See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/5969146

The Autism Observation Scale for Infants: Scale Development and Reliability
Data

Article  in  Journal of Autism and Developmental Disorders · May 2008

CITATIONS READS

212 6,842

5 authors, including:

Lonnie Zwaigenbaum Jessica A Brian

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Baby Sib project View project

ICF autism core set project View project

All content following this page was uploaded by Jessica A Brian on 01 December 2014.

The user has requested enhancement of the downloaded file.

ORIGINAL PAPER

The Autism Observation Scale for Infants: Scale Development

Ó Springer Science+Business Media, LLC 2007

Abstract The Autism Observation Scale for Infants Autism forms part of a spectrum of related neurodevelop-
(AOSI) was developed to detect and monitor early signs of mental disorders referred to as the autistic spectrum
autism as they emerge in high-risk infants (all with an older disorders (ASDs; hereafter, autism), the prevalence of
sibling with an autistic spectrum disorder). Here we describe which has been estimated recently at 6–7 per 1,000
the scale and its development, and provide preliminary data (Chakrabarti and Fombonne 2001; also see Bryson 1997;
on its reliability. Inter-rater reliability both for total scores Gillberg and Wing 1999). The prevalence, severity and
and total number of endorsed items is good to excellent at 6, life-long impact of these disorders render them a health
12 and 18 months; reliability is more modest for individual care priority. Evidence of improved outcomes with early,
items, particularly in 6-month-olds. Test-retest reliability of intensive intervention (Dawson and Osterling 1997; Harris
the AOSI at 12 months of age is within acceptable limits. et al. 1991; Rogers 1998) has served to underscore the
Evidence that the AOSI provides reliable data is the first importance of early identification and treatment. Indeed,
critical step towards evaluating its efficacy in distinguishing the US National Institutes of Health have emphasised that
high-risk infants who develop ASD. for autism ‘‘the urgency of early identification and treat-
ment puts the quest for infant screening and diagnostic
Keywords Autistic spectrum disorder
Early detection
instruments in the forefront of our priorities’’ (Bristol-
Behaviour
Assessment Power and Spinella 1999; also see Filipek et al. 2000).
Until recently, autism was rarely diagnosed prior to age
S. E. Bryson four (Howlin and Moore 1997; Siegel et al. 1988), often 2
Departments of Pediatrics and Psychology, Dalhousie University years after parents first raised concerns, although increas-
and IWK Health Centre, Halifax, NS, Canada ingly 2–3-year-olds are being diagnosed in centers with
specialised expertise and experience (e.g. Eaves and Ho
S. E. Bryson (&)
Autism Research Centre, IWK Health Centre, 5850 University 2004). Evidence that signs of autism are apparent within
Avenue, Halifax, NS, Canada B3K 6R8 the first 2 years of life comes from several sources, nota-
e-mail: [email protected] bly, parents’ retrospective reports (e.g. Dahlgren and
Gillberg 1989; Ohta et al. 1987), analyses of home videos
L. Zwaigenbaum
Department of Pediatrics, University of Alberta, Edmonton, AB, of children later diagnosed with autism (e.g. Adrien et al.
Canada 1992; Baranek 1999; Osterling and Dawson 1994) and case
studies (Dawson et al. 2000; Sheinkopf et al. 2000). These
C. McDermott
findings have resulted in a flurry of efforts to develop
York University, Toronto, ON, Canada
screening instruments aimed at detecting autism earlier in
V. Rombough life, typically at 24 but as early as 18 months (Adrien et al.
North York General Hospital, Toronto, ON, Canada 1992; Charman 2002; Robins et al. 2001; Siegel 1988;
Stone et al. 2000, 2004; Wetherby et al. 2004). Whether
J. Brian
The Hospital for Sick Children and Bloorview Kids Rehab, designed as first-level (population-wide) screens or second-
Toronto, ON, Canada level screens for children referred with developmental

