The effectiveness of early intervention programmes for autistic spectrum disorders.



A Report for the South East Regional Special Educational Needs Partnership.

Research Partners: Bexley, Brighton & Hove, East Sussex, Kent, Medway, Surrey, West Sussex.




Phil Reed, Lisa A. Osborne, & Mark Corness

University of Wales Swansea.


Please contact Professor Phil Reed if you have any queries:





Conclusions and Recommendations

  1. The current report failed to note any evidence of recovery from autism produced by any early intervention.
  2. In terms of intellectual functioning, Applied Behaviour Analysis and Special Nursery interventions proved to produce gains (of the same magnitude as many gains produced by previous longer-term clinic-based ABA programmes).
  3. The results from clinic-based ABA trials were partially replicated on a community-based sample; specifically with respect to intellectual and educational skills.
  4. Special Nursery placement was also found to be effective for improving adaptive behaviour and educational skills.
  5. There was no clear relationship between temporal input and gains, and a programme of around 15 – 20 hours appears to be optimal.
  6. Programmes should not be intense when the parental stress is high, as the latter will counteract the effectiveness of the programme.  Programmes could be offered along with some form of initial parental training/counselling.
  7. The importance of educational skills versus adaptive behavioural skills at pre-school for subsequent school functioning and school inclusion needs to be investigated.


Helping people with autism to function independently not only promotes their quality of life, but also does much to relieve the enormous psychological burden upon their families, and the financial strain upon the many external supporting agencies, such as educational, psychological, and health services. 

Research indicates that any intervention designed to target such individuals is far more effective if offered early, rather than remedially later.  Such early teaching interventions promote inclusion in education and society, enhance the future prospects of the individuals, and are now a United Kingdom government priority (Department of Health, 1998).  However, there is a clear need to evaluate approaches designed to help such pre‑school children, in terms of their impact on these disorders, so as to calibrate the likelihood of success of such interventions.

A number of early intervention procedures have been suggested as offering benefit to some autistic children.  Special Nursery Provision and Portage have been offered as approaches to managing the problems associated with autistic spectrum problems.  Both of these approaches have the benefit of being reasonably cost-effective, but both suffer from the problem of having a scant evidence-base with respect to their effectiveness promoting the child’s intellectual, educational and social functioning.

Much current debate has centred on Applied Behaviour Analysis (ABA) techniques, and in particular the 'Lovaas' model (Lovaas, 1987).  This model is outlined in a variety of sources (Lovaas, 1981; Lovaas & Smith, 1989), and involves 1:1 teaching of children with autism by adult tutors.  The approach uses a discrete-trial reinforcement-based method.  The intervention was initially developed for 40 hours a week, for three years.  The initial results reported by Lovaas (1987) concerning the effectiveness of this approach were remarkable.  The children undergoing this approach made mean gains of 30 IQ points, and just under half of these children appeared to ‘recover’, that is, they were not noticeably different from normal functioning children after three years.

However, there have been a number of critiques of this piece of research, many of which have focused on problems both with the internal and external validity of the Lovaas (1987) study (e.g., Conner, 1998; Gresham & MacMillan, 1997).  In terms of the internal validity of the study, it should be noted that different IQ tests were often used at baseline and at follow up.  This practice may well reduce the reliability of the measurement (Magiati & Howlin, 2001).  The group selection of the Lovaas (1987) study was not random, and, more importantly, the experimental and control groups differed on a number of salient features that confound the study.  In particular, there were a higher number of girls in the control group than in the experimental group, who would have a worse chance of recovery due to the generally greater level of autistic severity of females (Boyd, 1998).  In terms of the threats to the external validity of the Lovaas (1987) study, the reliance on IQ as a sole measure may be questioned, given that IQ is not necessarily the main problem in autistic functioning.  In fact, the picture with respect to the influence of ABA on other behaviours is actually quite mixed (Reed, 2004).  Secondly, the sample chosen for the study reported by Lovaas (1987) were verbal, relatively high-functioning, participants, who may have performed equally well with any intervention of a reasonable input (such as some of those noted above).  In the Lovaas (1987) study the control groups actually received a shorter period of the same treatment, and failed to show these IQ gains.  Finally, the study was clinic-based study, and may not generalise to applications as they are typically used in the parents’ home, severely compromising the usefulness of the study, and forming a main departure for the current study of community-based approaches.

