Lovaas, Ph.D., August 2nd, 2000
of California, Los Angeles, Department of Psychology, 1285 Franz Hall, Box
Angeles, California 90095-1563, FAX (310) 206-6380, www.lovaas.com
Unfortunately, misunderstandings about the UCLA model have been
very common, dating back as far as 1967 when Bettelheim accused Lovaas of
"stripping the patients of (their) humanity" treating them like
Pavlovian dogs" and comparing operant conditioning procedures to lobotomy
(Bettelheim, 1967: p. 410-411). Erroneous information has continued over time
and this is a serious problem because such information hinders parents and
professionals from making appropriate decisions about services for children with
autism. It may also cause funding agencies to set inappropriate priorities for
research and treatment and may create an atmosphere of mistrust that inhibits
collaborative work on identifying effective therapies.
Reluctantly, therefore, we have concluded that it is necessary
to summarize and correct some of the more persistent misunderstandings about the
UCLA model. Detailed and peer-reviewed descriptions of the UCLA model of early
intervention for children with autism have been presented in a treatment manual
and associated videotapes (Lovaas & Leaf, 1981; Lovaas et al., 1981;
currently under revision), as well as several articles (Lovaas, 1993; Smith
& Lovaas, 1998; Smith, Donahoe, & Davis, 1999). We have also reported
research on outcomes that children receive with this intervention (Lovaas, 1987;
McEachin, Smith, & Lovaas, 1993; Smith, Eikeseth, Klevstrand, & Lovaas,
1997; Smith, Groen, & Wynn, 1999; Smith, Klevstrand, & Lovaas, 1995).
Moreover, in these publications, we have discussed the strengths and weaknesses
of the UCLA model and research pertaining to this model.
The following discussion focuses on issues regarding the
children who have participated in our research, the experimental design of this
research, the nature of the treatment that children received, assessments of the
effect of this treatment on children and their parents, conclusions drawn from
these assessments, cost of intervention and concerns about testimony in Fair
Hearings. After reviewing all the distortions which have been expressed about
the UCLA project an attempt is made to account for these (see "Why all the
Distortions" (pp. 41-43). A distinction is made between Clinic-based versus
Workshop-based services as these services generate different kinds of treatment
outcome. The paper ends with an expression of optimism about future progress in
A. Did subject selection result in a representative sample of
(1) All children were independently diagnosed with autism by
M.D. or Ph.D. licensed clinicians, and there was high agreement on the diagnosis
between the independent examiners (see Lovaas, 1987). This provided evidence
that subjects met criteria for the diagnosis of autism. Altogether, twenty
pre-treatment variables considered to be descriptive of autism and/or related to
outcome were assessed. These variables are defined in detail in Lovaas (1987)
and include chronological age (CA) at diagnosis, mental age (MA) and CA at
beginning of treatment, diagnostic classification by M.D. or Ph.D. licensed
clinicians, and videotaped recordings scored by independent observers in a
reliable manner for kind and amount of toy play, self-stimulatory behavior, and
recognizable words. From parent interviews, data were obtained on the presence
of peer play, aggression toward family and/or self, suspected sensory deficits,
emotional attachment, toilet training, sex, siblings/family composition,
socioeconomic status of family, and age of walking. From medical records, the
presence of denotable neurological signs as determined by EEG and/or CAT scans
(2) Further evidence for the comparability of Lovaas (1987)
subjects to those diagnosed as autistic in other studies comes from Control
Group II incorporated into the Lovaas (1987) study. Control Group II subjects
were evaluated by another research team (the Neuropsychiatric Institute at
UCLA). Subjects in this group were selected on the basis of IQ scores similar to
the experimental group and Control Group I at intake according to the same
measures of intelligence used by the Young Autism Project (see Lovaas, 1987 for
a more thorough description of Control Group II). Control Group II showed
similar outcome data to those found from Control Group I and to reports of
outcome from other investigators (Freeman,et al. 1985; Lord & Schopler,
1989; Lotter, 1978; Wing, 1989).
(3) Schopler, Short, & Mesibov (1989) wrote that the Lovaas
(1987) intake criteria resulted in a restricted and biased sample of
high-functioning subjects who had a favorable prognosis regardless of treatment.
That is, they argued that the sample of autistic children used in the Lovaas
(1987) study was not representative of young children with autism. In addition,
Schopler & Mesibov (1988) claimed that the Lovaas (1987) criteria for
selecting subjects would have excluded 57% of their referrals.
A printout of intake data we received from Dr. Short shows that
Schopler et al.’s (1989) own partipcants had IQ scores quite comparable to the
Lovaas (1987) participants (i.e., mean IQ of 60 in Lovaas, 1987 versus 57 in
Lord & Schopler, 1989). However, the TEACCH sample had a preponderance of
children with a chronological age of 40-45 months. This distribution is atypical
of the general population of children with autism (which should be evenly spread
over the age ranges rather than concentrated in the oldest range). Also, only 5%
of the TEACCH sample spoke in recognizable words—a considerably lower rate
than the generally accepted figure of 50% (observed at UCLA and numerous other
sites). It appears to us that Schopler & Mesibov (1988) have made errors in
their calculations and/or the TEACCH sample is atypical of young autistic
children, not the UCLA sample (Smith, 1994). Note also that the
higher-functioning subjects in the Lord & Schopler (1989) study showed the
least improvement over time (rather than the most, as the authors hypothesized
in connection with the Lovaas, 1987 study). In the only other large-scale
investigation of IQ scores in preschool children, Freeman et al. (1985) reported
very similar results (i.e., higher functioning children with autism showed the
B. Did subject assignment result in equivalent groups?
(1) Assignments to the experimental or control group were made
on the basis of therapist availability which was determined prior to
family contact with the clinic. That is, if it was decided that enough
therapists were available to provide intensive treatment, then the next family
who contacted the clinic would be enrolled in the experimental group. If there
were not enough therapists available, the family would be assigned to Control
Group I. In the project’s initial proposal to the National Institute of Mental
Health (NIMH), a matched pair assignment procedure was planned (MH 11440,
"Experimental Studies in Childhood Schizophrenia," submitted 8/25/72,
p. 21). The use of matched pair random assignment was presented to parents in
Santa Barbara by Dr. Robert Koegel who participated in the early stages of the
Young Autism Project. However, parents threatened a protest at the next meeting
of the National Society for Autistic Children (the forerunner of the Autism
Society of America) in San Diego. NIMH was contacted about this situation and
Lovaas received approval to assign children to groups based on therapist
availability, which is a generally accepted procedure in clinical research
(Kazdin, 1980) and one that has been employed in several highly regarded studies
on children with autism (e.g., Bartak & Rutter, 1973; Howlin & Rutter,
1987). All subjects remained in the groups to which they were assigned at
intake. Only two subjects dropped out, and these subjects were not replaced.
Therefore, the original composition of the groups was essentially preserved.
(2) The experimental group and Control Group I received the same
assessment battery at intake. The groups did not differ on nineteen of the
twenty pre-treatment variables; the twentieth variable, CA at onset of
treatment, was not related to outcome.
(3) The number of control group subjects who were predicted to
achieve normal functioning had they received intensive treatment was
approximately equal to the number of experimental subjects who actually did
achieve normal functioning with intensive treatment. This prediction was based
on a multiple regression equation combining the twenty pre-treatment variables.
Thus, the subject assignment procedure yielded groups that were comparable to
each other prior to treatment on factors that predicted outcome, supporting the
inference that the assignment had produced equivalent groups.
(4) Control Group I, which received up to ten hours per week of
one-on-one behavioral treatment, did not differ at post-treatment from Control
Group II, which received no treatment from the UCLA project. Both groups
achieved substantially less favorable outcomes than the experimental group.
Because the three groups (experimental, Control Group I, and Control Group II)
were similar to one another at pre-treatment, this result confirms that our
subjects had problems that responded only to intensive treatment rather than
problems such as being unrepresentative of other children with autism, or
noncompliant and holding back, masking an underlying, essentially normal level
of functioning that would enable them to respond to small-scale interventions.
With regard to the Lovaas (1987) study, Baer (1993) described
the subject assignment procedure as "functionally random." It is also
important to note that in 1981, NIMH approved of a grant request which formed
the basis for the McEachin et al. (1993) follow-up study (MH 11440-15
"Experimental Studies in Child Schizophrenia"). This grant was
approved after an extensive review and site visit by NIMH pertaining to the
experimental design employed in the Lovaas (1987) study. Kazdin (1980), in
discussing random assignment in clinical research, states, "The essential
feature of random assignment is allocating subjects to groups in such a way that
the probability of each subject appearing in any of the experimental groups is
equal" (p.125). He also states, "The usual way to ensure that subjects
are assigned randomly is to determine the groups to which subjects will be
assigned prior to their arrival to the experiment" (p.125). He then
warns that matched pair random assignment, when used as a "method of
allocating subjects to groups[,] can produce groups that differ on all sorts of
measures" (p.126). Kazdin’s warning indicates that a matched pair random
assignment procedure (as initially proposed by the UCLA project) could have
generated unequal groups, especially when sample size is small and the variables
on which subjects are to be matched are not proven to be related to outcome.
Kazdin stresses the importance of "reassuring whether there are no
differences on measures prior to treatment because this situation partially
corroborates the assumption that random assignment has distributed subject
characteristics equally across groups" (p.127). The Lovaas (1987) study
appears to satisfy this condition. The Surgeon General has also provided a
favorable review: "A well-designed study of a psychosocial intervention was
carried out by Lovaas and colleagues (Lovaas, 1987; McEachin et al.,
1993)."(U.S. Department of Health and Human Services, 1999.) In short,
there is strong evidence that the superior functioning of the experimental group
after treatment was a result of the treatment itself rather than a biased
procedure for selecting and assigning subjects to the experimental group.