123
 J Autism Dev Disord

problems, the specificity and sensitivity of these instru- study of infant siblings of children with autism. Inter-rater
ments are at various stages of assessment. Of those reliability was assessed at 6, 12, and 18 months of age, and
developed as first-level screens, only one instrument, the test-retest reliability at 12 months of age.
Checklist for Autism in Toddlers (CHAT), has been sys-
tematically evaluated in a general population, with
sufficient follow-up to identify misclassification errors The AOSI and Its Development
detectable only later in development (Baird et al. 2000).
Despite its initial promise (Baron-Cohen et al. 1992), The AOSI is an 18-item direct observational measure
subsequent follow-up has indicated that over 80% of designed to detect and monitor putative signs of autism in
children diagnosed by age seven are missed by the CHAT infants aged 6–18 months. This is accomplished through a
at initial 18-month-old screening. Moreover, attempts to standard set of semi-structured activities, administered by
modify the cut-off criteria to increase sensitivity (from 18 an examiner who is both skilled at interacting with infants
to 38%) resulted in an unacceptable decrease in positive and knowledgeable about autism. The activities provide an
predictive value (from 75 to 5%). interactive context in which the examiner engages the
One critical outstanding question is whether these find- infant in play, while conducting a set of systematic presses
ings reflect real variation in the onset of autism, or whether to elicit particular target behaviours. The relative presence
the CHAT lacks sensitivity to the earliest signs as they or absence of these ‘‘pressed for’’ behaviours is rated by the
emerge in the first 2 years of life. Although the CHAT was examiner, as is an additional set of behaviours, which the
designed for screening at 18 months of age, like other early examiner targets for observation throughout the entire
screens for autism, its items are restricted to signs diag- assessment. The AOSI is conducted at a small table, with
nostic of autism in preschoolers. The problem is that the infant seated on his/her parent’s lap, across from and
currently very little is known about the earliest signs of facing the examiner. Parents are encouraged to assist in
autism or the timing of their onset. Evidence to date comes making the infant comfortable, but otherwise to assume an
from sources (parents’ retrospective reports, home videos observer role. The AOSI takes approximately 20 min to
and case studies) that may be biased or otherwise limited administer, and sessions are videotaped both to assist in
(Zwaigenbaum et al. 2006). It is thus possible that a better current behavioural ratings and to provide a database for
understanding of the early emergence of autism would future purposes.
allow us to develop more effective early screening and Development of the AOSI involved four steps: (1)
diagnostic instruments. identification of the behaviours to be targeted for assess-
In an attempt to address this need, we have embarked on ment, (2) development of activities appropriate to eliciting
a large, multi-site prospective study of high-risk infants, and assessing the target behaviours, (3) operalisation of the
defined as such by virtue of having a sibling with autism/ target behaviours and their ratings, and (4) revision and
ASD (Bailey 1996; see Bryson et al. 2007; Zwaigenbaum refinement of the instrument through pilot testing. Target
et al. 2005, for study details). Our main goal is to identify behaviours were drawn from available data on the earliest
and characterise both the nature and timing of early signs signs of autism, as derived from parents’ retrospective
of autism, and their natural variation, within this high-risk reports (e.g. Gillberg et al. 1990), early home videotapes
group (hereafter, infant siblings). For this purpose, we have (e.g. Adrien 1992) and case studies of children later diag-
developed the Autism Observation Scale for Infants nosed with autism (Dawson et al. 2000; Sheinkopf et al.
(AOSI; Bryson et al. 2000). While overlap exists in the 2000), and from our collective clinical experience with
constructs assessed (i.e., social-communicative develop- toddlers with autism. These behaviors include visual
ment, and atypical sensory-motor and/or repetitive tracking and attentional disengagement, coordination of
behaviours), the AOSI differs from the instruments out- eye gaze and action, imitation, early social-affective and
lined above in that it was designed to be developmentally communicative behaviours, behavioural reactivity, and
appropriate for infants. various sensory-motor behaviours (see Table 1 for
The potential utility of the AOSI as an early screen or description of behaviours assessed).
diagnostic instrument for autism remains to be determined.
Data on its validity in predicting infant siblings who are
subsequently diagnosed with autism will be addressed in an AOSI Items: Administration and Coding of Behaviours
upcoming paper. In the meantime, the purpose of the
present paper is to describe the AOSI and its development, Target behaviours are assessed within a standard set of
and to provide preliminary data on the inter-rater and test- seven activities, in which the infant is engaged by using
retest reliability of our newly developed measure. The various toys (several rattles, a bell, blocks, a book, soft
latter were evaluated in the context of our prospective balls, a rubber duck, a plastic stick and a blanket), and

123
J Autism Dev Disord

Table 1 Description of the behaviours assessed in the AOSI

Visual tracking Ability to visually follow a moving object laterally across the midline.
Disengagement of attention Ability to disengage and move eyes/attention from one of two competing visual stimuli.
Orientation to name Ability to move head and/or eyes toward and look at the examiner when name is called.
Differential response to facial emotion Ability to respond differentially through facial, head or other motor movements to a change in the
examiner’s facial expression from smiling to a neutral expression.
Anticipatory social response Ability to anticipate and enjoy social (vs. physical) cause-effect relationships.
Imitation Ability to reproduce an action produced by the examiner.
Social babbling Ability to engage in back-and-forth (reciprocal) vocalisations with the examiner.
Eye contact Ability to consistently establish appropriately sustained eye contact with the examiner.
Reciprocal social smile Ability to smile in response to the examiner’s smile.
Coordination of eye gaze and action Ability to co-ordinate gaze with actions on objects.
Behavioural reactivity General responsiveness, including under reactivity and over reactivity, to the activities and toys
introduced, and to the examiner’s actions.
Cuddliness Physical responsiveness to being cuddled by parent.
Soothability Ease of being soothed by parent using social means such as touch, other forms of human contact,
or verbal reassurance.
Social interest and shared affect Ease of engagement and interest in activities, and ability to share positive affect with the
examiner.
Transitions Ease and consistency with which toys are relinquished and movement is made from one activity to
another.
Motor control Degree to which motor behaviour is goal-directed, organised and modulated.
Atypical motor behaviour Presence of developmentally atypical gait, locomotion, motor mannerisms/postures or repetitive
motor behaviours.
Atypical sensory behaviour Presence of developmentally atypical sensory behaviours in any modality (e.g. smelling of toys,
staring at hands/shapes/objects, or feeling textures).

