Sensory Integration Therapy with Learning Disabled

Children:
A Critical Review
Robin Schaffer

Abstract
This paper is a reaction to the profound impact which sensory integration theory has had on theory
and practice in occupational therapy. Five key outcome studies employing sensory integration therapy
with learning disabled children are critically reviewed. In four of the five studies, the possibility of
Type 1 error, destroying their internal validity, is strongly suggested as a result of inadequate sampling
and matching procedures; lack of definition of the subject population; potential Hawthorne effect;
the use of gain scores; possibility of tester bias; and absence of relevant information regarding reliability
and validity of measures and levels of significance. In one of the studies, the possibility of a Type
II error is evident because of unequal samples and the considerable variability among subjects indicated
by the data. It is concluded that at least some of the research to date is fraught with serious methodological
errors and that there is the need for controlled, systematic investigation of sensory integrative therapy.

The study of learning disabilities in children is one of Sensory Integration and Learning
the most controversial and confusing areas in the clini- Disabilities Review of some Literature
—

cal/developmental literature. Controversy stems from

vague definitions, from the large number of theoretical Error Types
models of learning disability (Wong, 1979a, I979b),
research design problems (Cohen, 1976), and widely In any review of research, the first assumption of the
varying forms of treatment and claims of success (Sapir reviewer is that a Type I or Type II error may have
and Wilson, 1978). Comprehensive recent reviews of the been made. Although the Type I and Type II error
literature, which regard learning disabilities from differ- margins are set mathematically and the mathematical
ing perspectives, are available to the interested reader risk of either error is known, the actual probability that
(Accardo, 1980; Faas, 1980; Kinsbourne and Caplan, a Type I or Type II error has occurred is not a mathe-
; 979; Ross, 1976; Wiig and Semel, 1976). matical or a statistical problem. It is a structural question
that demands a common sense analysis of the actual
Most authorities have linked learning disabilities to
conditions under which a study was carried out
some form of subtle neurological dysfunction (Johnson
(Kershner et al, 1982). Type I error is the finding of
and Myklebust, 1967) or neurodevelopmental lag
a treatment effect when, in fact, there is none; Type
(Rourke, 1978). However, the study of brain-behaviour Il error states that there is no treatment effect where,
relationships is still in its infancy. Coles (1978) notes
in fact, there is one.
that serious methodological problems are characteristic
of the majority of the research which supports claims
Effect of Therapy on Academic Achievement
for neurological deficiencies in academically delayed
children. The first major research (Ayres, 1972) relating to SI
One approach that has not, as yet, been fully and therapy and learning disabilities tested the hypothesis
systematically investigated is Ayres' (1974) theory of that LD children with certain identifiable types of SI
sensory integrative (SI) dysfunction and learning dis- dysfunction who receive SI therapy will show greater
ability. While Ayres' influence on the field of occupa- gains in academic scores than LD children with the same
tional therapy has been immense, there are relatively degree and type of SI dysfunction who get an equal
few references to her work outside that field of literature. amount, in time, of academic work. The sample com-
In the following paper, five key outcome studies regard- prised two experimental and two control groups. It was
ing SI therapy with learning disabled (LD) children are concluded that the gains in academic scores were proba-
critically reviewed. bly related to enhanced SI resulting from the therapy.
Thus the possibility of Type I errdr, i.e., that significance
was found where none, in fact, existed, must be consid-
ered.
In considering whether adequate controls existed for
this study, sampling procedures and definition of the
population become important issues. The children com-
Robin Schaffer, M.Ed., B.A., B.O.T., 0.T.(C)., Director, Division of
Occupational Therapy, Department of Rehabilitation Medicine, Uni-
prising the initial sample were 'selected' by unspecified
versity of Toronto, Ontario, M5T 1W5. school personnel with no reference to whether the usual
APRILi.AVRIL 1984 73
Downloaded from cjo.sagepub.com at UNIV CALIFORNIA SAN DIEGO on June 8, 2015
criteria for defining learning disabilities (Chalfant & school personnel; however, unlike the earlier study, the
Scheffelin, 1969, p. 148) were adhered to. Some were identification was made with reference to two criteria
in contained classes for the educationally handicapped, frequently. included in a definition of learning disabili-
others were in regular classes receiving individual in- ties, namely, 1) children whose academic achievement
struction. Thus, no random assignment to experimental is below intellectual capacity, and 2) children with
and control groups occurred to control for both system- average or greater than average intellect who are unable