C. Did family characteristics effect outcome?
Children’s families ranged from high to low socioeconomic
status, with SES means in the two groups almost identical to the national
average (Lovaas, 1987). The correlation between SES and treatment outcome was
r(18)=-.13, indicating a insignificant trend for lower SES families to achieve
more favorable outcomes. Number of siblings was not predictive of outcome. Thus,
although the Lovaas (1987) treatment required family participation, a diverse
group of families was apparently able to meet this requirement.
D. Are follow-up data believable; are treatment effects durable?
(1) In the McEachin et al. (1993) follow-up study, a wide range
of measures was administered in order to avoid over-reliance on intelligence
tests which have limitations if used in isolation, such as bias resulting from
teaching the test, selecting a test that would yield especially favorable
results, failing to assess other aspects of cognitive functioning such as social
competence or school performance, and so on (Spitz, 1986; Zigler & Trickett,
(2) The follow-up evaluation used well-normed, standardized
tests and double-blind examination procedures (selection of examiners by
professionals not associated with the project who were blind to the UCLA clinic
as the source of the request, and the use of a normal comparison subjects to
prevent bias that might occur if examiners only tested subjects with obvious
signs of pathology). Such evaluations allowed for an objective, detailed, and
quantifiable assessment of treatment effectiveness. A particularly rigorous
assessment was given to the best-outcome subjects who demonstrated normal IQ and
school placement at age seven. These assessments included the Wechsler
Intelligence Scale for Children-Revised, the Personality Inventory for Children,
the Vineland Adaptive Behavior Scales, and a structured clinical interview
designed to identify subtle signs of autism. These assessments comprised more
than twenty five subscales with the scores indicating that the best outcome
group was indistinguishable from normal controls. The results of the McEachin et
al. (1993) follow-up, which extended several years beyond discharge from
treatment for most subjects, are an encouraging sign that these treatment gains
were maintained for an extended period of time.
Recently, five well-known commentators with diverse theoretical
orientations unanimously agreed that the Lovaas (1987) study and the later
McEachin et al. (1993) follow-up presented compelling evidence that the
experimental clients improved and that the improvement was due to treatment
rather than some extraneous variable ("Commentaries on McEachin, Smith,
& Lovaas," Baer, 1993; Foxx, 1993; Kazdin, 1993; Mesibov, 1993; Mundy,
1993). For this reason, further follow-ups of these subjects appear merited and
the UCLA clinic was recently awarded a grant from NIMH to conduct such a
follow-up. This follow-up study is entitled "Long-Term Outcome of Early
Intervention for Autism," 1 R01 MH51156-01A1 (the grant proposal for this
study received the highest priority rating NIMH can provide, namely 100). The
assessment instruments employed in this follow-up are extensive, reflecting
recent advances in detecting residual signs of autism. These assessments include
the WAIS-R, Wisconsin Card Sorting Test, Stanford-Binet Absurdities Subtest,
MMPI-2, Rorschach, Socio-Emotional Functioning, Theory of Mind, Narrative
Discourse Tasks, Social Discourse, and Abstract Semantics. The importance of
such an investigation is also magnified by the paucity of available information
on the long-term effects of treatment for clients with autism in any treatment
program (cf. Smith, 1993).
The safeguards discussed above provide considerable assurance
that the favorable outcome of the experimental subjects can be attributed to the
treatment they received rather than to extraneous factors such as improvement
that would have occurred regardless of treatment, biased procedures for
selecting subjects or assigning them to groups, narrow or inappropriate
assessment batteries, biased examiners, and so forth. For more detailed
discussions of these concerns, see Lovaas (1987), Lovaas, Smith & McEachin
(1989), and McEachin et al. (1993).
E. Is the treatment replicable?
McEachin et al. (1993) emphasized that "the most important
void for research at this time is replication by independent investigators"
(p. 370), a caution similar to that expressed in Lovaas (1987). Some writers
(e.g., Foxx, 1993) have repeated the concerns about the need to replicate the
Lovaas (1987) study. There are four potential problems pertaining to
(1) It has been claimed that the treatment procedures employed
in the Lovaas (1987) study were not adequately specified, and thus not
replicable. Answer: The following considerations suggest that the
treatment procedures are replicable: First, the components of the treatment were
published in a large number of peer-reviewed journals, developed by multiple
investigators, and often replicated across sites (cf. Newsom & Rincover,
1989; Schreibman, 1988). Second, a 250 page manual (Lovaas et al. 1981) and
associated videotapes (Lovaas & Leaf, 1981) describe the treatment in
extensive detail. Third, the project has shown that it is possible to train
therapists to provide high-quality treatment of the kind described in the Lovaas
(1987) study. Samples of the one-on-one treatment provided by therapists at
replication sites have been scored by project personnel at Washington State
University for quality control Level I of treatment involving the appropriate
use of instructions, prompts, and reinforcement (using definitions and scoring
systems provided by Koegel, Russo, & Rincover, 1977). The correct use of
these procedures ranged from 92% to 100%. Interrater reliability ranged from 94%
to 100% (see R01MH48863-01A1 "Multi-Site Young Autism Project"). To
reach quality control Level II on high-level programming requires basic
coursework in learning-theory and/or applied behavior analysis and has to be
followed by a nine-month full-time internship at UCLA or one of the replication
sites. For further details describing therapist training and supervision, see
"Two Kinds of Treatment Services" at end of manuscript.
(2) It can be claimed that contingent aversives were employed in
the Lovaas (1987) study, but that many states now prohibit the use of aversives,
and many or most parents may object to such treatment. Either of these factors
would prevent replication of the Lovaas (1987) treatment. However, while the
UCLA Clinic no longer use aversives, we have taken advantage of alternatives to
aversives developed during and after the time the 1987 study was completed
(treatment in the Lovaas, 1987 study took place between 1970-1984).
(3) It may be argued that most agencies are not able to provide
the required amount of treatment. Answer: Several collaborating sites are
involved in the collaboration. These include the New Jersey Institute for Early
Intervention (Cherry Hill, N.J.); Wisconsin Early Autism Project (Madison, WI);
Community Services for Autistic Adults and Children (Rockville, MD); Young
Autism Project of Pittsburgh, Southwood Community Services, Inc. (Pittsburgh,
PA); Iceland Young Autism Project (Kipavogure, Iceland); Early Autism Project,
Akershus College (Sandvika, Norway); Central Valley Autism Project (Modesto,
CA); Meredith Autism Project (Raleigh, NC); Early Autism Project, University of
Houston (Houston, TX); Early Autism Project (Edina, MN); Early Autism Project
(Spartanburg, SC); Early Autism Project (London, United Kingdom); Early Autism
Project (Sophia University, Tokyo, Japan) and Early Autism Project, Ramon Llull
University (Barcelona, Spain) are examples of sites which have been able to
secure staff who have met quality control standards so as to deliver the
intensive treatment as prescribed in the Lovaas (1987) study as well as testing
the effectiveness of different levels of intensity.
(4) It has been argued that the Lovaas (1987) treatment may not
be replicated since student-therapists were unequally distributed across
subjects and some may have developed particularly close relationships with
parents, adding unknown variables to the treatment. In fact, the ‘87 study
took steps to ensure that such variables did not come into play. To help reduce
these factors, each student-therapist worked six hours per week for six to nine
months at which point the practicum course was completed. During the six to nine
months, student-therapists were rotated across families to provide for a richer
exposure to children with autism. In addition, student-therapists were
instructed to behave as professional persons in a manner consistent with course
outlines (no visits to home outside assigned therapy hours, no meals with
family, no visits after course completion, no receipt of gifts, and so on). In
the Young Autism Project, student-therapists appointed as supervisors after nine
months may stay on the project for upward of six years, usually as graduate
students, but they are rotated and distributed equally across families.
F. Do treatment outcomes from TEACCH and the Neuropsychiatric
Institute (UCLA-NPI) or other centers equal the outcome from the Young Autism
Project in the Department of Psychology at UCLA (UCLA YAP)?
(1) Lovaas (1987) compared children in his intensive treatment
group with two other groups: (a) a group that received minimal treatment from
UCLA YAP and (b) a group of 21 children seen at UCLA-NPI by Freeman et al.
(1985) and never referred to Lovaas (1987). TEACCH also presented data on
children of the same age as those in the Lovaas and Freeman studies (Lord &
Schopler, 1989). The IQ scores were as follows:
||IQ at Age 2-3
||IQ at Age 7
|Intensive Tx (UCLA YAP)
|Minimal Tx (UCLA Therapy)
|UCLA NPI (Freeman et al.)
|TEACCH (Lord & Schopler)
Regarding the UCLA-NPI group, Freeman et al. noted that twelve
of the fifty three children in their study showed improvements in their IQ
classification (low, medium, or high), while five showed decreases. They further
noted that "no known special medical, behavioral, or educational
interventions" accounted for the changes in status. Regarding TEACCH, Lord
and Schopler noted that most of the improvements occurred with the lowest
functioning children (those who were initially nonverbal and had an IQ under
50). Lord and Schopler noted that no child with an initial IQ above 50 gained 20
IQ points or more. Overall, then, the intellectual functioning of most children
at UCLA-NPI and TEACCH remained stable. This is in contrast to the intellectual
functioning of children who received intensive behavioral treatment at UCLA YAP:
Their functioning increased substantially.