Despite these criticisms above, it should be noted that there have been a number of further studies of ABA, which have attempted to address some of these issues.  These studies have answered some of the above points.  For example, Smith, Eikeseth, Klevstrand and Lovaas (1997) studied more severely impaired children and noted only marginal IQ gains.  Smith, Annette, and Wynn (2000) found that a community-based treatment, led by therapists fared well, relative to a parent-led approach.  The range of effect sizes in some of these studies of ABA is displayed in Table 1.  However, a major issue is that in none of these intervention studies was the ABA treatment compared to another form of intervention, which is taken as a strong control condition (Hohmann, & Shear, 2002). 


Table 1:  Summary of IQ gains and effect sizes in studies of discrete-trial ABA approaches compared to control group and pre-intervention measures of the ABA children.



Comparison with control

Comparison with pre-intervention scores

IQ Gain

Effect Size

IQ Gain

Effect Size

Eikeseth et al. (1999)





Lovaas (1987)





Luiselli et al. (2000)





McEachlin et al. (1993)





Salt et al. (2002)





Shallow & Graupner (1999)





Sheinkof & Seigal (1998)





Smith et al. (1997)





Smith et al. (2000)






Thus, there are a number of issues that are clearly unresolved concerning the effectiveness of this form of intervention.  It appears critical to be able to assess the effectiveness of the ABA approach in a community-based setting with participants more typically of those who present to local education authorities.  It also appears important to utilise a wide range of instruments in the assessment procedure; not only examining intellectual functioning, but also educational functioning, adaptive behaviour, and the effects of the intervention on family stress.  Finally, providing evidence on the effectiveness of other interventions would not only allow these interventions to be assessed, but also would allow a well-matched alternate-treatment control group for the ABA studies, a comparison so far missing.  A range of interventions may also allow the importance of a number of the elements of the interventions to be teased apart; such as the temporal input, and nature of the intervention.

Given the above, the current study compared directly the impact of existing ABA, Special Nursery Placements, Portage and PACTS programmes on a variety of aspects of the children's abilities. 





            Participants were selected on the basis of four criteria: they were approximately three to five years old, they were at the start of their intervention, they received no other major intervention during the period of the assessment, and they had a diagnosis of an autistic spectrum problem.  A total of 66 participants were sampled.  Of these, five were excluded from the study (three for compromised treatment integrity, one for missing data, and one for highly discrepant GARS scores at baseline, and ceiling effects on follow up).

The 61 participants who completed the study are described in Table 2.  Inspection of these data shows that the participants in each of the four groups were well matched; the participants’ ages, and their autistic severity, were highly similar across the four groups, as were the group-mean scores on the overall measures at baseline.


Table 2:  Baseline measures for participants.  Numbers of participants, and their age at baseline (along with the range).  Means and standard deviations for Psychoeducational Profile (PEP-R), British Ability Scale (BAS), and Vineland Adaptive Behavior, measures (all standard scores, mean = 100, standard deviation = 15).







