describe positive
aspects of their child's behavior and development and try to minimize the
effect of having repeatedly to answer questions in a negative way.

Scoring and Interpretation

Coding methods remain the same as the ADI, with most items
coded no definite behavior of the type specified (0), behavior of the type
specified probably present but defining criteria not fidly met (1), and defi-
nite abnormal behavior of the type described in the definition and coding
(2), with a code of 3 used occasionally to indicate extreme severity.
Each item is scored for current behavior, with the exception of a few
664 Lord, Rutter, and Le Couteur

items where the behavior is relevant only during particular age periods.
For these items, specific age restrictions are given. For example, imagi-
native play, imaginative play with peers, and group play are coded only
between ages 4 and 10 years, and reciprocal friendships only above 10
years of age. Circumscribed interests is only scored for children 10 years
and older. These items were omitted from the present analyses because
of the group age of the sample. Each item describing the presence of
a qualitatively abnormal behavior (i.e., a positive item) is also scored
as to whether the behavior "ever" occurred for a period lasting as long
as 3 months. For these items, the "ever" coding is a lifetime measure
that includes the current period. Such behaviors can be coded for any -
time after the subject has achieved a mental age greater than 18
months.
Each item inquiring about the lack of a behavior or skill associated
with normal development (i.e., a negative item), is coded for its most ab-
normal manifestation between the ages of 4 and 5 years (i.e., 48-60
months), in addition to the current situation at the time of interview. The
rationale of the focus on the 4- to 5-year age period concerns the advan-
tages of a standard age period for comparative purposes, together with the
desirability of selecting an age that is high enough to provide an adequate
range of behavior and low enough to precede possible major changes in
behavior.

Algorithm

An algorithm for diagnosis comparable to that used in the original

A D I (Le Couteur et al., 1989) was generated by selecting A D I - R items
that most closely depicted the specific abnormalities described in the
clinical descriptions and diagnostic guidelines from the DSM-IV and
ICD~ (Volkmar et al., 1993; WHO, 1992). As shown in Table III,
using this procedure for lifetime diagnosis for subjects over the age of
4, 16 items were identified to measure reciprocal social interaction, 13
to measure communication, and 8 to measure restricted, repetitive be-
haviors. Within these areas, scores are also computed for each subdo-
main of abnormality as listed in the diagnostic criteria for research
(WHO, 1992), as shown in the tables. The scores are transferred di-
rectly from A D I - R items but, in order to ensure that undue weight is
not placed on individual items, severity codes of 2 and 3 are both
treated as if they were 2. Algorithm cutoffs were determined by gen-
erating R O C curves with the present data and identifying the point
within each area that yielded the best combination of sensitivity and
Autism Diagnostic Interview-Revised 665

specificity with both exceeding .90. The intention is to use one algorithm
for children from mental ages of 18 months through adulthood, with
three versions containing minor modifications: (a) a lifetime version;
(b) a version based only on current behavior; and (c) a version for use
with children under the age of 4 years. The last is necessary because
obviously all items based on behavior above this age are inapplicable.
Data collected on the earlier version of the ADI as part of the DSM-IV
field trials indicated relatively little differences between the current and
lifetime versions, with the greatest discrepancies emerging for high-
functioning adults (Lord et al., in press). Use of the ADI-R with very
young preschool children is the topic of a separate paper (Lord, Storo-
schuk, Rutter, & Pickles, 1993). Psychometric data presented here are
for a sample of preschool children with mental ages greater than 18
months, because of earlier findings that the standard algorithm did not
discriminate autistic children with mental ages below 18 months from
very young children with severe mental handicaps (Lord et al., 1993).
This sample was selected because children are now most commonly di-
agnosed during preschool years and ensuring adequate reliability and
validity for them seemed an appropriate place to start.
The algorithm, based on ICD-10/DSM-IV guidelines, specifies a
cutoff score of 8 on communication items for verbal subjects (who, as in
the original ADI, were operationally defined as individuals scoring 0 on
the "level of language" item, indicating use of three-word phrases, spon-
taneous or echoed, that sometimes contain a verb) and a cutoff of 7 for
nonverbal subjects. For all subjects, a minimum score of 10 on social
items and 3 for restricted and repetitive behaviors was identified. Because
the ADI-R provides much more detailed information on a wide range of
specific behaviors than that ordinarily available, we have followed a total
score approach. This follows the same principles as the official ICD-10
research criteria (WHO, 1993) in requiring abnormalities in all domains,
but avoids the dilemma of deciding how many individual items are re-
quired for each "symptom" to be present. This is advantageous because
the separation of individual symptoms within each domain is necessarily
somewhat arbitrary in view of the fact that all are thought to reflect the
same basic deficits. The total score approach also reduces error stemming
from the inevitable unreliability of single symptom judgments. To meet
ICD-10/DSM-IV draft diagnostic criteria for autism, an individual must
meet criteria in each of the three content areas in the guidelines for re-
search diagnoses, as well as exhibiting some abnormality in at least one
area by 36 months of age, as described by the caregiver or judged by the
interviewer.
666 Lord, Rutter, and Le C o u t e u r