G. Do alternate forms of treatment provide comparable outcomes?
A prepublication paper circulated by Drs. Dawson and Osterling
and presented at a tele-conference to medical personnel in the Northwest was
brought to our attention by parents who questioned the correctness of the data
reported in their paper. The paper was also presented at a conference in Chapel
Hill. We wrote Dr. Dawson March 19,1996 to inform her of mistakes in her
reporting of the outcome data from the UCLA Project, Project TEACCH and from the
Colorado Health Sciences. Dawson had reported that 3- and 4-year old children in
the TEACCH program gained an average of 15-24 IQ points. We indicated to her
that this statistic applied to only 44 of the 142 children in the study (those
who attained IQs under 50 and were classified as nonverbal at intake) and was
not representative of the sample as a whole. The remaining 98 children showed IQ
changes of 0 to -4, and, collectively, the 142 children in the study averaged an
IQ change of +5. In a letter of April 19, 1996, Dr. Dawson informed us that she
had corrected these errors and promised to send us the final version of her
chapter when it became available. However, she continued to describe the TEACCH
data as indicating that children typically made large improvements. Because this
is not what their data show, we wrote to her on May 21, 1996, to request a
clarification. We did not receive a reply nor the promised final version from
her. However, we did see a preprint of the chapter. In it, she had mostly
corrected her description of the UCLA data, though she continued to omit the
McEachin, Smith, and Lovaas (1993) follow-up data and did not report accurately
on the number of intensively treated children with normal IQs, giving us credit
for having one more child with average IQ at follow-up than we actually had
reported. Unfortunately, the description of the TEACCH data, while different
from the earlier draft, continued to have major inaccuracies. To their credit,
Lord and Schopler did not hide the absence of changes in test scores in their
original report (e.g., their title refers to the stability of these scores).
Dawson and Osterling also reported on the claim by Rogers and
colleagues from the Colorado Health Sciences that children in that program made
significant increases in their rate of cognitive, language, and other skills.
Instead of reporting IQ scores, Rogers and colleagues reported children’s
developmental ages at follow-up, as well as the developmental ages they would
have been expected to have without treatment. However, when data pertaining to
cognitive skills are converted to ratio IQ scores, the gains did not support
significant increases but appear quite small, ranging from 3 to 9 points for the
10 reported pre-post treatment comparisons and apparently lower for the 2
unreported comparisons. Thus, substantial inaccuracies pervade Dawson and
Osterling’s review. Also, they attributed data to peer-reviewed publications
when in fact most (LEAP, Princeton and Walden) actually appeared in non-peer
Essentials of the paper were subsequently published in the 1996
NIH State-of-the Science in Autism conference (Alexander, D., Cowdry, R.W.,
Hall, Z.W., & Snow, J.B., 1996. The state of the science in autism: A view
from the National Institutes of Health). The paper was also published in a
comprehensive volume on early intervention (Guralnick, M.J. (Ed.). The
effectiveness of early intervention. Baltimore, MD: Paul Brooks Publishing
Co., 1996 and was presented at a conference at Chapel Hill (North Carolina).
The widely circulated misleading information in the Dawson and
Osterling article risks having an adverse effect on professionals and parents
seeking adequate and truthful information to help guide them in their search for
effective treatments for their children. For example, Dr. Gerald Mesibov cited
the Dawson and Osterling data in a Fair Hearing to argue against parents’
requests of using the UCLA program to help their young autistic child (Sherman
v. Pitt County).
H. Do the best-outcome subjects show residual signs of autism?
A total of ten out of the nineteen subjects in the experimental
group showed obvious signs of autism and mental retardation in the McEachin et
al. (1993) follow-up. Some investigators have expressed concern that the nine
best-outcome subjects in the experimental group may evidence residual signs of
autism that escaped observation in our previous follow-ups (e.g., Mundy, 1993).
(See also Lovaas, 1987 "...certain residual deficits may remain in the
normal functioning group...answers...will soon be forthcoming in a more
comprehensive follow-up."). As described earlier, the McEachin et al.
(1993) follow-up study used comprehensive assessment batteries that allowed for
multiple opportunities to detect signs of autism. In addition, these batteries
were administered by examiners blind to the purpose of the assessment. However,
subtle signs of autism may still have escaped attention (or may have become more
obvious as subjects grew older). Also, many assessment instruments designed to
test for residual signs of autism were not available when NIMH approved funding
for the McEachin et al. (1993) study (MH 11440-15 "Experimental Studies in
Child Schizophrenia"). The recently awarded grant (1 R01 MH51156-01A1
"Long-Term Outcome of Early Intervention For Autism") is specifically
designed to detect residual signs of autism in the best-outcome group. The
assessments involved were referred to in paragraph D.
I. Have the outcome data been distorted?
(1) In addressing the Young Autism Project at an autism conference
("Autismus Heute") in Hamburg, Germany in 1988 attended by more than
one thousand parents and professionals from throughout the world, Professor
Ritvo stated that the results of the Lovaas (1987) study were "totally not
true," and added, "I regret that a lot of that stuff comes out under
the label of UCLA and it is unfortunately disseminated widely and untrue."
Professor Ritvo withdrew these remarks, however, when confronted with a UCLA
inquiry. Neither Professors Ritvo nor his colleague Dr. Freeman have examined
the best outcome children in the Lovaas (1987) study.
(2) Sometimes the accusations have been directed to others who voiced support
for the UCLA project. Dr. Gillberg, a well-known psychiatrist from Sweden,
resigned from Dr. Rimland’s journal (International Autism Research Review)
after Dr. Rimland had written a positive review of the UCLA program. Dr.
Gillberg cited in his letter of resignation (10/20/87) that "…I would not
like to be associated with a publication pushing negative reinforcement
therapies in autism" and "…I certainly don’t agree with his
(Lovaas) conclusion that autism couldn’t be a brain problem since it can be
cured with behavior therapy." The reader will recognize Dr. Gillberg’s
mistaken inferences about our supposed use of "cure." (See Section I
for misunderstandings about "cure" and paragraph (5) below. Negative
reinforcement is involved whenever a person (e.g. a parent) helps another person
(e.g. a child) overcome stress and fear. Behaviors (e.g. the child coming to the
parent) are strengthened because the latter removes an aversive event. All of us
use negative reinforcement in our interaction with others, including Dr.
Gillberg. There exists hundreds of scientific studies on the properties of
negative reinforcement. It is regrettable that our position on an organic
etiology in autism and our reference to the plasticity of the young nervous
system and the effect of early environmental enrichment (conf. Lovaas, 1987 and
McEachin, Smith & Lovaas, 1993) was not familiar to him.
(3) In a conference sponsored by the Bancroft Foundation
(Haddonfield, New Jersey), Eric Schopler, Ph.D., who was then the director of a
statewide service agency for children with autism (Project TEACCH) and the
editor of the Journal of Autism and Developmental Disorders (JADD),
remarked that he had serious reservations about the Lovaas (1987) study because
Dr. R (the real name has been withheld), who was a former student of Dr. Lovaas,
had sought grant funding and dispensed grant moneys in a fraudulent manner. Dr.
Lovaas wrote Dr. Schopler that Dr. R had been an undergraduate student in his
classes at UCLA, one of several thousands such students over the last 30 years,
and that Dr. R had worked much more closely with two colleagues of Dr. Schopler.
Dr. Schopler did not respond to Dr. Lovaas’ corrections.
(4) An issue which surfaces periodically is the assertion that
the UCLA project claimed to "cure" autism. Gresham and MacMillan
(1997) report, "We argue in this article that the (UCLA project) is far
from providing a cure for autism. With the exception of (studies from UCLA) none
claim to have a cure (of) autism." They go on to cite Rutter (1996) to back
up their accusation that the UCLA project has claimed a "cure."
Referring to the Lovaas (1987) and McEachin, Smith, & Lovaas (1993) studies,
Rutter (1996) argues that one of the reasons for being cautious about the
acceptance of the "strong claims" of these articles are based on
"the claims of cure." Rutter argues that the claims of cure run
"counter to both clinical experience and what might be expected on the
basis of prevailing theories" (p. 270). Answer: The UCLA project has never
claimed to cure autism. We have earlier warned that "certain residual
deficits may remain in the normal functioning group" (Lovaas, 1987, p. 8).
The term "cure" implies removal of the original cause of the problem
and because the cause of autism is unknown, claiming a cure would certainly be
unjustified and unethical. In contrast, it is possible to enable a child with
autism to achieve normal functioning without finding a cure for autism, just as
it is possible for a physician to recover patients to normal functioning
without having found a cure for their illness. Hodgkin’s disease is a case in
point. It can only undermine parents’ and professionals’ confidence in the
UCLA project to imply that we have made unethical claims. Part of the problems
of how to describe the outcome may have come from the use of term
"Recovered" in Table 3 of Lovaas (1987, p. 7). The term "normal
functioning" was used and defined in the title of Lovaas (1987) Behavioral
treatment and normal educational and intellectual functioning in young autistic
children. We warned that "…questions can be asked about whether the
children truly recovered from autism" (p. 8) adding that "…certain
residual deficits may remain in the normal functioning group that…can only be
isolated on closer psychological assessment, particularly as these children grow
older" (p. 8). Reference was made to McEachin’s Ph.D. thesis (1987) which
formed the basis for McEachin, et al. (1993) follow-up. Data from this follow-up
were familiar to Lovaas at the time of submitting the 1987 report.