Participants (gender)

12  (11m,  1f)

20 (18m, 2f)

16 (14m, 2f)

13 (12m, 1f)

Mean Age (months)

Age Range (months)


32 – 47


41 - 48


30 – 45


37 – 46

Autistic Severity:

GARS Autism Quotient

90.5 (14.0)

98.1 (9.7)

85.5 (25.1)

95.6 (11.6)

Intellectual Functioning

PEP-R: Overall Score

55.6 (13.8)

51.9 (20.1)

53.3 (16.1)

49.4 (13.2)

Educational Functioning

BAS: Cognitive Ability

56.8 (16.6)

57.8 (12.8)

53.1 (10.9)

52.5 (10.0)

Adaptive Behaviour:

Vineland Composite

58.2 (8.0)

53.0 (4.6)

59.0 (6.0)

56.2 (4.2)


            Assignment to group was on the basis of the intervention being offered to the child in their particular area.  For example, if a child was in an area that offered a special nursery placement, then that child was assigned to that group.  The areas involved in the study offered a similar socio-economic profile, all being in South-East England.  Thus, although the allocation to group was not truly random, the child’s characteristics did not influence group assignment.  This is seen in the well matched profile of the four groups.  Ethical approval for the study (University College London Hospital Trust Ethics Committee) was granted on this understanding.



            Four community-based early interventions for autism were studied (i.e. they were not part of a specially organised trial); Applied Behaviour Analysis (ABA), Special Nursery Placements (Nursery), Portage, and Parents of Autistic Children Training and Support (PACTS).  These interventions were selected as they represent some of the most commonly occurring community-based interventions, and they provided a broad spectrum of the types of approach currently on offer in the United Kingdom.

            Applied Behaviour Analysis.  The Applied Behaviour Analysis (ABA) programmes included in this study were provided by a range of organisations, who offered: discrete-trial ‘Lovaas-type’ interventions (3; see Lovaas, 1987), discrete-trial “Verbal Behaviour” programmes (4; see Sundberg & Michael, 2001), and CABAS-based approaches (5; see Greer, 1997).  Although nominally distinct in their theoretical orientation, all of these programmes shared key ABA-features.  All were home-based, and offered mostly 1:1 teaching for the autistic child.  The teaching was provided by a number of tutors in each programme under the guidance of an ABA Supervisor.  Typically, a session would last two-three hours, and comprise approximately 8 – 14 tasks or drills per session, depending upon the particular needs of the child.  These tasks would last typically about 5 – 10 minutes each, and would be repeated until some criterion performance was reached.  Each task would be separated by a 5 – 10 minute break, or down-time.  The programmes used an antecedent (question/task), behaviour (response) sometimes prompted if necessary, and consequence (reinforcement, usually a tangible such as food, but also praise and activities) procedure.  No aversive stimuli were used in any of the programmes.

            Special Nursery Placements.  The Special Nursery Placements (Nursery) occurred in a variety of provisions.  All were in Special Educational Needs Nurseries, but some of these were specialist autistic nurseries (8), and the rest were in special needs nurseries catering for all types of disabilities, including autistic spectrum problems (10).  Typically, the children would attend the nursery for a number of 2 – 3 hour sessions per week, depending on the severity of the child’s autism.  The nursery would engage in a range of teaching activities, mostly group-based (e.g., play-based activities, use of some social stories).  Many of the placements used visual scripting techniques (such as TEACCH).  Although mainly group-based, there were also some 1:1 tuition sessions, and most children had a dedicated support worker for some part of their nursery session.

            Portage.  Portage is a home-based teaching programme for preschool children with special educational needs (see Cameron, 1997).  The children are taught new skills through the use of questions and tasks, prompts, and rewards.  Parents and carers are shown how to apply this system by a weekly/fortnightly visit from a Portage supervisor.  The training sessions are brief, usually about 15 – 20 minutes per day, and are scheduled when the parent believes the child will be at their most receptive.  Typically, the parent will teach the child in a 1:1 situation, and will target one or two skills a week for teaching.  Monitoring and evaluation of progress occurs at the supervisors visits.  The Portage programme has been extensively used with children with developmental delay and is typically not intensive.  Most children in the current study received this form of intervention (12), however, some children (4) received a more intensive version of the Portage procedure that involved greater temporal input per day (around 2 hours in some cases).