atic and error variance. to profit from placement in a standard class. However,
The Hawthorne effect (i.e., an effect which is a result children with IQ's in the borderline retarded range were
of perceived change in physical conditions or an effect included in the study. The study concluded that there
due to attention) may have been influential since die is a certain type of vestibular disorder, detected in this
control groups, rather than receiving some 1-10k alternate study by a hyporesponsive postrotary nystagmus (HPN),
form of intervention — which would have been prefera- which intereferes with academic achievement and can
ble — only received additional classroom time. Perhaps be ameliorated through SI therapy.
the one-to-one attention inherent in the SI therapy In evaluating the internal validity of this study, several
situation might account for some of the improvement. methodological flaws point to the possibility of a Type
Although the independent variable, i.e., SI therapy, I error. Although there was a control group matched
is not delineated in enough detail for replication and to the experimental group for age, school and IQ, the
treatment personnel are not specified, this writer assumes experimental children with HPN were more intelligent
that SI therapy was carried out by qualified occupational than those experimental children without HPN, which
therapists whose expertise with this approach would be might account for some of the gain. While data regarding
available to other investigators. A lack of information the matching of children with HPN on age, IQ and
regarding the reliability and validity of the measures sex variables is presented, there is no indication as to
employed is evident; these factors must be considered the significance of the differences, i.e., the p levels. Thus
as possible contributors to Type I error. While the data IQ may constitute a source of extraneous variance
indicate significant improvement in raw scores, there interfering with the systematic variance of independent
is no indication whether these improvements would hold variables necessary for confidence in conclusions. There
for standard scores. While the study concludes that SI is no information regarding whether blind procedures
therapy resulted in enhanced auditory-language skill were implemented; thus, tester bias may have been
and SI function, no data is shown for the Southern influential, contributing to a Type I error. Moreover,
California Sensory Integration Tests (SCSIT) and the controls were inadequate for the Hawthorne effect since
Illinois Test of Psycholinguistic Abilities (ITAP) scores. the control group received only additional regular
In fact, only part of the pre-test battery was administered classroom experience. The study did attempt to control
at post-testing. for variations in schools by drawing experimental sub-
While the design of the study is subject to a three jects the first year, and control subjects the next, from
factor analysis of variance (group x test time x type each school. Random assignment to group was not
of SI dysfunction), F tests were not employed. Another implemented to control systematic and error variance.
issue which warrants consideration is the fact that of Assessment of the dependent variables in this study
the 148 initially selected children, 56 were eliminated begs consideration. No information is provided regard-
because they were not available for post-tests (N = 20) ing the reliability and validity of the various measures.
or because they did not perform poorly on the SCSIT Furthermore, Polatajko's study (1983) seriously chal-
(N =36). Inclusion of this second group might have been lenges the use of the Southern California Postrotary
an interesting test of the discriminant validity of the Nystagmus Test (SCPNT). Ayres proposes that the
SCSIT. SCPNT, a standardized test of postrotary nystagmus
In summary, the conclusion of this study that SI duration, reflects vestibular function; however, Polatajko
therapy results in enhanced academic achievement in maintains that the use of the SCPNT is questionable
LD children is untenable because of the strong possibil- on the basis of reliability considerations, since the
ity of Type I error. While these children may have stimulus is manually administered and, therefore, is
demonstrated SI dysfunction (assuming that the SCSIT subject to tester variation, and on the basis of validity
are valid and reliable — again, no information is provid- concerns, since to test vestibular function the head must
ed), it was not established that they were LD. A potential be maintained in a specific position; however, no method
Hawthorne effect, the use of gain scores, and the lack to control head position or correct for it is employed.
of stringent sampling and matching procedures — or In addition, because the test involves both visual and
inadequate information for the reader regarding these vestibular input, the resulting nystagmus cannot be
and other procedures — all contribute to a lack of deemed labyrinthine solely but is more likely optokinetic
confidence in any conclusions drawn by the investigator. nystagmus superimposed on vestibular nystagmus. Thus,
The second major study of the efficacy of SI therapy the SCPNT is a questionable measure of vestibular
(Ayres, 1978) attempted to examine the issue of which function.