systematic presses are designed to elicit particular target ‘‘peek-a-boo’’, the infant’s ability to anticipate seeing the
behaviours (Visual Tracking, Disengagement of Attention, examiner’s face is assessed. For Orientation to Name, the
Orientation to Name, Reciprocal Social Smiling, Differ- examiner moves away from and to one side of the infant,
ential Response to Facial Emotion, Social Anticipation and and, while the infant is looking elsewhere, assesses whether
Imitation). In the case of Visual Tracking, the infant’s the infant will orient to, and look at, the examiner when
attention is engaged by shaking a rattle at midline; the his/her name is called. For the Imitation item, the examiner
rattle is then positioned to one side of the infant, and his/ assesses the infant’s ability to reproduce either an oral-
her ability to laterally track the rattle is assessed as the facial movement (tongue protrusion or round mouth
rattle is moved silently at eye level across the midline from movement for 6-month-olds) or an action with an object
one side to the other. For Disengagement of Attention, a (e.g. patting a ball or tapping the table with a plastic stick)
rattle is shaken to one side of the infant, and, once his/her for 12- and 18-month-olds.
attention is engaged, a second rattle is shaken on the Each press is administered a predetermined number of
opposite side, and the infant’s ability to disengage and times, and trials are repeated if the infant is distracted or is
move his/her eyes/attention from the first to the second otherwise inattentive to the task at hand. The assessment
rattle is assessed. Reciprocal Social Smiling is assessed by typically begins with the presses for Visual Tracking and
observing whether the infant smiles in response to the Attentional Disengagement, although task order is flexible
examiner’s smile. For Differential Response to Facial and dependent on the interests of the infant. Presses for
Emotion, the examiner establishes eye contact with the Reciprocal Social Smiling, Orientation to Name, and Dif-
infant, and the infant’s facial expression, head or other ferential Response to Facial Emotion are interspersed
motor movements are assessed in response to a change in between other structured activities, as are two Free Play
the examiner’s facial expression from smiling to a neutral sessions, during which the examiner engages the infant in
expression. In the case of Social Anticipation, the infant is rolling a ball back and forth, playing with blocks and
engaged in a game of peek-a-boo, and when the examiner looking at a picture book, all designed to optimise the
is hidden behind a blanket (or her hands), preparing to say infant’s comfort and create opportunities for Social