            Parents of Autistic Children Training and Support.  This programme has been developed by Bexley Local Education Authority.  This is a home-based programme for pre-school children between 2:6 and 4:0 years old.  The programme has five distinct parts: an introductory training course for parents and carers which is intensive and takes five days.  There are home-based supervision and support sessions from a supervising psychologist.  Up to four home-based sessions each week of direct 1:1 teaching for the child are carried out by trained assistants.  There is also regular progress monitoring using checklists, observations, and video recording.  Finally, the aim of the scheme is a planed and supported transfer into school, usually a nursery/reception class.  The sessions are typically discrete-trial, reinforcement-based, and focus on social cooperation, communication, self-help, basic skills, and play.  The important facet of this scheme is it combines parental training and home-based intervention.

            The key characteristics of the different interventions, along with a description of their main features are shown in Table 3.


Table 3: Characteristics of the four interventions studied.  The mean number of hours input, type of teaching (individual versus group), mean number of tutors and family members involved in the programme, number of different service providers are all shown.







Mean intervention (hrs)

Range (hrs)

Interquartile range (hrs)


20 - 40

28 – 34


3 – 23

12 -15


2 – 15

3 – 9


11 – 20

12 – 13

1:1 teaching (hrs)





Group teaching (hrs)





Tutors (family tutors)

4.4 (1.0)

4.0 (1.0)

4.0 (2.0)

4.7  (1.4)

No. of Service Providers





Treatment  characteristics:     Based:


Led by:



Parent Training Course:
































            A wide range of measures were employed in this study to cover three broad areas (full details of these measures are given in Appendix 1).

Intellectual Functioning

                     The Psychoeducational Profile (Revised).

Educational Functioning

                     British Abilities Scale: Early Years - early number, naming vocabulary, verbal comprehension.


                     Vineland Adaptive Behaviour Scale.

                     Conner's Teacher and Parent Rating Scales (Short Form).

Family measures

                     Stress – Questionnaire on Resources and Stress (Friedrich version)

                     Style ‑ Parent‑Child Relationship Inventory

                     Support ‑ Huntingdon Client Feedback Questionnaire



The children were identified in conjunction with the LEA.  When identified, the children were visited by an Educational Psychologist, and the first set of measures taken (this 'assessment took about 120-180 minutes to complete).  The family of the child were also contacted at this point, and the purpose of the project explained.  A brief history of the child's provision to date (if any), was taken, and the family were asked if they would mind completing some measures on how the problems have impacted upon them.  After nine months, the final child-measures were taken by the Educational psychologist, and the family were asked to fill in the questionnaires again.




            On re-assessment, the GARS measures were retaken, and the mean change from baseline across all participants was a statistically insignificant increase of 4.21 in the overall autism quotient.  None of the groups diverged from this pattern in the overall assessment of the change in their autistic severity.


Overall functioning measures

The change in the intellectual functioning (PEP-R), educational functioning (BAS-GCA), and adaptive behaviour scores (Vineland overall scores), was assessed by taking a follow-up minus baseline difference score.  These change scores are shown in Figure 1.  It can be seen that the ABA group scored higher than the other groups on the intellectual functioning (PEP-R) and educational functioning (BAS) measures, while both ABA and Nursery outperformed the other two groups on the adaptive behaviour (Vineland composite) score.





Figure 1: Change in standard scores (follow up minus baseline) over the nine month assessment period for the four groups.