(5) Gresham and MacMillan liken the UCLA project to Hooked on
Phonics, which, as they point out, made false claims of treatment gains and was
successfully sued by the Federal Trade Commission. (See Gresham & MacMillan
pre-publication copy (February 14, 1997, p. 28 & 29). They provide no basis
for this thinly-veiled accusation of fraud.
J. Are the claims made of certain treatment aspects ignored?
(1) Children become dependent and do not acquire concepts.
Parents who have sought consultation about UCLA from TEACCH staff have been
warned that "trying to meld the two programs (UCLA and TEACCH) may result
in great confusion...and considerable anxiety...for the autistic child,"
and that the UCLA program "promotes dependence...focusing mainly on
compliance training, and does not help the children to understand
concepts." TEACCH further claims that "operant conditioning programs
are very effective with mentally retarded persons, but such procedures are not
as effective with autistic persons" (Mesibov, 1993). Greenspan (1992)
presented "behavioral schools of thought" as an example of a
"common unhelpful approach" that "ignores the delayed child’s
many needs" and would result in "disorder patterns to become
stereotyped and more perseverative as (the children) grow" (p. 5).
There are no data to support these claims. In contrast, there is
abundant evidence that behavioral treatment can help autistic children acquire
complex behaviors such as language and social interaction (Schreibman, 1988).
TEACCH apparently assumes that children diagnosed with autism suffer from a
life-long disability and, accordingly, has designed treatment programs which
involve the students’ enrollment in special education throughout their
educational career. Outcome data from TEACCH are consistent with this
pessimistic inference (Lord & Schopler, 1989). However, as mentioned
earlier, McEachin, Smith, & Lovaas (1993) follow-up data show major and
lasting improvements in the language, emotional, and social behavior of children
diagnosed with autism when treated with intensive behavioral intervention
(2) Investigators from UCLA-NPI and TEACCH have suggested that
while intensive behavioral treatment may raise intellectual functioning, it does
not foster independence. A direct comparison of independence as assessed by a
measure called the Vineland Adaptive Behavior Scales exists and can be used to
address this claim. The UCLA YAP data were collected on all intensively treated
children at an average age of 13 years (McEachin et al. 1993). The UCLA-NPI
(Neuropsychiatric Institute) data were collected on children originally seen at
age 2-3 when these children had reached an average age of 15 (Freeman et al.
1991). The TEACCH data were collected only on high functioning autistic children
at an average age of 14 years (Venter et al. 1992). The means at UCLA, UCLA-NPI,
and TEACCH, expressed as standard scores, were as follows:
|| UCLA YAP
|Daily Living Skills
As can be seen, intensively treated children from UCLA YAP
appear to function substantially more independently in all areas when compared
to children from UCLA-NPI or TEACCH. (Greenspan has yet to publish treatment
outcome data in a peer-reviewed journal).
(3) In his review of "Social and Behavioral
Interventions" from the prestigious NICHD conference (Autism 1995:
NIH-State-of-the-Science"), McIlvane (1996) appears to make the same
mistake as Gresham and MacMillan (1997, see below) when he states that
"flexible behavior over time are more meaningful than changes in
measurements such as IQ. Assignment to regular classes as the criterion for
successful outcome is often meaningless because it reflects local political and
legal mandates more than individual child need or status. As in other domains of
intervention research, studies are needed to determine the long-term effects of
all interventions (particularly early intervention)" (p. 148). It is
difficult to understand the purpose of making such statements when the author
was familiar with the McEachin, Smith, & Lovaas (1993) outcome data from the
(4) Gresham and MacMillan (1997) argued that the intensive
treatment group and the control group differed in terms of the
amount of attention they received, and that the children could have improved
because of therapist characteristics such as warmth and encouragement. There is
no data to support that attention and warmth by adults facilitates the
improvement of children with autism. Although Bettelheim based his treatment on
this faulty notion.
(5) These authors further imply that the increase in IQ scores
"was not due to cognitive gains, but rather due to changes in the child’s
ability to show what they already know" (p. 190). We know of no data-based
research supporting this assertion but we are familiar with many studies
refuting it (reviewed by Rutter, 1983).
(6) Gresham and MacMillan (1997) also claim that "subjects
were post-tested with different measures than they received at
pretest,…thereby making the results of these different testings
uninterpretable" (p. 189). Do Gresham and MacMillan suggest that one should
give a three year old the same tests as a thirteen year old? They refer to the
Bristol et al. (1966) NIH criteria for treatment outcome research to discredit
our research and add some concerns of their own as follows:
1. "The UCLA project is based, in part, on principles of
operant learning, however the primary teaching method relies on discrete
trial discrimination learning and compliance with simple
commands" (p. 6). Answer: Discrete trials and discrimination learning is
not separate from operant learning but an integral part of such learning.
Furthermore, compliance was not a primary teaching method in the UCLA
treatment method. These misunderstandings on the part of Gresham and MacMillan
reflect their lack of knowledge about the treatment employed.
2. Gresham and MacMillan’s states: "We do, however, fault
him (Lovaas) for failing to randomly assign subjects to treatments…."(p.
11). Answer: An account of why we refrained from using a matched pair random
assignment was provided earlier (see Section B). Matched pair random assignment
is considered an ideal method for assignment, but with a small number of
subjects (e.g. 19 in each group) the assignment could well end up with an
unequal distribution (Kazdin,1980). Evidence for random assignment in the UCLA
project is based on similarity between Control Group I and experimental group at
intake (similar scores on 19 of 20 pretreatment variables). Nevertheless, it is
possible that some variable, not yet identified as predictive of outcome, could
have been unequally distributed, a problem which faces all investigators.
If a parent of an autistic child was to be informed that their
child would be assigned to a no-treatment control group, the parents very likely
would not participate. Instead they would hire their own behavioral therapists.
Anyone attempting to use a matched pair random assignment in autism today would
not be able to do so including Gresham and MacMillan, when and if they
themselves decide to engage in treatment and treatment outcome research.
Difficulties encountered by patients in accepting random assignment to either
experimental treatment or standard procedures is a problem in medical research
as well. There are numerous alternatives, such as the use of wait-list controls,
multiple baseline designs, comparison groups, etc.
3. "We do fault him (i.e. Lovaas) for failing to employ
measures that tapped more discrete behaviors (more…than global IQ and
educational placement)" (p.11). Answer: Gresham and MacMillan were aware of
the McEachin et al. (1993) follow-up which assessed the subjects on many
dimensions relevant to autism (see Section D). McEachin’s Ph.D. thesis, which
formed the basis for the 1993 publication was also referenced in the 1987 study.
Why they chose to ignore data from these publications can perhaps only be
answered by Gresham and MacMillan.
4. "Pretest scores were "optimized" by
"reinforcing compliant behavior during testing" and "did not
describe how this compliant behavior was reinforced." By doing so they
"violated standardized test administration procedures" (p.8). Answer:
We did describe the procedure used during pre-testing, which were consistent
with those developed by Freeman (1976). The intent of Freeman’s procedure is
to obtain a more valid assessment of the child’s intellectual potential when
non-compliance is reduced. In effect, we biased outcome against us by using
Freeman’s procedure, not using that procedure is likely to have resulted in
lower intake IQ scores and a larger increase in IQ scores at follow-up.
5 "……Social class of the children was not measured
(which would) restrict the extent to which one can generalize findings" (p.
10). Answer: We did measure family characteristics [("… number of
siblings in the family (1.26 in each group), socioeconomic status of the father
(Level 49 vs. Level 54 according to 1950 Bureau of the Census standards,) boys
to girls (16:3 vs. 11:8")] Lovaas (1987) (p.6). (See also Section C).
6. ("Control 2 subjects were not treated by (the Young
Project) but were receiving some unspecified form of treatment" (p. 6).
Answer: Control group II subjects received services which are commonly available
to autistic children in the community, special education and speech therapy. The
poor outcome in this group is consistent with outcome data from other
investigations of such services.
7. "Smith and Lovaas are unwilling to concede any
limitations on which the effectiveness of (the Young Autism Project) resides…
What explains this apparent refusal to admit to a single limitation in these
investigations?" (p. 10). Answer: We have identified limitations and
cautions both in regard to outcome data and treatment integrity. Already 30
years ago Lovaas presented an invited address to parents and professionals on
"Strengths and Weaknesses of Operant Conditioning Techniques for the
Treatment of Autism" (Top Priorities for Mentally Ill Children. 2nd
Annual meeting and Conference of the National Society for Autistic Children.
June 24-27, 1970, (30-41) San Francisco). In the Lovaas 1987 report we warn that
"a number of measurement problems remain to be solved" (p.8) and
illustrate this by describing difficulty in measuring play behavior,
communicative play and IQ scores. We warned that "…the term normal
functioning raises questions about whether these children truly recovered from
autism" (p.8) "…Other cautions about the 1987 follow-up were
reviewed in Sections I to III. We also expressed caution about therapist
requirements ("…extensive theoretical and supervised practical experience
in one-on-one behavioral treatment…" (p. 8). "Many treatment
variables are left unexplored, such as effect of normal peers" (p.8).
Mention is made of the limitations "….on generalization of our data to
older autistic children" (p.8). McEachin et al. (1993) repeated the need
for replication: "The most important void for research to fill at this time
is replication by independent investigators" (p. 370) and repeated the
extensive training requirements on therapists. We then go on to describe another
void, namely to better serve the children who did not achieve normal
functioning, suggesting that "…perhaps such children require new and
different interventions that have yet to be discovered" (p. 370). Does this
sound like a "refusal to admit to a single limitation" as Gresham and
MacMillan have expressed? .