123
 J Autism Dev Disord

Babbling. Observations of the remaining target behaviours relationship to other, as yet unidentified, early develop-
(see Table 1) are made throughout the entire interactive mental features of the disorder. Below we outline the
play assessment. methods used to assess inter-rater and test-retest reli-
Target behaviours have been operationalised and, with ability of the AOSI, and provide initial data on both.
three exceptions (Eye Contact, Atypical Motor Behaviour
and Atypical Sensory Behaviour), are rated on a scale from
0 to 2 or 3, where 0 implies typical function, and scores of Method
1–3 represent increasing severity of impairment. Eye
Contact and Atypical Motor and Sensory Behaviour are Participants
rated on a scale that is confined to 0 (typical) and 2
(atypical). In general, 0 represents typical behaviour, 1 Participants were recruited from consecutive self-referrals
represents inconsistent, partial or questionable behaviour, 2 to our ongoing prospective study of infant siblings of
represents atypical behaviour, and 3 represents a total lack children with a formal diagnosis of autism or autistic
of the behaviour. In the case of Visual Tracking, for spectrum disorder (ASD; which includes autism, Asperger
example, a score of 0 represents the ability to smoothly syndrome and atypical autism or pervasive developmental
track a silent object moved laterally across the midline on disorder-not otherwise specified; APA 2000). Diagnoses of
two presses/trials; a score of 1 implies delayed or inter- the probands have been independently validated: all meet
rupted eye movements (i.e. in the absence of any external formal DSM-IV-TR criteria for autism or ASD, as assessed
event, infant looks away and then returns gaze to object or by the Autism Diagnostic Interview-Revised (ADI-R; Lord
does not cross midline); a score of 2 implies partial visual et al. 1994; using either standard ADI-R criteria for autism
tracking or only tracking of noisy objects; and a score of 3 or the criteria for ASD recommended by Risi et al. 2006)
implies that the infant does not track objects laterally (but and the Autism Diagnostic Observation Schedule (Lord
may track vertically). For Social Interest and Shared et al. 2000), and by the judgment of experienced clinicians
Affect, which is assessed throughout the AOSI, a score of 0 with expertise in autism. For the purpose of assessing inter-
represents sustained interest and pleasure directed at the rater reliability of the AOSI, 32 infants (15 males) were
examiner; a score of 1 implies inconsistent interest and/or seen at 6 months of age (M = 6.7 mos; SD = 12.2 days;
little, if any, pleasure; a score of 2 implies interest or range = 6.1–7.3 mos), 34 (19 males) at 12 months of age
pleasure only in response to self-directed actions, or to toys (M = 12.5 mos; SD = 10.6 days; range = 12.0–12.8 mos),
or physical events such as tickling; and a score of 3 implies and 26 (10 males) at 18 months of age (M = 18.9 mos;
that the infant shows no interest or pleasure. And similarly, SD = 11.8 days; range = 18.0–19.1 mos). Most of the
for Transitions, which codes for the ease and consistency infants (approximately 60%) were seen at each age. Of the
with which the infant relinquishes toys and moves from 34 infants seen at 12 months of age, 11 returned for a
one activity to another, a score of 0 represents no difficulty; follow-up visit, 2 weeks after the 12-month visit, to assess
a score of 1 implies some resistance but can be redirected test-retest reliability of the AOSI; an additional nine new
with no distress; and a score of 2 implies repeated and cases (i.e. infants not assessed for inter-rater reliability,
marked difficulty with distress or disruption to the assess- also seen 2 weeks after their 12-month visit, were included,
ment. Item content (except for the Imitation item, as for a total of 20 infants (10 males; M = 12 mos; SD =
described above) and the criteria for rating target behav- 11.4 days; range = 12.1–12.9 mos) assessed for test-retest
iours (except for Motor Control, which takes account of the reliability. All children were seen at the Autism Research
child’s age) are the same for infants ranging from 6 to Unit at The Hospital for Sick Children in Toronto, or the
18 months, with an emphasis on the quality and consis- Autism Research Centre at the IWK Health Centre in
tency of behaviors to help ensure that all items can be Halifax. Written informed consent was obtained from
meaningfully coded across this age range. parents of the infant participants.
Over a period of more than 2 years prior to the
present reliability assessment, the scale was revised and
refined through piloting various methods of eliciting and Procedure
rating the behaviours in low-risk and various high-risk
infants (notably, premature infants, those with infantile Inter-rater reliability was assessed at each of the three ages
spasms and those with a sibling with autism) aged 6– (6, 12 and 18 mos). One of four examiners (CM, VR, JB
18 months. Currently, the AOSI serves as a research and AW), each trained in the administration and scoring of
instrument, and in that capacity its purpose is both to the AOSI, assessed each infant individually. During each
detect and monitor the earliest signs of autism, and to infant’s assessment, the AOSI was scored by one inde-
yield a better understanding of their nature and pendent rater (CM, VR, JB or AW) and by the examiner.