PSYCHOMETRIC DATA: STUDY h RELIABILITY

Method

Subjects
Ten autistic (8 male, 2 female) and 10 mentally handicapped or lan-
guage-impaired (8 male, 2 female) children served as subjects for initial
analyses concerning the reliability of the ADI-R for preschool children.
Autistic children were selected from the Communication and Behavior
Disorder and Developmental Pediatrics or Clinics at Glenrose Hospital
in Edmonton, Alberta, Canada. All autistic children had received inde-
pendent clinical diagnoses based on DSM-III (APA, 1980) and ICD-10
draft criteria (WHO, 1987) by a clinical psychologist and child psychiatrist
within 6 months of participating in the study. Mentally handicapped and
language-impaired children were recruited from the same clinical and local
noncategorical preschool programs for mentally handicapped children.
None had ever received diagnoses of autism or PDD according to teacher
or parent reports or hospital records. Although all of the nonautistic chil-
dren attended a preschool program for mentally handicapped children,
two tested as having nonverbal IQs close to or greater than 100, but had
significant expressive and receptive language delays. Children from both
groups were all living with at least one biological parent; parents were all
native English speakers with at least a Grade 11 education. All children
in both diagnostic groups had shown significant language delays prior to
36 months of age.
Children who were nonambulatory or had other marked motor im-
pairments, had other than mild, remediable sensory impairments or iden-
tifiable syndromes (e.g., Down syndrome, Rett syndrome) or were judged
by their teachers to be functioning below the 12-month level overall were
excluded from the study because of expected differences in parental per-
ceptions and difficulty in separating deficits associated with autism from
sensory or motor impairments or profound mental retardation (Lord et al.,
in press). Ethnic representation was 15% Asian, 10% West Indian, and
75% white.
All children were assessed using the Men'ill-Palmer Scale of Mental
Tests (Stutsman, 1931), scoring only nonverbal items. Because the autistic chil-
dren typically had difficulty with several of the imitation items in the first level
of the test, a basal score on the Merrill-Palmer was operationally defined as
passing five performance items at the 18- to 23-month level or five perform-
ance items at the 24- to 29-month level. Two autistic children and one mentally
handicapped child did not reach either criterion and so were administered the
Autism Diagnostic Interview-Revised 667

Bayley Scales of Mental Development (Bayley, 1969). To be consistent across

tests, all scores reported here are ratio IQs.
The two groups, autistic and nonautistic mentally handicapped/lan-
guage impaired (MH/LI), did not differ in chronological age, with age
ranges in b o t h cases from 36-59 months and mean ages of 48.9
(SD = 12.2) and 50.1 (SD = 15.7) months, respectively. Mean nonverbal
IQ/DQ was 64.12 (SD = 32.86) for the autistic children compared to 63.80
(SD = 23.61) months for the nonautistic children. Mental ages ranged
from 21-74 months.

Procedures

Videotaped interviews with mothers of children described above were

conducted by one of three experienced interviewers, unfamiliar with the
children and blind to their diagnoses. The interview consisted of the ADI-R
with the following modifications: All items were administered for current
behavior, with "current" defined as a behavior that had occurred on a regu-
lar basis over the course of at least 1 month in the last 3 months. In ad-
dition, a number of items were asked retrospectively for specific ages,
usually 2 years (i.e., 24 months) or infancy (defined as "ever" occurring
between 6-18 months chronological age). Interviews lasted between 60 and
90 minutes. Each videotaped interview was independently scored by four
medical or graduate students also blind to diagnosis. Prior to the scoring,
coders had spent 6 weeks watching and scoring videotapes of the ADI-R,
giving practice ADI-Rs with supervision, observing assessments, treatment
sessions and groups for autistic children and adults, and visiting schools
and day programs for individuals with autism and other developmental and
psychiatric disorders. Before reliability coding began, each coder had
achieved reliabilities of greater than 90% on three consecutive scorings of
live or videotaped ADI-Rs with consensus codes determined by the first
author and at least one other coder.

Data Analysis

Reliability was calculated using percentage exact agreement and

weighted kappa (Cohen, 1968) for each of the six rater pairs. Multirater
weighted kappas (Conger, 1980), reported in the following tables were com-
puted across raters, as was mean percentage agreement. Items appropriate
only for children over age 4 years were not analyzed. Data for items not
 Table I. Interrater Reliability for ADI-R Algorithm Items from ICD-10/DSM-IVa
E
Ir
(n = 20) Agreement

Qualitative Abnormalities in Reciprocal Social Interaction

BI. Failure to use eye-to-eye gaze, facial expression, body posture and gesture to regulate social interaction (.93)
Direct gaze .70 .91
Social smiling .64 .90
Range of facial expression used to communicate .86 .96
B2. Failure to develop peer relationships (.93)
Imaginative play with peersb .89 .95
Interest in children .69 .90
Response to children .72 .90
Group play with peersb .88 .96
B3. Lack of social-emotional reciprocity and modulation to context (.96)
Use of other's body .84 .93
Inappropriate facial expressions .67 .92
Quality of social overtures .71 .91
Appropriateness of social responses .71 .95
Offers comfort .87 .95
B4. Seeking to share own enjoyment (.95) t"
Showing and directing attention .70 .91
Offering to share .85 .94
Seeking to share in own enjoyment .80 .93
-I
Qualitative impairments Communication and Language
ga.
CI. Delay or total lack of spoken language, not compensated by gesture (.95)
tr "
Pointing to express interest .80 .92
Conventional gestures .74 .88 o
Nodding head .86 .94
Head shaking .89 .96
C2V. Relative failure to initiate or sustain conversational interchange (.94)b
Social chat .77 .93
Reciprocal conversation .84 .93
C3V. Stereotyped and repetitive use of language (.94) b
Stereotyped utterances .85 .95 o
Inappropriate questions .70 .91 n
Pronominal reversal .76 .92 h"
Neologisms/idiosyncratic language .73 .95
C4. Lack of varied spontaneous make-believe or social imitative play (.94)
Spontaneous imitation .69 .90 Jr
Imaginative play .85 .93 ~.
Imitative social play .69 .90 ~.