8. Greshman and MacMillan do also cite Schopler et al. (1989) to
the effect that our subjects were "…clearly not representative of all
children with autism" (p. 9). Answer: In Section A we provided data
indicating how data from Schopler’s (1989) sample was not representative and
misleading. We also presented our data (on diagnosis, child characteristics,
etc.) supporting our claim that the samples used in the Lovaas (1987) study were
in fact representative of children with autism. Gresham and MacMillan did also
fail to indicate the inclusion of Control Group II in helping insure
representativeness of the Lovaas (1987) sample. Instead, Gresham and MacMillan
(1997) states"…it is unclear why this group was included in the design
other than because it represented a conveniently available, patched up control
group" (p. 187). Answer: The use of Control Group II was explicitly
described to help insure that "….the referral process did not favor the
project cases because there were no significant differences between Control
Group I and II at intake or follow-up, even though Control Group II subjects
were referred to others by the same agency" (p. 7). Furthermore,
"…the poor outcome in the similarly constituted Control Group I and II
would seem to eliminate spontaneous recovery as a contributing factor to the
favorable outcome in the experimental group" (p. 8). Gresham and MacMillan
refer to both studies. Why they have chosen to ignore these data can perhaps
best be answered by them.
Although it may seem to be reasonable to specify whether the
samples are representative of other clients diagnosed with autism, the value of
such an undertaking depends on the usefulness of the diagnosis of autism in
research and treatment. There seems to be some consensus building that there are
many causes of autism, hence there are likely to be many and different kinds
of autism, requiring different kinds of treatment and preventive measures. This
is analogous to what is happening in cancer research where many kinds of cancers
have been identified with different causes requiring different treatments and
yielding different outcomes. In treatment research, the definition of different
kinds of autism can best be determined by identification of different kinds of
causes and treatment outcomes, that is, based on a functional analysis rather
than the traditional identification based on communality in the appearance of
the behavioral topographies. This also holds, not just in the case for the
behavioral grouping called autism, but for the separate behaviors involved in
that grouping. For example, self-injurious behavior enjoy three different kinds
of determinants, positive, negative or sensory reinforcement which varies across
clients. The treatment for each is different so that a particular kind of
treatment (e.g. contingent social isolation) will decrease one kind of self
injury, increase another, and leave a third kind of self-injury unaltered. What
is required for the effective treatment of self-injurious behaviors is a
functional analysis (a cause-effect analysis) of the particular kind of
self-injurious behavior confronting the practitioner.
Kanner’s diagnosis of these children is based on behaviors of
similar physical appearances. That is, behaviors like mutism, echolalia, social
isolation, emotional aloofness, rituals and self-injury, low scores on IQ tests,
etc. may look similar to the observer. However, as stated above, similar looking
behaviors may have different causes and therefore may not belong together, i.e.
respond in a similar way to the same intervention. What is often forgotten in
traditional autism research is the fact that the diagnosis of autism is an hypothesis.
That is, the diagnosis is based on a guess that there exist an underlying
organizing force tying the behaviors ("symptoms") together. Kanner’s
diagnosis has shown itself to be a poor hypothesis, given that, in the 55 years
after Kanner proposed the diagnosis, no cause or treatment for autism has been
found. The most effective treatment addresses the problems of these children on
a behavior by behavior basis and there is little evidence of generalization of
treatment effects across behaviors. Although searching for the cause of and
treatment for autism may seem like a noble effort, Kanner’s hypothesis may
well be misleading and delaying the discovery of effective treatments for these
children. The sooner the hypothesis is rejected, the more productive of causes
and treatments our search may be. See also Hacking (1999, p. 112-115)
description of autism (and schizophrenia) as "social constructions"
and problems associated with large-scale public acceptance of such constructs.
9. "The added benefit of the Lovaas program, given the high
cost, should be examined" (p. 12). The "precious economic resources
for education" (p. 12) is sited as a reason to be concerned about cost of
treatment. Answer: An approximation to the substantial savings involved in
implementing behavioral treatment has been provided later in this paper (Section
The treatment of children with autism has cost a great deal of
money in other ways as well. In D.B. v. Greenwich, Frank Gresham
testified for one day as an expert witness to defend Greenwich, Connecticut’s
"10 hours per week" ABA program. Expert witnesses such as Gresham are
able to charge large fees for their services. Ultimately, Gresham admitted under
oath that the District’s "per hour" presentation of ABA programs was
only 1/4 of the frequency which Dr. Gresham considered to be appropriate.
In California, Gresham testified on behalf of another district
in an effort to defeat a child seeking reimbursement for a forty hour per week
program. The district lost and the hearing officer held that a 40 hour per week
ABA program was precisely what the child needed. To date, Gresham has only
testified on behalf of school districts and apparently has never been involved
with any district which has prevailed.
In May of 1999, in T.H v. Palatine, a federal district in
Illinois, after a protracted two-year court battle, held that a school
district’s offer of services in a multiple-handicap class was
"absurd" and directed the district to pay for the child’s home-based
ABA program, as well as the family’s attorneys’ fees. Significantly, the
family had offered to settle the entire dispute at the outset for only $18,000.
Ultimately, this district was compelled to pay out more than $300,000, including
the school district’s own attorneys’ fees.
10. We are in full agreement with Gresham and MacMillan’s concerns about
families receiving services from therapists who have received less than optimal
training and less supervision in delivering early and intensive behavioral
intervention. Earlier in this paper (see Section E) we described the training
requirements for sites involved in the replication project. It is of major
concern to us to learn of providers who present themselves as competent in
providing the UCLA model and may site the UCLA follow-up studies as support for
their treatments. Except for staff at the replication sites, we know of no
providers who have been recently trained in the UCLA model and sought to be
updated. Without updating treatment quality is likely to suffer (Wolf et al.
1995). Also, few if any provide clinic-based treatment as in the Lovaas (1997)
report, but rely on workshop-based treatments where the family hires staff to be
trained in one to three days of workshops, with supervision one to three month
apart. Hardly any provide outcome data. We estimate that outcome data
documenting normal functioning from such services will be less than 10 percent.
This does not imply that the children will fail to make improvements. Behavioral
interventions (as in Applied Behavior Analysis) is based on thousands of
scientific papers in Learning and Behavior reported in peer-reviewed journals
over more than a hundred years. There are ample data supporting the
effectiveness of behavioral treatments addressing self-help skills, reduction in
self-injury, community functioning, peer-play, school adjustment, language
development, and so forth, all of which should help the child become both more
competent and self-sufficient. The problem is that, for children who fail to
reach normal functioning and fail to become integrated and learn from typical
peers, services need to continue, for example by the parents’ efforts. If not
the client will regress, showing little if any benefit from earlier treatment as
was demonstrated in the Lovaas et al. (1973) outcome study. We are also
seriously concerned about efforts (some of which are supported by federal
grants) to compare the UCLA Young Autism Project against alternate forms of
treatment, using staff who have not been trained in UCLA procedures and have yet
to demonstrate quality control on treatment. These problems are also discussed
below (in Section V).
There are many issues involved here. One important issue is
centered on the failure of alternate treatments to document
effectiveness on normed tests in peer-reviewed journals. This includes services
such as special education, speech therapy, sensory motor training, floor-time,
option method, psycho-dynamic therapy, holding therapy, gentle teaching, etc.
Parents win support in Fair Hearings, not because the behavioral treatment their
children receive answers all or most problems. Rather, behavioral treatment can
demonstrate, in an objective data-based manner, that their children have made
improvements when other services fail to do so. The law is rational, and by that
is meant that it is data-based. One fortunate consequence of Fair Hearings may
be to stimulate research on treatments for autistic children and to report, at
an early date and in peer-reviewed journals, whether such treatments and
treatment combinations are effective or not. If not effective, the treatment
should not be offered and /or warnings about their potential ineffectiveness
and/or harm should be posted as is required in medicine. If not we will continue
to mislead parents and waste children’s lives.
A major and welcomed step ahead would be for Special Education
to take advantage of what is known in behavioral treatments. Behaviorists are
delivering a form of special education, the earliest efforts being represented
in the work of Itard (1932) some 200 years ago. Itard, considered the father of
special education, gave impetus to the work of persons like Montessori,
Sullivan, Kephart and Fernald. Anyone who reads Itard’s descriptions of his
treatments will recognize their similarity to behavioral ones. Itard’s
treatments did not survive and ended up going nowhere (except perhaps in the use
of sensory integration) because Itard did not know how to objectively document
treatment outcomes. Let us not once more make that mistake.
K. Is the treatment cost-effective?
The price of treatment is often mentioned as a reason for not
funding forty hours per week of one-on-one intervention. The cost is estimated
at $60,000 per year, or $120,000 for the average of two years required for
normal functioning at which point the best-outcome children no longer require
further special services. This contrasts to the cost of more than $2,000,000
dollars of life-long protective care. Regarding expense, 77% of the children in
the UCLA-NPI study were inpatients at the hospital. The average stay at UCLA-NPI
was about a month at a cost of approximately $1000 per day. These children then
attended special education classes which have yet to document their
effectiveness. The children are expected to need life-long protective care. At
TEACCH, the costs per year are lower (estimated by Lord and Schopler in a 1994
article as $1200 per year). However, the children continued to be enrolled at
the completion of TEACCH’s study, and the expectation seems to be that they
will be enrolled throughout their lives. For a more detailed analysis of the
cost-effectiveness of behavioral intervention, see Jacobson, Mulick and Green
L. Are parents satisfied with intensive behavioral treatment?
The UCLA project has been accused of being stressful for
parents. Our data show otherwise. On a 7-point rating scale (7 being
"completely satisfied") the UCLA project received a mean rating from
parents of 6.9 for clients served since 1991. On a different survey, data from
140 parents who rated UCLA workshops on a 5-point scale across six dimensions
(from "Was therapist polite and sensitive to parent’s needs?" to
"Would you recommend a similar workshop to others?") show ratings to
be 4.9, 4.9, 4.8, 4.9, 4.5, 4.9 (4.5 on the question "How effectively do
you think you can now implement behavioral techniques?").