123
J Autism Dev Disord

The rater was present in the room for the duration of the Table 2 Descriptive data on AOSI total marker counts and total
assessment. scores by age and rater
Test-retest reliability of the AOSI was assessed at Age Mean Standard deviation Range
12 months of age on two occasions separated by 2 weeks.
6 months
One of three examiners (CM, VR or TM) assessed each
infant on the two occasions. The examiner was blind to Total marker count
each infant’s previous AOSI results. Rater 1 3.17 1.37 1–6
Rater 2 3.67 1.34 2–7
Total score
Results Rater 1 4.46 1.79 2–9
Rater 2 5.25 2.07 2–11
For each infant at each age, scores of each of the two raters 12 months
(the examiner and independent rater) on each of the AOSI Total marker count
items were computed, and were summed to yield both a Rater 1 2.84 2.50 0–10
total number of items endorsed (‘‘total marker count’’, out Rater 2 3.03 2.36 0–9
of a possible 18; recodes scores of 1 and higher as 1, Total score
treating AOSI items as behavioural markers that are either Rater 1 3.77 3.48 0–12
present or absent) and a total score (out of a possible 50). Rater 2 3.97 3.37 0–14
Scores on some of the individual items were highly nega- 18 months
tively skewed due to the large number of 0 or ‘‘typical’’ Total marker count
ratings. Descriptive statistics on the AOSI (mean total Rater 1 4.00 2.76 0–10
marker counts and total scores, and respective standard Rater 2 3.78 2.88 0–10
deviations and ranges) are presented in Table 2. Total score
Rater 1 5.70 4.56 0–18
Rater 2 5.43 4.79 0–17
Inter-rater Reliability Across ages
Total marker count
We first examined the inter-rater reliability of individual Rater 1 3.28 2.33 0–10
AOSI item scores. For each of the 18 items at each age (6, Rater 2 3.45 2.28 0–10
12 and 18 months) and across ages, both binary scores Total score
(present/absent) and 0–2/3 ratings were examined using Rater 1 4.55 3.50 0–18
unweighted kappas. Note that the maximum AOSI total Rater 2 4.79 3.57 0–17
score is 50 (maximum score of 2 on 4 items and 3 on 14
items, for a total of 18 items/marker counts). As shown in
Table 3, reliability of individual items, both for binary Discussion
scores and 0–2/3 ratings, is generally good to excellent (i.e.
[.65; Landis and Koch 1977), although more modest for a The present study examined inter-rater and test-retest
subset of items (notably, Items 8, 11, 13, 14, 16 and pos- reliability of the Autism Observation Scale for Infants
sibly 18), particularly in 6-month-olds. We then computed (AOSI; Bryson et al. 2000), a newly developed direct
the inter-rater reliability of both total marker counts and observational measure designed to detect early signs of
total scores using intra-class correlations. The reliability of autism. This was accomplished within the context of an
AOSI total marker counts and total scores, respectively, is ongoing prospective study of high-risk infants, all with a
excellent at 6 (.68 and .74), 12 (.92 and .93) and 18 months sibling with autism/ASD. Inter-rater reliability both for
(.93 and .94), and across ages (.90 and .92). total marker counts (i.e. the number of endorsed items) and
total scores is good to excellent at 6, 12 and 18 months
(.68–.94). Reliability is more modest for individual items,
Test-retest Reliability particularly in 6-month-olds. Test-retest reliability, asses-
sed at 12 months of age, is well within acceptable limits
Test-retest reliability of the AOSI at 12 months of age was (.61 and .68, for total scores and total marker counts,
calculated using intra-class correlations. Although the respectively).
sample size is small (n = 20), test-retest reliability both for The present findings are encouraging, although several
total marker counts and total scores is fair to good at .68 points bear emphasizing. First, we acknowledge the
and .61, respectively. skewed distribution of scores on some items, reflecting the