Restricted, Repetitive Behaviors and Interests

DI. Encompassing preoccupations (.93)
Unusual preoccupations .64 .90
D2. Apparently compulsive adherence to nonfunctional rituals (.94)
Verbal rituals .83 .93
Compulsions/rituals .63 .90
D3. Stereotyped and repetitive motor mannerisms (,94)
Hand and finger mannerisms .69 .90
Other complex mannerisms .86 .94
D4. Preoccupation with part-objects or nonfunctional elements of materials (.95)
Repetitive use of objects ,67 .91
Unusual sensory interests .82 .95
a lntraclass correlations for each area are presented immediately after the area title. Intraclass correlation for all nonverbal communication items
(.94); Intraclass correlation for all verbal communication items (.94); lntraclass correlation for social total (.97).
bn = 11.
670 Lord, Rutter, and Le Couteur

included in the final version of the ADI-R algorithm are available from
the authors upon request.
Results

Table I provides reliability data for algorithm items concerned with

reciprocal social interaction; the remaining social items are reported in
the Appendix. Multirater weighted kappa levels exceeded .70 for 12 out
of 15 social algorithm items. Weighted kappas for all individual rater pairs
exceeded .63. Lower kappas generally reflected limited variability, either
with relatively few scores of 2 (e.g., inappropriate facial expression, for
which few children scored as markedly abnormal) or with few scores of
0, indicating no deficit (e.g., social smiling, for which even the language
impaired/mentally handicapped children rarely received scores of 0). Non-
algorithm items also showed consistently high interrater reliability with the
exception of arms up to be lifted, discrimination of parents, and attention
to voice for which weighed kappas were low (.52, .54 and .59, respectively)
because of low variability, though percentage agreement was high (90, 96,
and 87, respectively).
Items concerning verbal skills, particularly those describing language
abnormalities, were all nearly identical in content to those in the original
ADI and were scored for only the 11 verbal subjects. However, there were
many new or modified items describing nonverbal aspects of communica-
tion. These were scored for all 20 reliability subjects. Multirater weighted
kappas equaled or exceeded .69 for all 13 communication items. Nonalgo-
rithm communication items were also generally reliable.

Restricted and Repetitive Behaviors

Weighted kappa levels equaled or exceeded .63 for all 7 items in the
area of restricted and repetitive behavior (circumscribed interests was not
scored); mean percentage agreement was above 90 for all items. Weighted
kappas for individual rater pairs were consistently above .55. Kappas were
lower than other domains because of more frequent scores of 0 that had
occurred for social or communication items.

Reliability for Rater Pairs

For the six rater pairs, mean weighted kappas across all items
ranged from .73 to .78, with mean weighted kappas for algorithm items
exceeding .75 for each pair. Mean percentage agreement across all
Autism Diagnostic Interview-Revised 671

items ranged from 90 to 93 for each pair, with agreement for algorithm
items exceeding 92 for all pairs. As shown in Tables I and III, for al-
gorithm items, intraclass correlations for domain scores ranged from
.93 to .97.

lnternal Consistency

Cronbach's alpha was run for each domain separately to assess the
internal consistency of each area. For the 15 items in the social area, item-
total correlations ranged from .54 (for direct gaze) to .77 (quality of social
overtures), with an alpha of .95. For restricted and repetitive behaviors,
item-total correlations ranged from .30 (compulsions and rituals) to .53 (un-
usual sensory behaviors), with an alpha of .69. For the 11 verbal subjects
(those who had three-word phrases), item-total correlations ranged from
-.06 (inappropriate questions) to .77 (instrumental gestures) with an alpha of
.85. For all subjects, item-total correlations for communication items
ranged from .45 (imitative social play) to .70 (conventional, instrumental ges-
tures) with an alpha of .84.

Reliability Over Time

Six mothers (4 of autistic children, 2 of mentally handicapped

children) were reinterviewed 2-3 months later by a different inter-
viewer unfamiliar with the child and the previous interview. Although
the small samples limit the interpretability of statistics, exact agree-
ment exceeded 83% (Kw > .55) for all but 6 items, with a mean of
91% (M Kw = .72). On these 6 items, exact agreement occurred in
4 out of 6 rater pairs.

STUDY 2: VALIDITY

Method

Subjects

After the reliability study was completed, an additional 30 subjects

were recruited for the validity analyses (15 autistic, 15 nonautistic), result-
ing in a total of 25 subjects in each group. The same criteria for inclusion
in the study for autism and for mental handicap/language impairment were
672 Lord, Rutter, and Le Couteur

used as described earlier. Additional subjects were recruited from the

sources in Edmonton, Alberta, described previously, from the Greensboro-
High Point TEACCH clinic in North Carolina and from local noncategori-
cal programs for language-impaired and mentally handicapped children in
Greensboro and the surrounding area.
Children were recruited into two groups of equivalent chronologi-
cal age (CA) in months: autistic (M CA = 46.76, SD = 10.73) and
nonautistic (M CA = 44.72, SD = 13.74), shown in Table V. Ethnic
distribution was 12% African American and West Indian, 82% white,
6% Asian, Hispanic, and Native American. It was equivalent across lan-
guage and diagnostic groupings. Social class, measured by occupational
status of father (Treiman, 1977) covered a large range, with mean val-
ues for each diagnosis ranging from 43.69-52.14; all in the middle class.
Additional subjects were assessed using the Merrill-Palmer and the
Bayley, as described for the reliability study. Mean IQ/DQS were 71.88
(SD = 21.33) for the autistic children and 71.48 (SD = 20.09) for the
nonautistic children, with mental ages of 34 and 32 months, respec-
tively.

