These notes represent a summary of two further sources of information and observation. 

Firstly, conference presentations at ABA Ireland and, secondly, a paper describing a replication of a longitudinal study of ABA intervention with two matched groups differing in the level of intensity of input and frequency of supervision.

The common conclusions highlight the benefits of the ABA approach for some, but not all, young children with autism; and there is reference made to variables which appear to have predictive validity in determining which children will make rapid and positive progress and those whose progress will be limited.


M.J.Connor                                                                                            February 2006


Notes from Autism Conference – December 2005


Summaries of presentations made by Gina Green and by Jane Howard in respect of the current state of play with regard to ABA were made available to the present writer (MJC) and are helpful in highlighting both the benefits for many young autistic children of this approach and some continuing questions and reservations.


Gina Green describes how the essence of ABA is a careful observation, recording, and analysis of behaviour, adapted from the initial work of Skinner, which seeks to develop skills and to reduce maladaptive behaviours in individuals with and without disabilities. 


The emphasis is upon positive reinforcement by which to increase the probability of occurrence of certain behaviours.

Component teaching involves the breaking down of skills into small steps which can be specified and observed and measured, and “Discrete Trial Training” is used to teach these component elements through an ABC approach … antecedent, behavioural response, consequence, with adult cues and prompts by which gradually to shape the behaviour as required and which can be faded over time.


A close structure is seen as important, albeit within the child’s natural situation of home or nursery, and minimal time is left without direct attention and stimulation directed towards the child in order to avoid opportunities for the practice of interfering and characteristically-autistic behaviours.


Any maladaptive behaviours are not subject to any reinforcement, and the goal is to replace such behaviours with acceptable responses via the task analysis and the systematic/selective reinforcers.


The quantity of empirical evidence is noted, demonstrating the effectiveness of this kind of approach for many young children with autism.


However, not all children show positive progress, and the present writer - MJC - remembers both the advice of Lovaas himself to the effect that the procedures are intensive and potentially very challenging for both child and teachers or parents with the implication that one should not pursue such a course unless there is growing confidence in its likely effectiveness; and the findings set out in the series of papers from Reed et al (University of Wales, Swansea) which have indicated that ABA is not always appropriate (with higher levels of existing parental stress as a clear contra-indicator).


This being so, there continues to be a particular interest in defining those characteristics of the children and of their familial environment which appear to be consistently associated with a higher probability of positive outcomes, in terms of significant developmental progress, from the implementation of an (intensive) ABA programme. 


Green has reviewed a range of studies and suggest that successful programmes seem to be those which :


Green concluded that early and intensive ABA can lead to large improvements in multiple areas of functioning for many children, but that …..


The presentation by Jane Howard focused upon her own recent study which compared the use of an intensive ABA-type intervention with a set of “eclectic” approaches among samples of young children with autism.


Her introduction also indicated that a common goal among interventions is the closing of gaps in the developmental trajectories of the children in question compared with their normally-functioning peers, but that there are a number of unknowns in this field.

For example, she refers to the lack of ability to predict what progress can or will be made in response to the timing or type or intensity or duration of ABA intervention; nor is there very much information from direct and methodologically-valid comparisons of defined interventions such as ABA vs. TEACCH.


The study that Howard completed with her team compared 3 groups of children … those receiving a generic early years’ programme operating for about 15-17 hours each week in a small group involving access to speech therapy and a range of early years activities; an autism-specific class operating for up to 30 hours each week  and allowing some individual working and access to speech therapy, and elements of PECS, TEACCH, DTT, and circle time; and an intensive ABA group involving individual working up to 35 hours per week with the targets of enhanced imitation and attention, etc.

It is acknowledged that the children were not randomly assigned but that parental preference was the determining factor.

All the children were below the age of 48 months (the ABA group younger still) and all had a clear diagnosis of autism with no secondary medical complications.


After one year of intervention, the ABA group was shown to have made the greatest improvements in the areas assessed.

It was held that the issue was not simply the number of hours (intensity) and evidence was cited from an existing study to the effect that comparisons of children receiving a similar intensity of ABA and of an eclectic combination of approaches showed the greater benefits of the ABA working.