These data were analysed by a multivariate analysis of covariance (MANCOVA) with the change in PEP-R, BAS and Vineland scores as dependent variables, the four intervention types as independent variables, and the participants’ age at intake and their initial GARS scores as covariates.  This means that any statistically significant results obtained for the interventions are independent of any differences in the initial age of the participant, or the severity of their autism.  As the numbers of participants per group were relatively low (if acceptable) for this form of analysis, Pillai’s Trace Criterion was chosen as the test statistic, as this is the most robust test statistic option (Olson, 1979).  This analysis revealed a statistically significant effect of intervention, Pillai’s Trace Criterion = 0.297, F(9,165) = 2.01, p < 0.05.  Separate univariate analyses of variance (ANOVAs) revealed statistically significant differences between the interventions on educational (BAS), F(3,55) = 3.81, p < 0.05, and adaptive behaviour (Vineland), F(3,55) = 2.89, p < 0.05, but not on intellectual functioning (PEP-R), F(3,55) = 2.19, 0.10 > p > 0.09, measures.  Further follow-up tests (Tukey’s Honestly Significant Difference (HSD) tests) revealed that for the educational functioning (BAS), ABA had a statistically significantly higher score than all the other interventions, all ps < 0.05.  For the adaptive behaviour scores (Vineland) both ABA and Nursery each differed significantly from both the Portage and PACTS interventions, ps < 0.05.

The extent to which each individual intervention produced statistically significant changes over the intervention was also analysed by means of paired t-tests testing the statistical significance of the change over and above zero over the assessment period.  These results are shown in Table 4.  Inspection of these analyses show that for intellectual functioning (PEP-R) both ABA and Nursery produced statistically significant improvements.  All groups produced statistically significant improvements in educational functioning (BAS).  Finally, for adaptive behaviour, only Nursery produced a significant improvement over the assessment.


Table 4:  Results of paired t-tests against zero for all four groups for each change score (* = p < 0.05, ** = p < 0.01).







Intellectual Functioning (PEP)





Educational Functioning






Adaptive Behaviour (Vineland)







Impact on Family Stress

            The effect of the intervention on family stress over the assessment period was examined by calculating a change score for the total family stress index derived from the QRS-F scale.  These data can be seen in Figure 2 and shows that stress reduced in all groups except for the Portage group, with the reductions being the largest in the ABA group.


Figure 2.  Change in total family stress for the four groups


            These data were analysed by a one-way ANOVA with group as the independent variable, and a statistically significant difference was found, F(3,57) = 3.49, p < 0.05.  Tukey’s HSD tests revealed that this difference was due to the ABA and Portage groups differing significantly from each other, p < 0.05.  None of the other differences between the groups proved to be significant, all ps > 0.10.  In addition to analysing group differences, these data were analysed to examine if each group produced a statistically significant reduction in stress over the course of the study.  All of the groups except Portage, t < 1, showed a significant reduction in stress: ABA, t(11) = 6.81, p < 0.01; Nursery t(19) = 2.63, p < 0.05; PACTS, t(11) = 2.59, p < 0.05


Moderating Factors

            The above analysis shows that there were differences in terms of the interventions on educational and adaptive behaviour measures.  These differences were independent of initial severity and age at intake.  However, number of other factors might have contributed to these scores.


Time input

The most obvious is time of the intervention.  Of course, analysis of this variable is confounded by the fact that the nature of the time in each of the four interventions is made different by the differing teaching approaches occurring during the same period.  Nevertheless, an attempt was made to analysis this variable.

            The data present a multi-level model problem with each of the participants having a time and outcome score, but also being nested into for different intervention types, for which the relationship between time and outcome may well be different.  This analysis s made difficult as it is not necessarily the case that the relationship between time and outcome has the same slope for each of the interventions.  For this reason separate correlations were conducted on each of the four interventions for each of the three main outcome measures.


Table 5:  Pearson’s correlation coefficients for the relationship between time of intervention and change in measure for the four interventions (* = p < 0.05, ** = p < 0.01).







Mean Time (Hrs)

Range (Hrs)

Interquartile range (Hrs)


20 - 40

28 – 34


3 – 23

12 -15


2 – 20

3 – 9


11 – 20

12 – 13


(PEP-R: Overall)




- .488



- .252




Adaptive Behaviour

(Vineland Composite)






            Inspection of these data reveals a somewhat complex pattern of results.  For the Portage group, there were clearly moderate to strong positive correlations between increasing the level of temporal input and gains.  There was little relationship between these two variables for the Nursery group, and negative relationships between these scores for the PACTS and ABA groups.