Procedures

Interview procedures were similar to those in the reliability study

except that coding for the additional subjects was made on the basis
of live interviews of mothers or both parents together. Three additional
interviewers were added to the research team over the course of data
collection. Prior to data collection, each accompanied the experienced
research associates on interviews, scoring along with them, and then
giving interviews that were scored by their companion as well until over
90% agreement was reached for three consecutive interviews carried
out by the new interviewer. During data collection for the validity study,
training tapes for the ADI-R became available. All interviewers coded
these tapes, blind to diagnosis, with percentage agreement for exact score
completed item by item consistently above 90%, and averaging about
94%. Over the course of the study, reliability checks were made by hav-
ing two research associates conduct approximately one in five inter-
views together.
Parents of all but 5 of the additional autistic children and 7 of the
additional mentally handicapped/language-impaired children were inter-
viewed as part of an outpatient diagnostic assessment. Interviewers were
unfamiliar with the child but did have some background information in-
eluding reasons for referral and previous diagnoses. Since typically about
Autism Diagnostic Interview-Revised 673

40% of children referred to the clinics known for seeing autistic children
were diagnosed as not autistic, and about 10% of the children referred to
the clinic for developmental disorders were diagnosed as autistic, this was
felt to be similar to interviewers remaining blind. Only children who later
received DSM-III diagnoses of either autism, language impairment (LI),
or mental handicap (MH), without autism/PDD made by a psychiatrist
and/or psychologist who was unfamiliar with the ADI-R score were in-
cluded in the study. For the validity study, consensus scores for the reli-
ability subjects were created through discussion among the interviewers and
raters.

Algorithm

The appropriateness of the algorithm generated for the ADI-R was

tested with the preschool sample by comparing individual items and area
summary scores across diagnostic groups and by comparing the number of
children who met formal criteria for autism. One-way fixed-effect ANOVAs
(2 diagnoses) were performed on all item scores and algorithm area scores.
Because variances were unequal, scores were converted to ranks before
analyses were performed. Raw scores are reported for all measures for the
sake of interpretability. Because of the large number of analyses, only those
with values of p < .02 are treated as significant. When computing the di-
agnostic algorithm, items coded 3 for particularly severe manifestations or
7 for indications of abnormality that differed from the dimension in ques-
tion were recoded as 2 and 0, respectively. Only items appropriate for chil-
dren 4 years and under were analyzed in the validity studies because of
the young age of the sample.

Social

As shown in Table II, all 13 algorithm items from the social area
that could be scored in preschool children showed significant diagnostic
differences across autistic and mentally handicapped groups. Several non-
algorithm items that described behaviors typically attributed to autistic
children were not significant, such as cuddliness and discrimination of par-
ents. Though the diagnostic group difference for current separation anxiety
was significant; nearly one third of the autistic children were scored as
not abnormal (8/25 did show separation anxiety) or mildly abnormal (11
out of 25).
 "4

Table 1I. Mean Scores for ADI-R Algorithm Items for Reciprocal Social Interaction
Mentally handicapped/
Autistic Language-impaired
(n = 25) (n = 25)
M SD M SD ANOVAs: F(I, 48)
BI. Failure to use eye-to-eye gaze, facial expression, body posture and gesture to regulate social interaction
Direct gaze 0.92 0.76 0.36 0.64 7.97a
Social smiling 1.28 0.89 0.20 0.50 27.95 b
Range of facial expression used to communicate 1.32 0.69 0.20 0.41 48.75 b

B2. Failure to develop peer relationships

Interest in children 1.72 0.54 0.44 0.58 64.67b
Response to children 1.40 0.64 0.36 0.57 36.54 b

B3. Lack of seeking to share own enjoyment

Showing and directing attention 1.84 0.47 0.40 0.71 71.67 b
Offering to share 1.68 0.56 0.76 0.88 19.53b
Seeking to share own enjoyment 1.48 0.65 0.32 0.63 41.02b
r"
o

B4. Lack of social emotional reciprocity and modulation to context

Use of other's body 1.44 0.82 0.20 0.41 45.76b
Inappropriate facial expressions 1.16 0.80 0.12 0.33 36.05 b
Quality of social overtures 1.64 0.49 0.40 0.58 67.05 b
Appropriateness of social responses 1.64 0.64 0.40 0.58 51.95 b e~

Offers comfort 1.48 0.71 0.04 0.20 94.25 b

ap < .01. O
bp < .001.
Autism Diagnostic Interview-Revised 675

As shown in Table III, all social subdomain totals, based on ICD-

10/DSM-IV draft guidelines (e.g., integration of nonverbal behaviors, peer
relationships), also yielded significant differences between the autistic and
MH/LI groups. In a one-way fixed-effect ANOVA performed on the total
social algorithm scores converted to ranks, autistic subjects had significantly
higher scores (indicating greater abnormality) than the MH/LI children,
F(1, 48) = 243.38, p < .0001.
For the social algorithm total, 2 autistic children had a score of 11
(10 was the cutoff); the remaining 24 autistic children had scores of 13 or
greater. In contrast, no children from the MH/LI group had social domain
scores exceeding 10 out of 29. Two MH/LI children had scores of 10.