Her own hypotheses are that the children in question do benefit from a targeted and specific type of intervention (and this would fit to the common finding of the difficulty among children with autism in shifting attention especially when the shift is not just from one task to another but when there is a shift in the type of task and the necessary use of different skills or sensory modalities).

The combination of eclectic components do not so readily fit the style of the children in question, and the “jack of all trades ” programme does not provide sufficient focus or continuity or intensity of any one approach.


She also ponders whether the eclectic approach remains popular because there is no guaranteed success of any one intervention so that there is a kind of hedging of bets, and a hope or belief that a combination of approaches will include something to which there will be a positive response.


Gina Green then offered further thoughts to highlight the critical need for evidence-based practice by which to offer some protection to both providers and consumers.

This will involve the need for the close monitoring of interventions, both in terms of positive and negative outcomes, such that one can avoid unwarranted claims and questionable marketing while also gaining some confidence that the investment in time or financing will be legitimate. 

The availability of a valid choice of interventions, with their clearly delineated benefits and costs, would also be reassuring for parents as well as logical in seeking to match the range of needs.


What is needed, she argues, is a kind of protocol where a panel of experts from a range of related disciplines can evaluate performance data while ensuring that the evidence is based upon properly implemented and controlled methodology ….. including information about dependent and independent variables, the research design, controls, selection of participants and their allocation to conditions, treatment effects, and generalising of skills across settings and time.


 Outcomes and Predictors


One can now turn to a study which has offered at least a partial replication of the original Lovaas research findings in highlighting markedly positive benefits for many of the children participating in a controlled study of the effects of ABA under different conditions. 


The paper by Sallows and Graupner (2005) begins with a review of the various studies, beginning with the  Lovaas findings of 1987, which have demonstrated that intensive behavioural intervention can bring about dramatic improvements for a significant proportion of children.

However, they also recognise the methodological criticisms that were directed towards that initial 1987 paper in terms of the use of the term “recovery” to describe children whose measured ability fell into the average range and who were being educated within mainstream schools (when children with higher functioning autism could well achieve such outcomes via other interventions and that a follow-up of some of the recovered children at age 13 revealed a continuation of significant behavioural issues).

Questions were also raised about the use of aversives within the methods adopted which would no longer be deemed acceptable.


Importantly, Sallows and Graupner note that only about half of the children in the Lovaas target sample showed marked gains, and they highlight the ongoing need to identify features which will act as predictors of outcomes thus to clarify which children are likely to benefit from the ABA-type approach.


With regard to aversives, these authors refer to a number of studies, such as Smith et al (2000), where partial replication of the Lovaas findings have been sought without the use of aversives and where most have found that some children showed marked improvements in IQ. 

Fewer children reached an average level of functioning, but this was attributed to differences in methodology such as lower intensity and duration of intervention, different sample characteristics, and less training and supervision for the staff.


(The point about sample characteristics implies again the logic of seeking to specify what is as about the child or the circumstances which would predict the greater probability of positive outcomes to an ABA approach.  In this regard, one notes that the Smith et al study sample is said to have included an unusually high number of children who had no expressive language with the reasonable implication that it is the child with the relatively greater level of language skill at the time of initiating the intervention who is likely to respond more positively.)


The authors also refer to the results of behavioural intervention provided by staff working outside university settings in largely parent-managed treatment with the supervision of a consultant who is less frequently available than is typical for programmes completed within a recognised treatment centre.


One such study is that of Sheinkopf and Siegel (1998) who described results for children who received around 19 hours of treatment per week over a 16 month period.

6 of 11 children had achieved an IQ in the average range and 5 of them were in regular classes albeit showing residual behavioural needs.

However, it was argued that the sample of children was not comparable to high achievers noted in other studies because the cognitive measures used included the Merrill-Palmer Test which is geared towards non-verbal skills and may yield scores higher than those of more balanced tests which reflect both verbal and non-verbal skills.


(Another possible implication is that the child with the greater probability of positive outcomes is the one who is the relatively higher scorer on initial tests of ability which reflect verbal as much as non-verbal capacities.)


A further study quoted is that of Bibby et al (2002) who reported results for children receiving an average of 30 hours of parent-managed treatment (range 14 – 40 hours) over a 32 month period. 