123
 J Autism Dev Disord

Table 3 Inter-rater reliability of individual AOSI items by age and Table 3 continued
across ages
Item 6 mos 12 mos 18 mos Across ages
Item 6 mos 12 mos 18 mos Across ages
17. Atypical motor behaviour
1. Visual tracking Binary score – .72 1.0 .81
Binary score 1.0a .92 .66 .87 0–2 rating – .72 1.0 .81
0–3 rating 1.0 .92 .66 .87 18. Atypical sensory behaviour
2. Disengagement of attention Binary score – .62 – .47
Binary score .76 .79 –b .78 0–2 rating – .62 – .47
0–2 rating .66 .79 – .72 a
Unweighted kappas
3. Orients to name called b
Insufficient variability in scores
Binary score 1.0 1.0 1.0 1.0
– Insufficient data
0–3 rating 1.0 1.0 1.0 1.0
4. Differential facial emotion
Binary score .64 .67 .91 .77 relatively low frequencies of non-zero scores. On the
0–3 rating – .62 .92 .73 assumption that AOSI items measure early signs of autism,
5. Anticipatory response it is expected that they would be more commonly observed
Binary score .92 .81 .90 .87 in siblings who have the disorder than in those who do not.
0–3 rating .86 .67 .81 .77 As such, the overall rates of the behaviors of interest may
6. Imitation be relatively low, even within a high-risk sample of sib-
Binary score .68 .53 .69 .66 lings. To some extent, this is an inevitable feature of any
0–3 rating .73 .54 .62 .67 early identification tool that focuses on behaviors that have
7. Social babbling varying degrees of sensitivity and specificity to the target
Binary score .81 .68 .91 .81
disorder, but does create challenges in the assessment of
0–3 rating .71 .63 .83 .73
reliability. We have selected analytic techniques (i.e. kappa
8. Eye contact
and intra-class correlation) that correct for chance agree-
ment (although this may be less for ICCs), so that estimates
Binary score .37 1.0 1.0 .71
of reliability are not inflated by multiple ratings of zero.
0–3 rating .38 1.0 – –
Notably, total scores on the AOSI, the distributions of
9. Reciprocal social smile
which are less skewed, generally have higher levels of
Binary score .69 .64 .76 .70
reliability than individual items. Accordingly, our initial
0–3 rating .68 – .78 .69
data suggest that total scores are a more robust predictor of
10. Gaze-action co-ordination
autistic symptoms at 24 months than scores on individual
Binary score .63 .78 .79 .74
items (Zwaigenbaum et al. 2005). Moreover, of those who
0–3 rating .65 .67 .61 .67
received a blind, clinical best estimate diagnosis of autism
11. Reactivity
at the age of 3 years, 11 of 14 had an AOSI total score of 9
Binary score .50 .57 .66 .58
or more at 12 months (Brian et al. 2006). Infant siblings
0–3 rating .29 .59 .55 .48
diagnosed with autism at 3 years received mean AOSI total
12. Cuddliness scores of 10.9 (SD = 6.8; range = 0–24) at 12 months and
Binary score 1.0 .78 – .74 13.2 (SD = 7.5; range = 3–25) at 18 months; mean total
0–3 rating 1.0 – – – scores obtained at 6 months (M = 6.1; SD = 2.9; range =
13. Soothability 2–11) overlapped more with total scores of non-diagnosed
Binary score .48 1.0 .40 .60 siblings and typical controls.
0–3 rating .50 – .27 .50 Although the reliability of individual AOSI items is
14. Social interest and affect generally good, lower estimates are obtained for a subset of
Binary score –.05 .65 .75 .55 items, notably Eye Contact, Behavioural Reactivity,
0–3 rating –.05 .72 – – Soothability, Social Interest and Shared Affect and Motor
15. Transitions Control, particularly in 6-month-olds. Less reliable mea-
Binary score .77 .72 .75 .77 surement of these constructs in the youngest group is
0–2 rating .77 – .73 .73 complicated by their less differentiated behaviour. Other
16. Motor control possible sources of unreliability include inadequate oper-
Binary score .33 .44 .60 .45 ationalisation of relevant AOSI items or their differential
0–3 rating .33 .44 .60 .45 ratings, and/or the confounding of our reliability