M. Does the UCLA project focus on stimulus control and
A recent grant funded by the Office of Education for the amount
of $750,000 to Heartland Area Education Agency (AEA) of Iowa is entitled
"Development, Testing, and Dissemination of Nonaversive Techniques for
Working with Children with Autism: Demonstration of a ‘Best Practices’ Model
for Parents and Teachers." This project is intended to improve upon the
UCLA project. The grant proposal, however, is replete with errors similar to
those described above, and it has serious methodological problems as well. What
is new is the assertion that the UCLA project does not focus on stimulus control
and functional analysis of behaviors. In reality, the UCLA project pioneered
work in these areas (e.g., Lovaas, Freitag, Gold, & Kassorla, 1965) and has
continuously presented additional empirical data published in peer-reviewed and
readily available journals for the last thirty years.
The grant notes that the UCLA project uses
"consequent-based techniques" and "ignores the behavior
management literature that emphasizes control of antecedents." The
misleading nature of these statements can be seen in the many studies from the
UCLA project in which discrete trials and "consequent techniques" are
used to enhance verbal and nonverbal imitation. The child’s acquisition of
imitation, a product of discrete trial teaching techniques, allows adults and
typical children to model behaviors for the child with autism. A model for a
behavior is one kind of antecedent stimulus. Another example is conversation,
which clearly involves antecedent stimuli (responding appropriately to
statements made by others and varying responses depending on these statements).
Antecedent stimuli are always present in discrimination learning, which is the
major learning-based intervention underlying the UCLA programs. Reinforcement
withdrawal instigates frustration which is a major antecedent of tantrums and
self-injury, and has been the subject of research by the UCLA project for a
number of years, starting with Lovaas, et al (1965). Thus, we have devoted
extensive research to antecedent stimuli. In order for the child to learn
discriminations, however, "consequent techniques" (reinforcement for
correct responses, no reinforcement for incorrect responses) must be utilized.
Without such techniques, children have no way of knowing what skills they are
supposed to acquire, and no incentive to find out. It is dismaying that the
authors fail to recognize this, yet intend to embark on a project which includes
the employment of consequent (i.e., reinforcement) techniques.
It is also stated in the grant that the UCLA project relies
"heavily on aversives and negative reinforcement." The combination of
concepts such as "aversives" and "negative reinforcement" in
this sentence is unfortunate. First of all, the UCLA project has not used
aversives for years. Secondly, all therapists, teachers and parents use negative
reinforcement when they reduce a child’s stress or discomfort contingent on
the child’s behaviors. The behavior is reinforced (strengthened) when it
results in the removal of an aversive event. For example, if a child is hurt and
seeks out the parents who then reduces the hurt, then this is likely to
strengthen closeness to the parents.
Of considerable concern is the use of federal moneys to fund
research projects based on distortions of the UCLA project. In addition to
containing distortions of the UCLA project, the proposal contains numerous
serious methodological and conceptual problems. For a review of these, see Smith
and Lovaas (1997).
N. Do discrete trial procedures result in limited generalization
Misleading statements about the use of operant, discrete trials
are common. A recent example is provided by Mesibov (1997) which states,
"the children following these techniques have been unable to generalize
what they have learned" (p.27). This is in contrast to the TEACCH
curriculum which facilitates communication ‘in many situations.’" On
the basis of the data reviewed earlier (see the Vineland data reviewed earlier
in this paper), and Mesibov’s familiarity with the McEachin, Smith, &
Lovaas (1993) follow-up study it appears that Mesibov is providing misleading
information to parents and colleagues.
O. Quality Control on Treatment.
Sheinkopf and Siegel (1998) article introduces a number of
misunderstandings about what constitutes replication of the Lovaas (1987) study.
The 1987 study was based on clinic-supervised treatment, not on workshop
based consultant services which comprise the focus of the Sheinkopf-Siegel
article. Both clinic-based and workshop based services take place in the home,
the difference being the amount of supervision and the quality of
staff-training. This difference is substantial and there are no data to suggest
that the treatment outcomes would be similar. To illustrate, clinic-based
services allows for daily supervision of treatment by senior staff and employs
one-on-one therapists who have been selected on basis of superior performance in
academic and practicum work and stayed on the project for 6 months to several
years. The intensity of the supervision and demands for quality control on
treatment has limited the maximum number of clients to 15 per site at any one
time. We have earlier warned that "it is unlikely that a therapist or
investigator could replicate our treatment program for the experimental
(intensive treatment) group without prior extensive theoretical and supervised
practical experience in one-to-one behavioral treatment....." (Lovaas,
1987, p. 8). Given these considerations, various statements in the Sheinkopf and
Siegel article can be dealt with as follows:
1. (Page 16) "Parents reported that their children had
received treatment based explicitly on the methods outlined by Lovaas and
colleagues" (Lovaas et al. 1981). Correction: As we have already
stated, reading the teaching manual (Lovaas et al. 1981), attending a workshop
led by UCLA certified consultants, practicing behavioral therapy on several
families or spending a short time at the UCLA affiliated site, does not make a
person qualified to provide UCLA based treatment. Even professional behavioral
analysts who have not completed an internship at UCLA affiliated
replication-site would also not, in our judgment, have appropriate experience to
replicate the UCLA program. In the same context, Sheinkopf and Siegel claim
(Page 18) that "We were unable to directly observe therapy and as such to
account for treatment fidelity." It is highly unlikely that neither the
parents, or Sheinkopf and Siegel, would be able to ascertain whether children
were receiving the UCLA based treatments because neither have been trained in
2. (Page 18) The parents received services from "Three
behavior therapists in the San Francisco Bay area for guidance in treatment
implementation." Correction: So far as we can determine, none of the
therapists who treated the children in the Sheinkopf and Siegel article had been
trained at UCLA with the possible exception of one who served as a senior
therapist on the UCLA program some 15 or more years ago. Consequently, we think
it is quite unlikely that they would be able to serve competently as a
consultant to families in home-based programs. Such therapists, when they are
working with us, are not permitted to conduct workshops, and we recommend that
they not do so in the future. Several have become workshop consultants and have
opened clinics on their own despite our recommendation, and this is a source of
great concern to us if they claim to provide services based on the UCLA model.
Data show that without updates and regular refresher courses, service providers
fail to maintain quality control standards on treatment (Wolf et al. 1995). We
would expect less than 10% rate of normal functioning as a result of children
receiving workshops of a unspecified nature from unspecified consultants.
3. (Page 21) "It is possible that Lovaas overestimated the
minimum number of treatment hours per week needed for therapeutic effect." Correction:
There are no data to support that an intervention of less than 40 hours per week
will result in 47% rate of normal functioning. If we had evidence that 30 hours
or less a week would generate the same outcome, then we would have been able to
serve that many more children which would be the goal of any treatment agency.
Furthermore, given the unknown and non-optimal nature of the treatment provided
in the Sheinkopf and Siegel article, we are in no position to argue that more or
less of that form of treatment would make a difference. Finally, Pomeranz (1998)
questions Sheinkopf and Siegel’s assertion that children who received 20 hours
a week of therapy gained as much as those who received 30 hours a week. Dr.
Pomeranz comments that the children who received 20 hours a week received many
more months of therapy. It is a major mistake to delete such information. There
also remains a question of the children’s treatment history, prior to the
treatment provided by the therapists in the Sheinkopf and Siegel study.
We are greatly concerned that when these children are assessed
as teenagers (as in McEachin, et al. 1993) outcome data will in all likelihood
show that the children have not maintained their treatment gains, but regressed
unless the workshop services were continued. One can protect a child from
regressing by helping the child develop friendships with typical children,
learning to play and talk with them, and learning to learn in a classroom
environment with typical children. It takes a lot of skill and numerous hours of
treatment to achieve this end, and we have no reasons to believe that the
workshop-based program by persons described in the Sheinkopf and Siegel article,
who do not adhere to the UCLA program, will be able to do so.
It is difficult to ascertain the full effect of the Sheinkopf
and Siegel article on parents and service providers. It is worth noting that Dr.
Schopler used the Sheinkopf and Siegel article to once more discredit the UCLA
project (Schopler, 1998, pp. 3-4).
P. Additional Misunderstandings
With the opening of a UCLA replication site (P.E.A.CH.) in
London, the United Kingdom, we are witnessing a flurry of misunderstandings
about the UCLA Project in newspapers and technical journals in Europe. One such
publication by Dr. Patricia Howlin (1997) appeared in the prestigious and widely
circulated journal European Child and Adolescent Psychology. The major
errors in this article are as follows:
1. (Page 60) "...the IQ of the control group children had
risen by only 8 points" (in the 1987 study). Correction: Both
control groups lost IQ points.
2. (Page 60) "McEachin, Smith, & Lovaas...reported on
the same children." Correction: Control Group II children were not
made available for assessment in that follow-up study.