10 of 66 children achieved an IQ in the average range and 4 were attending mainstream classes, but the less positive outcomes were attributed to the significant minority with a pre-intervention score of below 37, to the fact that more than half the sample were older than 48 months at the start, many received less than 20 hours per week, and the trainers were less experienced than university-based staff and did not achieve a level of at least weekly supervision of the people directly implementing the intervention.


(Again, an implication about predictor variables can be identified in terms of age of child and relative [measured] ability at the time of the start of the programme.)


Sallows and Graupner summarise the findings from a number of studies which have included references to pre-treatment variables and, while acknowledging some lack of consistency, they suggest that the most commonly observed predictors of outcomes include :


The use of multiple regression techniques has been used in seeking to determine which combinations of pre-treatment variables have the greatest association with outcomes.  Verbal imitation skill plus IQ plus age appear the combination most highly correlated with results. 

Meanwhile, rapid learning during the first 3 or 4 months of treatment has been associated with eventually positive outcome; as has rapid acquisition of verbal imitation plus non-verbal imitation plus receptiveness to instruction (as reflected in later re-administration of the CARS and the Vineland measures).


The study by Sallows and Graupner themselves set out to determine if a community-based programme and one without aversives can achieve positive outcomes similar to those cited in the Lovaas data; whether significant residual symptoms remain despite average post-treatment scores; and whether pre-treatment variables can be identified as accurate predictors of the outcomes.


The children were recruited from special educational settings with the criteria of age at entry between 24 and 42 months, a mental age index (mental age divided by chronological age) giving a figure of at least 35, within normal neurological limits, and a formal diagnosis of autism from a experienced practitioner with a match to the guidelines of DSM-IV and the Autism Diagnostic Interview.

13 children began the programme in 1996, 11 in 1997, and 14 in 1998-1999.  23 of the children (19 boys and 5 girls) had completed 4 years of intervention or had “graduated” earlier at the time of the report written by the current authors.  Children were matched on pre-treatment IQ and assigned randomly to a clinic-directed group

(similar to the Lovaas target group in receiving 40 hours per week of treatment) or a parent-directed group intended as a less intensive intervention  (around 32 hours with the parents receiving 6 hours per month of supervision plus a consultation every 2 months with a senior clinician).

The average age at pre-test was 33 months, and 35 to 37 months was the typical age when intervention began.


The measures used included the Bayley Scales of Infant Development, Merrill-Palmer for a measure of non-verbal IQ, Reynell Language Scales, and subtests of the Vineland Adaptive Behaviour Scales.  Follow up testing was implemented annually for 4 years with age appropriate tests coming into use, such as the WPPSI or WISC-III, and the Leiter Scale, alongside those used at pre-treatment testing where still appropriate.

Social functioning post-treatment was assessed by the Autism Diagnostic Interview Revised, and the Personality Inventory for Children. 

At age 7, the children were rated by parents and teachers on the Child Behaviour Checklist, and achievements were measured by the Woodcock-Johnson tests.


One further and idiosyncratic measure was used, namely the Early Learning Measure developed at the University College of Los Angeles and designed to assess the rate of acquisition of skills during the first few months of treatment.

Every 3 weeks during the 3 months leading to the start of treatment and for 6 months afterwards, the same list of 40 items known only to the experimenter (involving verbal and non-verbal imitation, following verbal instructions, and expressive object labelling) was used.  The measures derived indicated the number of correct items achieved before the start of the programme, and the number of weeks required for the child to achieve 90% of the verbal imitation items as a reflection of rate of progress. 


The overall results summarised by the authors demonstrated that the Lovaas-style early behavioural treatment could be implemented in a clinical setting outside a university and that favourable outcomes could be achieved without the use of aversives.


Following between 2 and 4 years of treatment, 11 of 23 children achieved full IQs in the average range as well as increases in language and adaptive behaviours.  At age 7, these children were working successfully in mainstream classes showing average attainments, adequate language, and regular play interaction with peers.


Parent-directed children did largely as well as those in the first group despite the lower intensity and less supervision and advice available to the adults. 

This finding was described as unexpected, and it was hypothesised that it was a matter of the parents taking on the senior therapist role, filling gaps in  the day to day input, and involving neighbours and others to provide the consistency of attention to, and play with, the children.  Further, it was held that the parents may have experienced problems reflected in the slow progress or non-compliance of their children by trying to push their children too far or too fast or using inappropriate interventions, but were able to gain advice without delay and to avoid repeating the circumstances such that their own skills advanced rapidly.