Family Stress 

            The role of the families stress prior to the start of the intervention was also examined as a possible mediating variable.  Initial analysis of the correlation between the families’ stress at baseline, as measured by the total QRS-F score, and the various overall measures of functioning (PEP-R, BAS-GCA, and Vineland Composite) showed no direct influence (PEP: r(59) = - 0.015; BAS: r(59) = - 0.004; Vineland: r(59) = - 0.117).

However, this simple analysis proved to mask a more suitable relationship concerning the interaction of family stress levels and the amount of temporal input required by the intervention.  Performing a split on the data at the mean points for both the family stress at baseline (mean = 28), and the mean for intervention time (mean = 16 hours), created four groups of participants (irrespective of the type of intervention experienced).  Figure 3 shows the outcome gains made for these four groups on all the three main child outcome measures.


Figure 3:  Change in standard scores (follow up minus baseline) over the nine month assessment period for the four groups.

            Inspection of these data shows that for the intellectual functioning measure (PEP-R) the high time input groups out performed the low time input groups, with little impact of family stress.  However, for the other two outcome measures, although it was generally true that high time input produced greater gains than low time input, the low stress high time group outperformed the high stress high time group.  This suggests that for the educational (BAS) and adaptive behaviour (Vineland) measures, high family stress may inhibit the gains otherwise associated with high temporal input.


Table 6: Pearson’s correlation values between intervention time input and outcome change measures for low and high stress groups (* = p < 0.05, ** = p < 0.01).



Low Stress

High Stress

Correlation Difference


(PEP-R: Overall)



High > Low






Low > High


Adaptive Behaviour

(Vineland Composite)



Low > High



            These observations were confirmed by the analysis of the pattern of correlations observed between temporal input of the intervention and outcomes, which are displayed in Table 6.  This table illustrates a different pattern for the intellectual functioning measure compared to the other two child-outcome measures.  For intellectual functioning (PEP-R), the relationship between time input and outcome is a stronger positive relationship for the high compared to the low stressed groups.  This pattern of data is reversed for the educational (BAS) and adaptive behaviour (Vineland) measures, which show a stronger positive relationship for the low stress parents compared to the high stressed group.




            The present research examined the effectiveness of a range of early interventions as they occurred in the community, rather than examining specially tailored clinic based interventions.  In this way it was hoped to give some greater insight into the relative benefits of the different interventions as they are likely to occur in educational settings.  The present work went beyond previous work by comparing range of interventions across a wide range of measures, and attempted to isolate some of the contributing factors to the success of the interventions.

            The main findings that emerged from the work reported above are that there was no sign of overall recovery from autism, in that the Autism Quotient did not dramatically improve in nay of the children.  This stands in contrast to the findings reported by Lovaas (1987) of almost 50% rates of recovery.  Of course, there are several factors which may account for this; the length of time that the intervention was assessed for shorted in this study than in Lovaas (1987).  However, other studies have assessed children for longer and similarly found no recovery (e.g., Boyd & Corley, 2001).  It would also be highly surprising if no gains after nine months would suddenly develop into full recovery with extension of the programme/  The length the temporal input was not 40 hours, but if anything the current study found increasing the temporal input on ABA, and reinforcement-based approached (e.g., PACTS) was counterproductive.  This finding is borne out by a meta-analysis of the ABA literature (Reed, 2004).

            In terms of the comparison between the interventions, the results demonstrate that in terms of intellectual functioning, as measured by the PEP-R, ABA, Nursery and PACTS all generated gains from baseline (Portage did not produce great gains on this measure).  However, these data fell short of statistical significance.  In terms of educational improvements, all of the interventions generated a gain.  However, ABA produced greater gains than the other interventions.  In terms of adaptive behaviour gains, Nursery placements produced gains, whereas ABA, Portage and PACTS did not produce such great gains.  Taken together, this suggests that both ABA and Nursery appeared somewhat more effective than the Portage or PACTS programmes overall.