Communication

Two subdomains of communication were coded for children at all

language levels. These were the areas of gesture (C1) and play (C4). These
subdomains are discussed first and then the remaining subdomains consid-
ered separately for children with phrases (verbal) and children with single
words only or no words (nonverbal).
As shown in Table IV, there were significant main effects for diagnosis
for all gesture (C1) items. Autistic youngsters were described as using fewer of
all types of gestures than the MH/LI children. The three items in the subdomain
of play (C4), spontaneous imitation, imaginative play, and imitative social play, all
yielded significant main effects of diagnosis. Children with autism showed higher
(more abnormal) scores in all aspects of play and imitation.
Items from the two subdomains of verbal communication were compared
across diagnosis for children with phrases only. These domains were conver-
sation (CV2), and stereotyped use of language (CV3). Out of six items, there
were significant main effects of diagnosis for only one: social chat. For the
five other items, scores for the autistic children indicated greater abnormalities;
however, because of small samples and floor effects on some items in this
relatively young sample, differences were not significant.
Several nonalgorithm communication items also revealed diagnostic group
differences. The autistic children acquired their first words at a later age and were
descn~oed by their parents as less likely to babble in a socially directed way as
infants than the MH/LI children. They also differed in their use of enactive and
emotional gestures, comprehension of language, understanding gestures, and elic-
ited vocal imitation. Significant diagnostic differences for communication items
administered only for the verbal children occurred in immediate echolalia and
understanding the plots of simple stories.
 Table IlL Means and Standard Deviations for ADI-R Subdomains and Domain Scores
Mentally handicapped/
Autistic Language-impaired
(n = 25) (n = 25)

M SD M SD ANOV~

Quantitative Impairments in Reciprocal Social Interaction

BI. Failure to use eye-to-eye gaze, facial expression, 3.52 1.61 0.76 1.01 FO, 48) = 52.66 ~
body posture, and gesture to regulate social interaction
B2. Failure to develop peer relationships 3.12 0.93 0.80 1.04 /7(1, 48) = 69.24c
B3. Lack of seeking to share own enjoyment 5.00 1.26 1.48 1.58 F(I, 48) = 75.67':
B4. Lack of social emotional reciprocity and modulation 7.36 1.93 1.16 1.21 F(1, 48) = 184.34c
to context
B. Social domain total 19.00 3.76 4.20 2.88 F(I, 48) = 243.38c

Quantitative Impairments in Communication and Language

CI. Delay or total lack of spoken language not compensated 6.60 1.47 2.08 2.50 F(I, 48) = 60.73 c
by gesture
C2V. Relative failure to initiate or sustain conversational 3.22 1.09 1.84 1.52 F(I, 21) = 19.40c
interchange"
t"

C3V. Stereotyped and repetitive use of language" 3.11 1.90 1.36 0.63 F(1, 21) = 10.38a
II

C4. Lack of varied spontaneous make-believe or social 3.44 1.51 1.84 1.52 F(1, 48) = 38.44 c
gl
imitative play e,

g-
CNV. Communication total (nonverbal children)a 11.62 1.96 5.09 4.28 F(1, 25) = 28.91 c
o
CV. Communication total (verbal children)a 16.33 2.96 5.57 3.06 F(1, 21) = 69.60':
 Ig

Table IV. Mean Scores for ADI-R Algorithm Items for Communication a
El
Mentally handicapped/
Autistic Language-impaired
~t
(n = 25) (n = 25)
M SD M SD ANOVAs
,g
CI, Delay or total lack of spoken language, not compensated by gesture
Pointing to express interest 1.68 0.47 0.48 0.87 F(I, 48) = 36.49': 3.
Conventional gestures 1.72 0.46 0.60 0.87 /7(!, 48) = 32.67':
Nodding head 1.80 0.50 0.68 0.85 F(I, 48) = 32.11 c
Head shaking 1.40 0.76 0.32 0.56 F(I, 48) = 32.64c ~..

C2. Relative failure tO initiate or sustain conversational interchange a

Social chat 1.55 0.53 0,28 0.47 F(1, 21) = 36.52 c
Reciprocal conversation 1.67 0.71 0.93 0.83 n$

C3V. Stereotyped and repetitive use of language

Stereotyped utterances 1.55 0.73 0.71 0.73 ns
Inappropriate questions 0.44 0.73 0.14 0.36 ns
Pronominal reversal 0.87 0.76 0.07 0.25 ns
Neologisms/idiosyncratic language 0.22 (I.44 0.07 0.27 ns

C4. Lack of varied spontaneous make-believe or social imitative play

Spontaneous imitation 1.72 0.54 0.68 0.85 F(I, 48) = 26.51 *
Imaginative play 1,48 0.77 0.68 0.80 F(i, 48) = 12.94 b
Imitative social play 1.24 0.66 0.48 0.51 /7(I, 48) = 20,63 c

an = 16 For nonverbal autistic and n = 11 for nonverbal MH/LI; n = 9 for verbal autistic and n = 14 for verbal MH/L1.
bp < .01,
Cp < .001.
678 Lord, Rutter, and Le Couteur

As shown in Table III, algorithm summary scores for communication

for verbal children yielded a significant main effect for diagnosis, F(1,
21) = 69.6, p < .0001. All 9 verbal autistic youngsters fell in the range of
autism (10 or greater); 5/14 verbal MH/LI children also scored in this range.
There was also a significant effect of diagnosis for the nonverbal children,
F(1, 25) = 28.91, p < .001. All nonverbal autistic children met the com-
munication criterion, as did 4 MH/LI children.