Meanwhile, it was recognised that, while not severe or of clinical significance, some residual symptoms of autism were observed, especially among those children who achieved the average IQ rating only after several years of treatment.

Among the more rapid learners, around a third were observed to have mild delays in social skills.  “Appearing preoccupied” was also a common problem for which 3 children were allocated classroom aides to provide reminders about staying on task.


With regard to predicting outcomes, the strongest pre-treatment features were the levels of …..

Also strong predictors were the rapid acquisition of new material as measured by the Early Learning Measure, and change in IQ after 1 year.


The authors refer to a model which appears to have the potential for predicting which children will prove the rapid learners … one of whose components is that of pre-treatment verbal imitation … but further validation studies are required.

One likely problem with the model is the limitation in discriminatory power among children above a certain age level, not yet clarified, given that children often acquire imitation skills by school age.


The comparability of the Bayley Test with the Wechsler Test was demonstrated by the authors such that the switch from the Bayley at initial assessment to the WPPSI at subsequent assessment was not seen as a confounding factor. 

In addition, there was raised the question whether the positive findings among the rapid learning group could be due to maturation as much as to the actual input; but the maturation hypothesis was thought to be discountable in the light of the negligible improvement of children receiving (non-intensive) community services observed in a number of longitudinal studies …. such as that of Eikeseth et al (2002).


The authors do acknowledge some methodological issues including the small numbers of children involved, and the number of tests used, with the risk of some spurious findings and some question over drawing implications for a wider population of children with autism. 


Two further findings were highlighted by the authors.


Firstly, the ratings of parental involvement were only weakly related to outcomes suggesting that more specific efforts to increase parental feelings of competence and self-belief in being able to contribute to treatment planning could enhance effects.


Secondly, the acquisition of social skills was positively associated with the amount and duration of supervised play, and it may be that variation in the relevant scores among the children reflected the variation in parental opportunities or willingness to arrange play sessions with the children of neighbouring families …. although it was hoped that disadvantages in this one aspect of social skill learning could be compensated by further experience of alternative stimuli such as video modelling.


The authors noted that their study is unusual in approaching the level of input for the clinical group that was available in the initial Lovaas study, but they still conclude that high hours and intensive supervision were not sufficient to make up for low levels of pre-treatment skills.


Instead, they hold that, consistent with existing studies, low IQ scores (below 44) and absence of language (no words at 36 months) predicted limited progress; while rate of learning, imitation, and social relatedness predicted favourable outcomes.


                                    *          *          *          *          *          *


M.J.Connor                                                                                            February 2006





Bibby P., Eikeseth S., Martin N., Mudford O., and Reeves D.  2002   Progress and outcomes for children with autism receiving parent-managed intensive interventions.   Research in Developmental Disabilities  23  81-104


Eikeseth S., Smith T., Jahr E., and Eldevik S.  2002   Intensive treatment at school for 4- to 7-year old children with autism.   Behaviour Modification  26  49-68


Green G. 2005 / Howard J.  2005    Presentations to the conference  “ Facing Autism : ABA Ireland ”   Coleraine , University of Ulster.  16-17 December


Howard J., Sparkman C., Cohen H., Green G., and Stanislaw H.  2005   A comparison of intensive behaviour analytic and eclectic treatments for young children with autism.  Research in Developmental Disabilities  26  359-383


Lovaas O.  1987   Behavioural treatment and normal educational and intellectual  functioning in young autistic children.   Journal of Consulting and Clinical Psychology  55  3-9


Sallows G. and Graupner T.  2005   Intensive behavioural treatment for children with autism.   American Journal on Mental Retardation  110(6)  417-438


Sheinkopf S. and Siegel B.  1998   Home-based behavioural treatment of young children with autism.   Journal of Autism and Developmental Disorders  28  15-23


Smith T., Groen A., and Wynn J.  2000   Randomised trial of intensive early intervention for children with pervasive developmental disorder.   American Journal on Mental Retardation  105  269-285



This article is reproduced by kind permission of the author.

© Mike Connor 2006.

Back to NAS Surrey Branch Welcome Page