These data replicate some of the findings reported in other ABA intervention studies, and extend this to a community-based sample, sing children who are relatively severely autistic for these types of study.  In addition, these are the first data to demonstrate the effectiveness of nursery placements, and stand in contrast to the often reported failures to produce gains in the nursery control groups of ABA intervention studies (see Lovaas, 1987).

            There were also findings about the possible moderators and mediators of this effect.  The impact of severity and age at intake was controlled in this study, and did not impact on the results reported here.  However, a number of other factors did vary between the interventions, and it is possible to suggest how these factors impacted on performance.  In terms of temporal input, there is no straight forward relationship between the time spent on the programme and the gains produced that can be applied across the board.  It appears that for the Portage programme, increasing the temporal input did generate greater gains for the children.  This was not the case for the Nursery placements, and the opposite effect was noted for ABA and PACTS.  These data suggest two possible factors at work in determining the relationship between temporal input and outcome.  Firstly, in general terms a relatively low levels of input (e.g., in the Portage group) increasing the input does produce increase in gains.  This relationship is less pronounced in the other groups, suggesting that there may be an optimal time for input of approximately between 10 and 20 hours a week.  A second feature is that the two more focused behavioural approaches (ABA and PACTS) showed largely negative relationships between temporal input and gains within there range, suggesting for these approaches again between 15 and 20 hours may be optimal.  This relationship was not seen in the Nursery group, which provides a strong comparison to the PACTS group in terms of temporal input.  There is reason to expect that this would be the result based on basic research that shows the effectiveness of reinforcement diminishes over the length of a training session; with longer sessions responding actually declines as the session progress, an effect attributable to habituation of the reinforcement process (Aoyama & McSweeney, 2001).  This effect suggests the importance of variety in an intervention programme session.

            However, the role of temporal input, for educational and adaptive behaviour outcomes (but not for intellectual functioning) was itself moderated by the levels of family stress.  Children whose parents exhibiting high stress at the outset of the programme did not perform as well with high temporal input on the educational and adaptive behaviour measures, as children whose parents were relatively less stressed.  This finding, taken together with the data from the ABA child outcomes, the effects of temporal input, and the differences in baseline in stress measures for the ABA parents (typically high stress), may offer a further explanation of why increasing the amount of input does not directly influence child outcomes.

            Of course, there are limitations on the possible generality of the current study.  With a study of community-based approaches it is impossible to be precise about the exact nature of the intervention which occurs during each and every session.  It may be that the current results are limited to the interventions studied in the present report.  However, the fact that significant findings were found across a range of different provisions within each intervention does offer some cause for optimism about the likelihood of generalisable finding.  There is also the issue of random allocation to groups, which was not strictly practised in this report.  Although acknowledging this problem, it should be stated that the match of children in the four groups was very close, and there is nothing in the child demographics or characteristics that either allocated them to a particular group, or would predict a better prognosis for one of the groups.  Additionally, there are ethical concerns about the random allocation of participants to interventions that they may not otherwise have received; and this act of random allocation destroys some of the external validity of the present report, from which it draws considerable strength relative to other clinic-based approaches.




Aoyama, K., & McSweeney, F.K. (2001).  Habituation may contribute to within-session decreases in responding under high-rate schedules of reinforcement.

Animal Learning and Behavior, 29, 79-91.

Boyd, R.D. (1998).  Sex as a possible source of group in equivalence in Lovaas (1987).

Journal of Autism and Developmental Disorders, 28, 211‑215..

Boyd, R., & Corley, M. (2001).  Outcome survey of intensive behavioural intervention for young children with autism in a community setting.

Autism, 5, 430-441.

Cameron, R.J. (1997).  Early intervention for young children with developmental delay: The Portage approach.

Child: Care, Health and Development, 23, 11-27.