Restricted and Repetitive Behaviors

As shown in Table V, four items from the ICD-10/DSM-IV domain

of restricted and repetitive interests and behaviors showed significant
differences in distribution for autistic and nonautistic children. These included
verbal rituals, unusual sensory behaviors, hand and finger mannerisms, and
whole body mannerisms. Repetitive use of objects did not show a significant
difference; comp,dsions/rituals and unusual preoccupations were marginally
significant. However, because of findings from previous research with older
populations of the diagnostic usefulness of these items (Le Couteur et al.,
1989; Lord et ai., in press), they were retained in the algorithm. They re-
quire continued evaluation, however. Nonalgorithm items such as abnormal
idiosyncratic negative response, unusual attachments to objects, resistance to
change in own routine, and response to trivial changes in the environment
did not differ across diagnosis.
The mean score for the algorithm area of restricted, repetitive be-
haviors was 4.92 (SD = 1.80) for the autistic children, compared to 1.96
(SD = 1.64) for the mentally handicapped children/language impaired chil-
dren, F(1, 48) = 37.56, p < .0001 (performed on ranks). Of the 25 autistic
children, 1 had a score of 2 (not meeting criteria), 7 children scored a total
of 3 in this area; the remaining 17 autistic children scored 4 or more. Of
the 25 MH/LI children, 6 received scores of 3 or higher.

Items Outside Diagnostic Areas

Few differences were significant for nonalgorithm items outside the

three diagnostic areas. Only curiosity, initiation of appropriate activities, food
fads, sensitivity to noise, self-injurious behavior, and unusual fears were reli-
ably different across diagnostic categories. Severity of aggressive behavior,
pica, tantrums, motor difficulties, and sleep problems did not distinguish
the diagnostic groups.
 gg

r."
Table V. Mean Scores for ADI-R Algorithm Items for the Area of Restricted, Repetitive Behavior and Interests gl

Mentally handicapped/
Autistic Language-impaired
gl
(n = 25) (n = 25)

M SD M SD ANOVAs: /7(i, 48)

DI. Encompassing preoccupations

Unusual preoccupations 1.04 0.93 0.48 0.77 5.35 a
m.

D2. Apparently compulsive adherence to nonfunctional rituals

Verbal rituals I. 16 1.21 0.40 0.87 14.26b
Compulsions/rituals 0.68 0.80 0.20 0.50 6.45 a

D3. Stereotyped and repetitive motor mannerisms

Hand and linger mannerisms 1.20 0.76 0.20 0.58 27.27c
Other complex mannerisms 0.88 0.93 0.24 0.60 8.420

D4. Preoccupation with part-objects or nonfunctional elements of materials

Repetitive use of objects 1.00 0.76 0.72 0.68 ns
Unusual sensory interests 1.12 0.60 0.36 0.57 21.130

"p < .02.

bp < .01.
Cp < .001.
 Qo

Restricted, Repetitive Behaviors and Interests

DI. Encompassing preoccupations 1.04 0.93 0.48 0.77 ns

D2. Apparently compulsive adherence to nonfunctional rituals 1.16 1.21 0.40 0.87 ns

D3. Stereotyped and repetitive motor mannerisms 1,36 0.76 0.32 0.69 F(I, 48) = 25.75r

D4. Preoccupation with part-objects and nonfunctional 1.36 0.57 0.76 0.66 F(1, 48) = 11.7r
elements of materials
D. Restricted repetitive behaviors and interests total 4.92 1.80 1.96 1.64 F(I. 48) = 37.56c

an = 16 For nonverbal autistic and n = 11 for nonverbal MH/LI; n = 9 for verbal autistic and n = 14 for verbal MH/LI.
bp < .01.
Cp < .001.
o

g~
g~
es_

o
Autism Diagnostic Interview-Revised 681

Overall Algorithm

All but one of 25 clinically diagnosed autistic children met A D I - R

criteria for a I C D - 1 0 / D S M - I V draft diagnosis of autism. The one
exception was a high-functioning 4-year-old boy with autism who re-
ceived a score of 2 for restricted, repetitive behaviors (one less that the
cutoff). Two for the 25 mentally handicapped/language impaired chil-
dren were classified as autistic by the ADI-R, but did not receive clini-
cal diagnoses of autism. Both met criteria exactly or by 1 point in two
out of three areas; both boys were nonverbal, young 3-year-olds with
mental ages of 21 and 26 months, respectively. They had received clinical
diagnoses of general developmental delay, severe receptive-expressive lan-
guage impairment, and in one case, attention deficit/hyperactivity disor-
der and in the other, oppositional disorder.