3. (Page 61) "...the lack of random assignment." Correction:
Pre-treatment assessment showed the groups to be equivalent at intake which is a
test of whether random assignment was used and successful.
4. (Page 61) Findings should be questioned because of
"...the variability of measures used...before and after treatment." Correction:
Few would suggest that 3 year olds should receive the same assessment as 13 year
5. (Page 61) There are questions about the
"...representativeness...of groups involved." Correction:
Control Group II represents children from another agency; data show the children
in the UCLA study to be comparable to those in Control Group I.
6. (Page 61) "...the failure to use independent assessors
in the evaluation..." Correction: The McEachin et al. (1993)
follow-up study describes the double-blind assessment which involved independent
assessors. Dr. Howlin should be familiar with this study since it is referred to
in her 1997 article.
7. (Page 61) "By far the greatest controversy,...the use of
terms such as 'cure'..." Correction: The term "cure" has
never been used by the UCLA Project. To state that we claimed cure can only
serve to undermine others’ confidence in our research.
8. (Page 61) "...there are no measures of social
interaction, friendships, conceptual abilities, social communication,
obsessional and ritualistic behaviors..." Correction: Dr. Howlin
gives reference to the McEachin et al. (1993) follow-up study and in Table 4 (p.
67) of Howlin's article, the McEachin et al. study is described as reporting
significant improvements in social functioning and decreases in maladaptive
9. (Page 66) In Table 4, it is stated that the Rogers (1991)
study demonstrated "significant changes in cognition." Correction: In
fact, when the children’s MA scores are converted to IQ scores, the gain is
between 3 and 9 points. This is not a significant gain (Section G).
Q. How valid are the testimonies at Fair Hearings?
Many parents who have chosen the UCLA program for their children
end up in Fair Hearings to determine whether educational and other state
agencies will agree to pay for services. At such hearings, "expert
witnesses" representing the state are called in to testify for or against
the parent’s requests. The number of misleading statements provided by these
expert witnesses is large indeed. One case may be of particular interest, namely
that of Robert P. vs. HISD in Houston, Texas, in which Dr. B.J. Freeman provided
Dr. Freeman began her testimony by claiming that she was
familiar with the UCLA project. If nothing else, she was a member of John
McEachin’s Ph.D. committee, and approved of the research which formed the
basis for the McEachin, Smith, & Lovaas (1993) follow-up study. During her
testimony, Dr. Freeman repeated most or all of the misinformation provided
earlier. Specifically, she claimed that she was concerned about "subject
selection bias," referring to Schopler, Short, & Mesibov (1989). She
added that "measures of outcome were not adequate," that UCLA project
used "untrained graduate students" to examine for signs of autism, and
implied that the UCLA project failed "to follow those children beyond the
first grade," withholding her knowledge of the findings of the McEachin,
Smith, & Lovaas (1993) follow-up study. When asked if she personally
evaluated subjects from the Lovaas (1987) study, she answered, "Yes, I
have." However, she has not evaluated the best outcome children, although
she may have evaluated some of the non-best outcome children. She also claimed
that she has had patients who did as well or better than the Lovaas (1987)
children, even though her own outcome data (Freeman et al. 1985) demonstrate
that the children showed no improvement, consistent with the Lord & Schopler
(1989) data. IQ scores do not increase over time with higher functioning
children, as Dr. Freeman also claimed in her testimony, but in fact such
children lose IQ points at follow-up, as is shown in her own data (Freeman et
al. 1985). She further claimed that control group children were not matched in
the UCLA project, when in fact they were. She stated that she considers it
"absurd" that preschool activities are taught first in the home, and
asserted that forty hours a week of one-on-one treatment is
"counterproductive for children with autism." She claimed that a few
hours of one-on-one and group instruction provides equally favorable results,
but she has offered no data to support this assertion. She argued that discrete
trial learning results in "failure to generalize because it hasn’t taught
a concept." These statements are not only incorrect but also disturbing in
view of her knowledge of the McEachin et al. (1993) follow-up.
Sadly, Robert P. was denied funding and it is our opinion that
the inaccuracies in Dr. Freeman’s testimony contributed to this denial. Note
that Dr. Freeman’s distorted testimony is not unique in Fair Hearings.
[Appendix Contains Rebuttals of Widely Distributed
A. Comments on a Circular distributed by Love & Mesibov
(Circular Written by Steven R. Love, Ph.D., and Gary B.
Mesibov Ph.D., September 5, 1995. For a reprint please contact: Division of
TEACCH Administration and Research, CB# 7180, 310 Medical School, Wing E,
The University of North Carolina at Chapel Hill, Chapel Hill, North Carolina
27599-7180). Dr. Love is a Clinical Assistant Professor and Clinical
Director of the Asheville TEACCH Center; Gary Mesibov is a Professor and the
Director of TEACCH and is currently the Editor of JADD (Journal of Autism
and Developmental Disorders).
On September 5, 1995, Dr. Love and Dr. Mesibov distributed a
circular providing misleading information about the UCLA Young Autism Project.
The September 5, 1995 document is one of several circulars, letters, and
publications generated by TEACCH which repeats statements containing misleading
information. Although UCLA staff have brought these problems to the attention of
Drs. Love and Mesibov earlier, both continue to circulate misleading
information. There is also evidence that such misleading information was
presented to parents who sought reimbursement (in court) for services from the
UCLA Clinic. What follows are our answers to the misleading information in the
September 5, 1995 circular.
1. It is argued in the circular that instruction from TEACCH
leads to "greater independence and adaptability" because it focuses on
"meaningful routines and visual clarity to teach autistic people how to
function without constant adult assistance." In contrast, the UCLA program
uses "discrete trials" and with its use of prompts is "likely to
set-up dependence on the therapist and result in greater prompt dependence.
Autistic children will be less likely to use their skills in vocational and
residential programs." Answer: Drs. Mesibov and Love note that there
is empirical evidence that behavioral treatment is effective. However, no such
evidence exists for TEACCH. Both UCLA and TEACCH have published data from the
same measure of independence, namely the Vineland Adaptive Behavior Scales.
These data were reviewed earlier in this paper (Section J). Obviously, children
from the UCLA treatment program perform better than those from TEACCH on
well-normed measures of independent living.
2. In the circular, Love and Mesibov claim that "operant
paradigms are typically more effective with persons with mental retardation than
autistic people." They further argue that "many people with mental
retardation show a greater responsiveness to social reinforcement and external
motivation and their understanding of contingencies are often more highly
developed than those of people with autism." Answer: We know of no
empirical evidence to support these assumptions. The more we learn about
children with autism, the more we become aware of the abundance of external
reinforcers that can be utilized in treatment.
It is also argued that the UCLA project "may not focus as
much on what children with autism personally find motivating and
interesting." Answer: An enormous amount of effort is directed at
identifying effective reinforcers for helping the children treated develop
motivation to learn. A great deal of emphasis is placed on individual
differences as they relate to what a particular child will find reinforcing.
Without effective reinforcers, no learning will occur. The UCLA project has
amply demonstrated that the children can learn.
3. "Another difference between Lovaas and TEACCH is the use
of negative consequences of punishment." Answer: The UCLA project no
longer uses aversive interventions, and it has not done so for several years,
relying instead on what has been published in the literature about building
behaviors incompatible with self-injurious and self-stimulatory behaviors.
TEACCH, on the other hand, endorses the use of aversive interventions if other
interventions fail (Schopler, 1986). In fact, TEACCH has recommended the
following intervention (Schopler, Lansing, & Waters, 1983): "Each time
that Dave spits, put the edge of a washcloth that has been soaked in Tabasco
sauce in his mouth for a moment" (p. 227). (Tabasco sauce can cause tissue
damage). Schopler (1994) has endorsed the use of aversive therapy to help reduce
severe self-injurious behavior. A Consensus Development Conference sponsored by
the National Institutes of Health issued a preliminary report (NIH, 1990)
recommending aversives or "behavior reduction procedures" restricted
by appropriate peer review and consent, for short-term use.
Drs. Love and Mesibov also argue that TEACCH "relies
primarily on positive interactions between TEACCH and student." There is a
wealth of positive interactions between teachers and students in the UCLA
project as well. In addition, parents are involved in administrating treatment
at the UCLA project. In contrast to what others have argued, we have not found
parents to be lacking in the love they show for their children.
4. "One claim from the UCLA project which Drs. Love and
Mesibov find particularly troublesome is the suggestion that the intensive
Lovaas intervention will lead to a cure (or normal behavioral and intellectual
functioning) for almost one half of the children." Answer: As
mentioned earlier in this chapter, the UCLA project has never claimed that it
cures autism. However, there is ample objective and empirical evidence that
almost one half of the intensively treated children obtain scores falling within
the normal range of intellectual, social, emotional, educational, and self-help
skills on standardized and well-normed psychological tests (McEachin, Smith,
& Lovaas, 1993).
5. "There is not good evidence, however, that progress from
the UCLA (treatment) is any better than through other individualized, less
demanding interventions emphasizing social and communication skills nor that
long-term outcomes are any better". Answer: There is good evidence
that young children who are subjected to the TEACCH program fail to improve on
standardized psychological tests (Lord & Schopler, 1989). In contrast, there
is abundance of empirical evidence, published in APA peer-reviewed journals,
that intensive and early behavioral intervention is the treatment of choice for
children with autism. Objective long-term outcomes from other treatment
facilities (including TEACCH) are simply not available (Section G).