Connor, M. (1998).  A review of behavioural early intervention programmes for children with autism.

Educational Psychology in Practice, 14, 109‑117.

Eikeseth, S., Jahr, E., & Eldevik, E. (1999).  Intensive school-based behavioral treatment for four to seven year old children with autism: A one-year follow-up.

PEACH Putting Research into Practice Conference, London.

Greer, R.D. (1997).  The Comprehensive Application of Behavior Analysis to Schooling CABAS(R)).

Behavior and Social Issues, 7, 59-63.

Gresham, F.M., & MacMillan, D.L. (1997).  Autistic recovery? An analysis and critique of the empirical evidence on the Early Intervention Project.

Behavioral Disorders 22, 185‑201.

Hohmann, A.A., & Shear, M.K. (2002). Community-based intervention research: coping with the "noise" of real life in study design.

American Journal of Psychiarty, 159, 201-207.

Lovaas, O.I. (1981).  The ME book: Teaching developmentally disabled children. Austin, TX: PRO-ED Inc.

Lovaas, O.I. (1987).  Behavioral treatment and normal educational and intellectual functioning in young autistic children.

Journal of Consulting and Clinical Psychology, 55, 3-9.

Lovaas, O.I., & Smith, T. (1989).  A comprehensive behavioral theory of autistic children: Paradigm for research and treatment. 

Journal of Behavior Therapy and Experimental Psychiatry, 20, 17‑29.

Luiselli, J.K., Cannon, .B., Ellis, J.T., & Sisson, R.W. (2000).  Home‑based behavioral interventions for young children with autism/pervasive developmental disorder: A preliminary evaluation of outcome in relation to child age and intensity of service delivery.

Autism, 4, 426‑438.

Magiati, I., & Howlin, P.A. (2001).  Monitoring the progress of preschool children with autism enrolled in early intervention programmes: Problems in cognitive assessment.

Autism, 5, 399-406.

McEachin, J.J., Smith, T., Lovaas, O.I. (1993).  Long‑term outcome for children with autism who received early intensive behavioral treatment.

American Journal on Mental Retardation, 97, 359‑372.

Olson, C.L. (1979).  Practical considerations in choosing a MANOVA test statistic: A rejoinder to Stevens.

Psychological Bulletin, 86, 1350-1352.

Reed, P. (2004).  Behavioral early interventions for autism.

Psychological Bulletin.

Salt, J., Shemilt, J.C., Sellars, V., Boyd, S., Coulsdon, T., & McCool, S. (2002).  The Scottish Centre for Autism preschool treatment programme. II: The results of a controlled treatment outcome study.

Autism, 6, 33-46.

Shallows, G.O., & Graupner, T.D. (1999).  Replicating Lovaas' treatment and findings: Preliminary results.

PEACH Putting Research into Practice Conference, London.

Sheinkopf, S.J., & Siegel, B. (1998).  Home-based behavioral treatment of young children with autism. 

Journal of Autism and Developmental Disorders, 28, 15‑23.

Smith, T. (1999).  Outcome of early intervention for children with autism.

Clinical Psychology: Science and Practice, 6, 33‑49.

Smith, T.G., Annette, D., & Wynn, J.W. (2000).  Randomized trial of intensive early intervention for children with pervasive developmental disorder.

American Journal on Mental Retardation, 105, 269‑285.

Smith, T., Eikeseth, S., Klevstrand, M., & Lovaas, O.I. (1997).  Intensive behavioural treatment for preschoolers with severe mental retardations and pervasive developmental disorder.

American Journal on Mental Retardation, 102, 238‑249.

Sundberg, M.L., & Michael, J. (2001).  The benefits of Skinner's analysis of verbal behavior for children with autism.

Behavior Modification, 25, 698-724.


This article is reproduced by kind permission of the authors.

© Phil Reed, Lisa A. Osborne & Mark Corness 2005.

Back to NAS Surrey Branch Welcome Page