123
J Autism Dev Disord

assessment with the ages of those assessed. Most infants reliable is a critical first step. We do know that our high-
were assessed at each of three ages, the implication of risk infants show marked variation in total AOSI scores,
which is that we began with the 6-month-old assessments. and that elevated AOSI scores at 12 months are predictive
Thus, as the infants aged, the examiners became more of social-communication problems at 24 months (Zwai-
experienced and possibly more skilled at detecting the genbaum et al. 2005). Preliminary, more current data,
behavioural differences of interest. The potential signifi- derived from our entire cohort assessed to date, indicate
cance of this is underscored by our impression that such that a total AOSI score of 9 or more at 12 months is pre-
differences are typically subtle, if not inherently ques- dictive of an independent (blind) ‘‘gold standard’’ diagnosis
tionable, in 6-month-olds relative to those observed in of autism/ASD at 3 years of age. An upcoming paper will
older infants (Bryson et al. 2007). We plan to address this address the critical question of whether the AOSI is able to
issue in a future reliability study by adopting a cross-sec- distinguish early in life infant siblings who go on to
tional design in which either the oldest groups will be develop autism/ASD. In the meantime, evidence is pro-
assessed first, or, at a minimum, different age groups will vided here that our newly developed scale provides a
be assessed in parallel across time. Alternatively, it remains reliable means of documenting early signs thought to be
possible that behavioural signs of autism are simply less or indicative of autism.
even not reliably evident in 6-month-olds and rather
emerge later in development, as early as between 6 and Acknowledgement We are grateful to all of the families who par-
ticipated in this research. We also thank Brad Frankland for statistical
12 months of age (Bryson et al. 2007; Zwaigenbaum et al. advice, and Ann Wainwright (AW) and Theresa McCormick (TM) for
2005). Assuming that this is so, we have refined the AOSI assisting with the reliability assessments. The work was supported by
based on the present findings, and plan to systematically grants from the Hospital for Sick Children Foundation, the US
monitor development at several intervals during the 6- to National Alliance for Autism Research and the Canadian Institutes of
Health Research.
18-month age period.
We also recognise that the use of multiple examiners/
raters introduces an additional source of variance into our References
reliability data, and that our data on test-retest reliability of
the AOSI may be limited by the small sample size. The Adrien, J. L., Bathelemy, C., Perrot, A., Roux, S., Lenoir, P.,
Hameury, L., & Sauvage, D. (1992). Validity and reliability of
latter represents practical constraints on the part of parents, the infant behavioral summarized evaluation (IBSE): A rating
most of who find it difficult to return for a repeat visit close scale for the assessment of young children with autism and
in time to the normally scheduled 12-month assessment. developmental disorders. Journal of Autism and Developmental
We do note that the distribution of AOSI scores of the Disorders, 22, 375–395.
American Psychiatric Association. (2000). Diagnostic and statistical
infants who were assessed for test-retest reliability is manual of mental disorders (4th ed.). Washington, DC: Amer-
comparable to that of the larger sample of 12-month-olds, ican Psychiatric Association.
and that there is no evidence that data from one or two Bailey, A., Phillips, W., & Rutter, M. (1996). Autism: Towards an
infants might account for our findings. Rather, the consis- integration of clinical, genetic, neuropsychological, and neuro-
biological perspectives. Journal of Child Psychology and
tency in individual scores across the two sessions suggests Psychiatry, 37, 89–126.
that the AOSI is indeed reliable. Baird, G., Charman, T., Baron-Cohen, S., Cox, A., Swettenham, J.,
Evidence that the AOSI yields reliable data may reflect Wheelwright, S., & Drew, A. (2000). A screening instrument for
any one or some combination of the following factors. autism at 18 months of age: A 6-year follow-up study. Journal of
the American Academy of Child and Adolescent Psychiatry, 39,
First, and despite possible limitations of some items, con- 694–702.
siderable care has been taken in operationalizing the Baranek, G. T. (1999). Autism during infancy: A retrospective video
relevant constructs, and in refining the methods for eliciting analysis of sensory, motor and social behaviours at 9–12 months
and rating the behaviours of interest. We also emphasise of age. Journal of Autism and Developmental Disorders, 29,
213–224.
that our examiners are clinically skilled, knowledgeable Baron-Cohen, S., Allen, J., & Gillberg, C. (1992). Can autism be
about autism, and have been trained in the administration detected at 18 months? The needle, the haystack and the CHAT.
and scoring of the AOSI. Moreover, our impression is that British Journal of Psychiatry, 161, 839–943.
extensive, ongoing discussion regarding rating differences Brian, J., Zwaigenbaum, L., Bryson, S. E., Roberst, W., Szatmari, P.,
& Smith, I. (2006). Predictive validity of the autism observation
and confusion about the rating criteria and the behaviours scale for infants. Poster Presented at the International Meeting
themselves has resulted in improved reliability over time. For Autism Research, Hyatt Regency Montreal, Montreal,
Indeed, our experience serves to underscore that such dis- Canada.
cussions should ideally form an integral part of both Bristol-Power, M., & Spinella, G. (1999). Research on screening and
diagnosis in autism: A work in progress. Journal of Autism and
clinical and research training on such instruments. Developmental Disorders, 29, 435–438.
The clinical usefulness of the AOSI remains to be Bryson, S. E. (1997). Epidemiology of autism: Overview and issues
established, although evidence that the instrument is outstanding. In D. J. Cohen & F. R. Volkmar (Eds.), Handbook