DISCUSSION

The ADI-R is a reliable and valid instrument for making diagnoses

of autistic children of preschool age. Interrater reliability is good, with
kappas ranging from .62 to .89 and equivalent to those found for the origi-
nal ADI for communication and social items. Interrater reliability for
items in the area of restricted and repetitive behaviors and interests is
adequate, with a mean kappa of .70, but lower than that found with the
earlier version of the ADI, in part because of slightly lower kappas for
compulsions and rituals, unusual preoccupations, hand and finger manner-
isms, and both resistance to change items. Percentage agreement for each
of these items was over 90%, but frequencies of occurrence were suffi-
ciently low to restrict the range of kappas. Since the present data were
collected by interviewers less experienced with autism (though given sub-
stantial experience with the ADI-R) from more naive parents of children
with a mean age of 31/2-4 years, these findings are encouraging, particu-
larly for clinical use of the instrument. Further samples of older children
and adults require study as well.
As for the earlier version of the ADI, intraclass correlations are very
high. Internal consistency is also quite good, particularly for communication
and social items. Reliability across time is adequate but needs to be studied
further with a larger sample.
Individual items were much better at discriminating autistic from
mentally handicapped/language-impaired preschool children than had been
expected. Given earlier findings that "ever" scores and scores that targeted
behaviors during the 4- to 5-year period were most discriminative, the finding
682 Lord, Rutter, and Le Couteur

that all social and nonverbal communication algorithm items yielded sig-
nificant differences for diagnosis was surprising. The algorithm items that
did not differentiate young children included some verbal communication
items, for which results were limited by the small samples, and several items
from the area of restricted and repetitive behaviors that may have been
affected by the young age of the subjects. When clear differences had been
found for the item in the DSM-IV field trials (Lord et al., in press) or for
the earlier version of the item with the older sample and the item had
been changed very little from the ADI, the decision was made to keep it
in the algorithm.

Conchtsions

This paper describes a revision of the Autism Diagnostic Inter-

view, the Autism Diagnostic Interview-Revised (ADI-R), a semistruc-
tured, investigator-based interview for caregivers of children and adults
for whom autism or pervasive developmental disorders is a possible di-
agnosis. The revised interview has been reorganized, shortened, ad
modified to be appropriate for children with mental ages from about
18 months into adulthood. It is now linked more closely to ICD-10 and
DSM-IV criteria. Its psychometric properties are strong for preschool
children but require additional investigation, particularly with older
children and adults. Further comparisons between well-defined autistic
samples and other groups with pervasive developmental disorders and
related difficulties will also be important in indicating the usefulness
of the A D I - R in diagnosing autism versus other pervasive developmen-
tal disorders:

3Use of the ADI-R for research purposes requires training in both administration and scoring
by a person experienced in use of the instrument who has established reliability with other
experienced individuals. Training workshops are now available annually in the U.K. and in
North America. Training videotapes that describe the organization and purpose of the ADI-R
and provide detailed examples o f administration and scoring of social, communication, and
restricted and repetitive interest items are now available as part of the training package.
These tapes allow trainees to compare their codings with those of the consensus codes of
the authors. Each tape is provided with a commentary and justifications for decisions. Before
using the ADI-R in a research project, investigators are requested to attend a training
workshop, to show reliability with consensus codings for at least two standard interviews and
to demonstrate the ability of another trained researcher to reliably score two examples of
interviews administered by the investigator.
Autism Diagnostic Interview-Revised 683

APPENDIX
Nonalgorithm items for ADI-R: Reliability and Validity
with a Preschool Sample a
~r Aggrement F(1, 48)

Social and Play Items

Arms up to be lifted (infant) .52 .90 11.82c
Arms up to be lifted (current) .80 .90 7.400
Infant direct gaze .89 .95 ns
Infant separation anxiety .78 .90 9.73 b
Separation anxiety (current) .62 .90 8.7r
Cuddly as infant .76 .91 ns
Cuddly as child (current) .75 .90 ns
Infant social smile .72 .89 ns
Social smile at 2 years .75 .89 18.42c
Secure base .81 .97 5.6r
Attention to voice (current) .59 .87 14.66c
Affection .85 .94 9.09b
Come for comfort at 2 years .80 .93 15.77c
Come for comfort (current) .68 .91 11.68c
Discrimination of parents .54 .96 ns
Sense of humor .78 .90 11.69c
Join others' activities .72 .90 15.57
Sharing others' pleasure .85 .95 21.43 c
Greeting .81 .92 18.61c
Social disinhibition .70 .94 ns
Communication Items All Levels
Form of babble as an infant .70 .91 5.806
Age in months at first words .89 .94 7.72 b
Age in months at first phrase .89 .95 ns
Emotional gestures .74 .90 17.90c
Enactive gestures .71 .93 9.37 b
Elicited vocal imitation .69 .90 9.6r
Reciprocal vocalization .77 .93 ns
Understanding gesture .82 .93 23.35 c
Comprehension of language .79 .92 33.94c

Verbal Only
Amount of social language .89 .94 ns
Immediate echolalia .85 .95 5.32b
Gestures accompanying speech .77 .91 ns
684 Lord, Rutter, and Le Couteur