6. "The other troubling assumption that the Lovaas group
promotes is that more is better." Answer: There is abundant evidence
that more is better. The circular continues, "Forty hours of
intensive intervention a week makes it difficult for the parents to meet their
child’s needs to be a child. Young children with autism are still children and
need time for structured play and other less demanding activities." Answer:
There is no evidence that "structured play" and "less demanding
activities" are helpful in improving the functioning of children with
autism. No doubt parents of children with autism find time to engage their
children in play and less demanding activities, yet their autistic children do
not improve despite these efforts. In contrast, there is every reason to believe
that without early and intensive behavioral intervention, development of a child
with autism will not be enhanced.
It is important to note that on November 29, 1993, Dr. Love
circulated a letter to several parents of children with autism. This letter
contained most of the same misleading statements as the September 5, 1995
circular. Dr. Lovaas sent his reply to the November 29, 1993 letter to Dr. Love
on April 20, 1994. In this letter, Dr. Lovaas questioned the factual basis for
each one of Dr. Love’s misleading statements. In his reply to Dr. Lovaas on
May 3, 1994, Dr. Love admitted that he may not have had up-to-date information
about the UCLA program and that in failing to do so he may have based his
statements on personal opinions that could be considered outdated. Dr. Love’s
May 3, 1994 letter left open the door for a dialogue between the UCLA and the
TEACCH programs. Such a dialogue may have helped parents gain information from
which they could make the most appropriate decisions for their children’s
future. In that regard it is regrettable that subsequent to this exchange, in
September of 1995, we have a re-statement of the same misleading information to
B. Comments on a circular entitled "Early Intervention
for Young Children with Autism and Their Families: Recommended
Practices"(Circular distributed by the Pennsylvania Department of
Education, Bureau of Special Education, Harrisburg, PA)
The circular "Early Intervention for Young Children with
Autism and Their Families: Recommended Practices" reads "there is
sufficient evidence at this point to suggest a number of guiding principles in
the design of individual services for young children with autism and their
families" (p.1). It goes on to read "the organization and delivery of
services to children and families with autism should be based on the most
current and credible research. To do otherwise in a field undergoing such
profound growth in knowledge generation is to risk denying access to appropriate
services" (p.1). The intent is to "share with parents, on a continuing
basis and in a supporting manner, complete and unbiased information" (p.1),
to "respect the family’s right to accept or reject support and
services" (p. 2), to apply "state-of-the-art early intervention and
early childhood practices" (p. 2), and to "use evaluative data to
improve programs and services outcomes" (p. 2).
Answer: The stated intent of the circular, to inform parents
in an unbiased manner about what constitutes an adequate data-base from which
they may make decisions about their child’s treatment, represents a major
advance which undoubtedly will improve the future of children with autism.
However, after having read the circular, it is our opinion that parents and
professionals alike are in fact being misinformed and misled. Following are
certain major mistakes included in the circular.
It is noted in the circular that "in the largest study
conducted to date on early intervention for children with autism, all 51
children studied were integrated successfully with typical classmates (Strain,
1994). Significantly, this same study demonstrates long-term outcomes that have
not been surpassed by any other service-delivery model" (p. 3). We are
respectful of the work of Dr. Strain. On the other hand, so far as we know, the
only study published by Dr. Strain pertaining to outcome and published in a
peer-review journal is the study by Strain, Hoyson, and Jamieson (1985). This
study employed six subjects, reported no firm diagnosis of autism, and did not
employ a control group which would allow for a comparison of outcome so as to
assess the effectiveness of the treatment provided. With regard to the Strain
(1994) reference, "successful mainstreaming" cannot be used as a
significant outcome variable without also specifying the extent to which such
mainstreaming was accompanied by increased intellectual, educational, adaptive
and self-help behaviors, emotional adjustment, and so forth. This becomes
particularly important when one considers the limited gains made by children
with autism, as a group, when placed in the company of typical children. [This
note should not be used to encourage placement of children with autism into
groups of other children with autism. It just means that mere exposure to
typical children is not sufficient for improving the functioning of children
Of considerable concern to us is the use of the Strain (1994)
paper entitled "Peer-Mediated Treatment of Autistic Children," NIMH
Competitive Renewal, The Early Learning Institute, Pittsburgh, PA. This appears
to be a study in progress and should be referred to as such. However, the
circular may have led the reader to view the Strain (1994) paper as referring to
a study already completed and peer-reviewed.
Turning now to the many other studies referred to in the
circular in support of relaying unbiased information about appropriate
treatment, we call attention to the following:
1. Several of the studies cited do not report socially and
educationally meaningful outcome variables submitted to peer-review.
2. Many of the reported studies of social skills deal with
efforts to generalize skills rather than build new ones. Teaching other skills,
such as language, is not described at all, even though language is very
important in facilitating socialization.
3. No study advocates replacing discrete trial procedures.
Although there have been some efforts to develop a more "natural"
intervention, discrete trials remain the standard by which other procedures are
judged. In fact, several of the reports make use of the Lovaas et al. (1981)
teaching manual or the Lovaas (1977) language book, both of which emphasize
4. It is argued that "for young children with autism we
know of no experimental demonstration that has compared two or more levels of
intervention intensity" (p. 5). The Lovaas (1987) study, however, which is
so frequently cited in the circular, reports that two groups receiving different
intensities of intervention were employed, and demonstrated that a group which
received ten hours or less of one-on-one treatment did no better than a
no-treatment control group.
5. The circular refers to "in-class services ranging from
three to six hours, five days per week" as support for what is considered
adequate and appropriate treatment (p. 5). We know of no data to support such an
inference. Surprisingly, the same paragraph refers to the Lovaas (1987) study as
supporting this claim. The Lovaas (1987) study does not support this claim, nor
does the UCLA project.
6. It is noted in the circular that "total intervention
services, including in-class time and time spent at home and in community
settings, ranged between four to six hours, five to seven days per week"
(p. 6). Again, the Lovaas (1987) study is reported as supporting this claim
pertaining to in-class time, which is not the case. There is also a reference to
McGee et al. (1993) supporting this claim. We are not familiar with any study
published by Professor McGee that deals with outcome data.
7. In the circular it is stated that "within the range of
intensity across effective models there is no evidence that more is better"
(p. 6). We must again point out that data does not exist to support such a
statement, and that this statement may also be misleading.
8. A statement is made that "...the identification of
children as having autism younger than thirty six months is still relatively
unusual" (p. 6). This statement can not be supported by data. In fact, the
diagnostic instrument which currently employs the best data-base is labeled the
ADI-R by Rutter and Lord, and it places the cut-off at twelve to eighteen
months, nonverbal MA. There are also other tests designed specifically for early
identification of autism.
We could identify a number of other misleading statements in the
circular, but we hope that the information presented is sufficient to allow for
consideration of the withdrawal of this publication, and we strongly encourage
that this course of action be undertaken.
Let us return to reemphasize the goals that were stated in the
circular, namely the importance of providing adequate information to parents
which would help them choose the most appropriate treatments based on empirical
data to date. We strongly endorse such a position.
C. Comments on a Circular entitled "AUTCOM
(Memorandum was written by the "Autism National
Committee" Ardmore, PA)
In the following comments regarding the "AUTCOM
Memorandum," the terms "behavioral treatment" and
"one-on-one in-home treatment" will be substituted for the term
"Discrete Trial Training" (DTT). Discrete trials are used in
behavioral treatment, but comprise only a portion of such treatment and are
largely used during its beginning phases. Comments pertaining to the twelve
points listed in the memorandum are clarified below in the order in which they
appear in the memorandum.
1. It is argued in the memorandum that behavioral treatment
teaches a child that he or she is "damaged goods," consequently, the
child develops a self-image that he or she is "not good enough." Answer:
Behavioral approaches emphasize that children with autism have atypical nervous
systems and therefore have failed to understand what well meaning adults have
tried to teach them. Such children enter teaching situations with expectations
of failure. The treatment program is arranged, however, to maximize success and
minimize failure so as to build the child’s self confidence and trust in
adults (Lovaas & Smith, 1989). Behavioral psychologists express the most
optimism that children with autism can form healthy self-images.
2. The memorandum notes that children need to participate in
natural life; it is argued in the memorandum that behavioral treatment precludes
this. Answer: Behavioral treatment makes every effort to help the child
participate in the natural life of family and community and to develop
friendships and other social relationships. However, placing the child in a
"natural environment" without helping the child develop the skills
needed to interact in that environment will in all probability set the child up
for failure. There is every reason to believe that children with autism or PDD
grow up in normal and typical family environments (in contrast to what
psychodynamic theories have proposed) and yet fail to develop normally in those
3. The memorandum claims that family members, especially
siblings, will suffer when behavioral intervention is implemented. Answer:
Data show that parents who participate in one-on-one, home-based behavioral
intervention do not experience more stress than parents of children with autism
who do not participate in such programs (Smith & Lovaas, 1993). There is no
reason to believe that siblings of a child who undergoes treatment will fail to
develop positive relationships with that child.
4. Child development experts such as Dr. Stan Greenspan are said
in the memorandum to warn parents that behavioral treatment is likely to
increase perseveration and rigidity. Answer: This warning has no basis in
empirical data because outcome data show that children who receive behavioral
treatment are less rigid than a comparable group of autistic children who do not
receive such treatment (McEachin, Smith & Lovaas, 1993.
5. It is noted that Dr. Greenspan argues that behavioral
programs emphasize simple learning tasks which appeal to the child’s
"splinter skills" and fail to teach more complex beh