123
 J Autism Dev Disord

of autism and pervasive developmental disorders (2nd ed., pp. Lord, C., Risi, S., Lambrecht, L., Cook, E. H. Jr., Leventhal, B. L.,
41–46). New York: Wiley & Sons. DiLavore, P. C., Pickles, A., & Rutter, M. (2000). The autism
Bryson, S. E., McDermott, C., Rombough, V., Brian, J., & diagnostic observation schedule-generic: A standard measure of
Zwaigenbaum, L. (2000). The autism observation scale for social and communication deficits associated with the spectrum
infants. Unpublished Scale, Toronto, ON. of autism. Journal of Autism and Developmental Disorders, 30,
Bryson, S. E., Zwaigenbaum, L., Brian, J., Roberts, W., Szatmari, P., 205–223.
Rombough, V., & McDermott, C. (2007). A prospective case Ohta, M., Nagai, Y., Hara, H., & Sasaki, M. (1987). Parental
series of high-risk infants who developed autism. Journal of perception of behavioral symptoms in Japanese autistic children.
Autism and Developmental Disorders, Special Edition, 37, 12– Journal of Autism and Developmental Disorders, 17, 549–563.
24. Osterling, J., & Dawson, G. (1994). Early recognition of children with
Chakrabarti, S., & Fombonne, E. (2001). Pervasive developmental autism: A study of first birthday home videotapes. Journal of
disorders in preschool children. Journal of the American Medical Autism and Developmental Disorders, 24, 247–257.
Association, 285, 3093–3099. Risi, S., Lord, C., Gotham, K., Corsello, C., Chrysler, C., Szatmari, P.,
Charman, T. (2002). Progress in screening and surveillance for autism Cook, E. H., et al. (2006). Combining information from multiple
spectrum disorders: The Q-CHAT. Symposium presentation at sources in the diagnosis of autism spectrum disorders. Journal of
the International Meeting for Autism Research, Orlando, Florida, the American Academy of Child and Adolescent Psychiatry, 45,
Nov. 1–2, 2002. 1094–1103.
Dahlgren, S. O., & Gillberg, C. (1989). Symptoms in the first two Robins, D. L., Fein, D., Barton, M. L., & Green, J. A. (2001). The
years of life. European Archives of Psychiatry and Neurological modified-checklist for autism in toddlers: An initial study
Science, 238, 169–174. investigating the early detection of autism and pervasive
Dawson, G., & Osterling, J. (1997). Early intervention in autism: developmental disorders. Journal of Autism and Developmental
Effectiveness and common elements of current approaches. In Disorders, 31, 131–144.
M. J. Guralnick (Ed.), The effectiveness of early intervention (pp. Rogers, S. J. (1998). Empirically supported comprehensive treatment
307–326). Balitmore, MD: Paul H. Brookes. for young children with autism. Journal of Clinical Child
Dawson, G., Osterling, J., Meltzoff, A. N., & Kuhl, P. (2000). Case Psychology, 27, 167–178.
study of the development of an infant with autism from birth to Sheinkopf, S., Lester, B., Lagasse, L., Bauer, C., Shankaran, S., Bada,
two years of age. Journal of Applied Developmental Psychology, H., et al. (2000). A case study of autism in early infancy. Paper
21, 299–313. presented at the American Academy of Child and Adolescent
Eaves, L. C., & Ho, H. H. (2004). The very early identification of Psychiatry 47th Annual Meeting, New York, NY, October 27,
autism: Outcome to age 4 1/2–5. Journal of Autism and 2000.
Developmental Disorders, 4, 367–378. Siegel, B., Pliner, C., Eschler, J., & Elliott, G. R. (1988). How
Filipek, P. A., Accardo, P. J., Ashwal, S., Baranek, G. T., Cook, E. H. children with autism are diagnosed: Difficulties in identification
Jr., Dawson, G., et al. (2000). Practice parameter: Screening and of children with multiple developmental delays. Journal of
diagnosis of autism: Report of the Quality Standards Subcom- Developmental and Behavioral Pediatrics, 9, 199–204.
mittee of the American Academy of Neurology and the Child Stone, W. L., Coonrod, E. E., & Ousley, O. Y. (2000). Brief report:
Neurology Society. Neurology, 55, 468–479. Screening tool for autism in two-year-olds (STAT): Develop-
Gillberg, C., Ehlers, S., Schaumann, H., Jakobsson, G., Dahlgren, S. O., ment and preliminary data. Journal of Autism and
Lindblom, R., et al. (1990). Autism under age 3 years: A clinical Developmental Disorders, 30, 607–612.
study of 28 cases referred for autistic symptoms in infancy. Stone, W. L., Coonrod, E. E., Turner, L. M., & Pozdol, S. L. (2004).
Journal of Child Psychology and Psychiatry, 31, 921–934. Psychometric properties of the STAT for early autism screening.
Gillberg, C., & Wing, L. (1999). Autism: Not an extremely rare Journal of Autism and Developmental Disorders, 6, 691–701.
disorder. Acta Psychiatrica Scandinavica, 99, 399–406. Wetherby, A. M., Woods, J., Allen, L., Cleary, J., Dickinson, H., &
Harris, S. L., Handelman, J. S., Gordon, R., Kristoff, B., & Fuentes, F. Lord, C. (2004). Early indicators of autism spectrum disorders in
(1991). Changes in cognitive and language functioning of the second year of life. Journal of Autism and Developmental
preschool children with autism. Journal of Autism and Devel- Disorders, 34, 473–493.
opmental Disorders, 21, 281–290. Zwaigenbaum, L., Bryson, S. E., Rogers, T., Roberts, W., Brian, J., &
Howlin, P., & Moore, A. (1997). Diagnosis in autism: A survey of Szatmari, P. (2005). Behavioral manifestations of autism in the
over 1,200 patients in the UK. Autism Intervention Journal of first year of life. International Journal of Developmental
Research and Practice, 1, 135–162. Neurosciences, 23, 143–152.
Landis, J., & Koch, G. (1977). An application of hierarchical kappa- Zwaigenbaum, L., Thrum, A., Stone, W., Baranek, G., Bryson, S. E.,
type statistics in the assessment of majority agreement among Iverson, J., Kau, A., Klin, A., Landa, R., Lord, C., Rogers, S.,&
multiple observers. Biometrics, 33, 363–374. Sigman, M. (2006). Studying the emergence of autism spectrum
Lord, C., Rutter, M., & Le Couteur, A. J. (1994). Autism diagnostic disorders in high risk infants: Methodological and practical
interview-revised: A revised version of a diagnostic interview for issues. Journal of Autism and Developmental Disorders, 37,
caregivers of individuals with possible pervasive developmental 466–480.
disorders. Journal of Autism and Developmental Disorders, 24,
659–685.

123

View publication stats