Report of events .81 .92 ns

Talk expressing interest in others .72 .87 ns
Vocal expression .81 .91 ns
Understanding plot .76 .92 1.82c
Intonation, rhythm, rate .72 .92 ns
Other
Unusual attachments to objects .64 .90 ns
Aggression to family member .75 .89 ns
Aggression to non-family members .77 .93 ns
Tantrums .96 .98 ns
Destructive behaviors .83 .94 ns
Self-injury .96 .99 5.96 b
Food fads .75 .86 5.77 b
Pica .94 .97 ns
Unusual fears .65 .78 8.16 b
Cry because of pain .65 .81 ns
Cry for social reasons .68 .8l ns
Faint or seizures .72 .95 ns
Fine motor skills .83 .96 ns
Gait .78 .92 ns
Clumsiness .57 .42 ns
Initiation of appropriate activities .85 .95 9.94 b
Curiosity .82 .97 10.63b
Unusual musical ability .64 .56 ns
Abnormal idiosyncratic negative response .30 .93 ns
Distress over changes in own routine .77 .91 ns
Distress over trivial changes in enviornment .86 .96 ns
Unusually good memory .83 .93 ns
Overactivity at home .73 .90 ns
Problems going to bed .77 .93 ns
Sleep problems .79 .92 ns
Rocking .90 .97 ns
Sensitive to noise .83 .91 5.62/'
Overall level of language .92 .97 ns
Concerns about hearing .66 .84 ns
a All items refer to current behavior (in last 3 months) unless otherwise
noted. Reliability analyses (kw, agreement) were performed on 11 subjects
for verbal items and 20 for all others. Validity analyses (F scores) were
performed using 30 subjects for verbal items and 50 for all others. F scores
are for main effect of diagnosis performed on ranks.
bp < .01.
Cp < .001.
Autism Diagnostic Interview-Revised 685

REFERENCES

American Psychiatric Association. (1980). Diagnostic and statistical manual of mental disorders
(3rd ed.). Washington, DC: Author.
American Psychiatric Association (1987). Diagnostic and statistical manual of mental disorders
(3rd ed., rev.). Washington, DC: Author.
American Psychiatric Association. (1993). Options book for DSM-IV. Washington, DC: Author.
Baron-Cohen, S., Leslie, A. M., & Frith, U. (1985). Does the autistic child have a "theory of
mind?" Cognition, 2L 37-46.
Baron-Cohen, S., Tager-Fiusberg, H., & Cohen, D. (Eds.). (1993). Understanding other minds:
Perspectives on the theory of mind hypothesis of autism. Oxford, U.K.: Oxford University
Press.
Bayley, N. (1969). Manual for the Barley Scales ofhlfant Development. New York: Psychological
Corp.
Cohen, J. (1968). Weighted kappa: Nominal scale agreement with provision for scaled
disagreement of partial credit. Psychological Bulletin, 70, 213-220.
Conger, A. J. (1980). Integration and generalization of kappas for multiple raters.
Frith, U. (1989). Autism: Explaining the enigma. Oxford, U.K.: Basil Blackwell.
Hagberg, M. M., Aicardi, J., Dias, K., & Rasmos, O. (1983). A progressive syndrome of autism,
dementia, ataxia, and loss of purposeful hand use in girls: Rett syndrome: Report of 35
cases. Annals of Neurology, 14, 471-479.
Le Couteur, A., Rutter, M., Lord, C., Rios, P., Robertson, S., Holdgrafer, M., & McLennan,
J. D. (1989). Autism Diagnostic Interview: A semistructured interview for parents and
caregivers of autistic persons. Journal of Autism and Developmental Disorders, 19, 363-387.
Lord, C., Pickles, A., McLennan, J., Rutter, M., Bregman, J., Folstein, S., Fombonne, E.,
Leboyer, M. and Minshew, N. (in press). Diagnosing autism: Analyses of data from the
Autism Diagnostic Interview. In A. Frances (Ed.), DSM-IV sourcebook. Washington:
APA.
Lord, C., & Rutter, M. (1994). Autism and pervasive developmental disorder. In M. Rutter,
E. Taylor, & L. Hersov (Eds.), Child and adolescent psychiatry (3rd ed.) (pp. 569-593).
Oxford: Blackwell Scientific.
Lord, C., Storoschuk, S., Rutter, M., and Pickles, A. (1993). Using the ADI-R to diagnose
autism in preschool children. Infant Mental Health Journal, 14, 234-252.
Olsson, B., & Rett, A. (1990). A review of Rett syndrome with a theory of autism. Brain and
Development, 12, 11-15.
Short, C. B., & Schopler, E. (1988). Factors relating to age of onset in autism. Journal of
Autism and Developmental Disorders, 18, 207-216.
Stutsman, R. (1931). Guide for administering the Merrill-Palmer Scale of Mental Tests. In
L. M. Terman (Ed.), Mental measurement of preschool children (pp. 139-262). New York:
Harcourt, Brace & World.
Treiman, D. J. (1977). Occupationalprestige in comparative perspectives. New York: Academic
Press.
Tsai, L. Y. (1992). Is Rett syndrome a subtype of pervasive developmental disorders? Journal
of Autism and Developmental Disorders, 22, 551-561.
Voikmar, F. R., Klin, A., Siegal, B., Szatmari, P., Lord, C., Cicchetti, D. V., & Rutter, M.
(in press). DSM-IV autism/pervasive development disorder field trial final report.
American Journal of Psychiatry.
Wolff, P., Gardner, J., Paccia, J., & Lappen, J. (1989). The greeting behavior of Fragile X
males. American Journal on Mental Retardation, 93, 406-411.
World Health Organization. (1987). ICD-IO 1986 draft of chapter 5 categories FOO-F99. Mental,
behavioral and developmental disorders. Geneva: Author.
World Health Organization. (1992). The ICD-IO classification of mental and behavioral
disorders: Clinical descriptions and diagnostic guidelines. Geneva: Author.