neural conditions predispose the LD child to responsi- In summary, this study's claim that SI therapy ame-
veness to SI treatment as well as related issues and to liorates the dysfunction identified by HPN and promotes
explore the effect of SI treatment on disordered sensory efficiency of academic learning is suspect since there
processing in LD children. Again, as in the previously is a strong possibility of Type I error based on extraneous
discussed study, the sample was identified as LD by and error variance, the possibility of tester bias and a
74 CJOT - VOL. 51 NO. 2
Downloaded from cjo.sagepub.com at UNIV CALIFORNIA SAN DIEGO on June 8, 2015
Hawthorne effect, and unreliable and invalid measures. the stated conclusions would point to consideration of
In addition, confidence in the study's conclusions is a possible Type I error, the data warrant analysis in
weakened by lack of evidence, i.e., the absence of terms of a possible Type II error, i.e., no significance
relevant data tables as well as of p values in the tables was found where, in fact, it existed.
that are present. As Kershner and Cooke (1982) state: Several methodological factors support claims for
"Rejecting the null hypothesis demands statistical sup- error in this study. One is that the samples ma). be
port" (p. 13). unequal, accounting for a possible Type I or Type II
The third study for evaluation (Angelo, 1980) error. The experimental and control groups are not
proposed that SI therapy would improve the reading matched for size; there is no indication whether this
skills of low-achieving college students. The sample difference as well as differences between the groups on
consisted of twelve women who were either in the School the independent variables of age, IQ, and sex are
of Occupational Therapy at Texas Women's University significant. While this author believes that the groups
or had made application to it; had a low grade point should have been matched for degree of choreoathetosis
average or low scores on the Scholastic Achievement and for deficits in eye-hand coordination, in fact, the
Test or Aptitude College Test; or had poor reading skills, experimental group displayed more neuromuscular in-
visual perceptual problems, or difficulties with motor volvement; however, again no information is provided
performance. regarding the significance of this difference. As Kershner
The possibility of a Type I error is raised by the study's and Cooke (1982) note: "Where the samples for treat-
conclusion that SI therapy produced a statistically sig- ment conditions are unequal, randomness is destroyed
nificant improvement in comprehension and reading and a systematic bias may be added to the differences
skills. The major source of error is the absence of a among means which cannot be desentangled from the
control group and lack of randomization. No control influence of the treatment effects" (p. 11).
exists for the Hawthorne effect nor for the possible effect As in the two studies discussed previously, a possible
of an interest in occupational therapy. There is no Hawthorne effect exists for the same reason; an effect
indication that 'blind' procedures were implemented to may also be due to the individual attention received
control for tester bias. Regarding proper specification by the experimental subjects. No information regarding
of independent variables, motivation was deemed to be tester and treatment personnel is provided; again, it is
an important variable; however, its effect was not moni- unknown whether blind procedures were followed. Re-
tored nor was it evaluated by objective measures. In- garding the issue of specification of independent vari-
stead, independent ratings or indicators of motivation ables, the use of Schilder's Arm Extenstion test is
were taken by a therapist and an observer which resulted questionable in terms of reliability and validity since
in 'similar' rankings. Serious questions regarding the this is a non-standardized test involving clinical observa-
reliability and validity of this procedure must be raised. tions of choreoathetoid movements. In terms of assess-
In the area of proper assessment of dependent variables, ment of dependent variables, no information regarding
the SCSIT was employed to measure motor performance the reliability and validity of the MAC is provided.
and visual perception. However, this battery is not Support for a Type II error may be based on the
standardized for this age range. Thus, the internal considerable variability the data indicated among sub-
validity of this study is challenged on the basis of a jects. "Some improved appreciably while others had
strong possibility of a Type I error. lower post-test raw scores than pre-test raw scores"
(Ayres, 1977, p. 292). Some of the variability may have
Effect of Therapy on Eye-Hand Coordination been due to the difficulty LD children have in attending
carefully to a demanding task, resulting in reduced
A fourth study (Ayres, 1977) examined the effect of SI reliability of test scores. The large variations in change
therapy on eye-hand coordination in LD children who increase the difficulty in demonstrating statistically sig-
exhibit choreoathetosis. Ayres notes that "many children nificant changes, contributing to Type II error. Finally,
with minimal brain dysfunction show very mild involu- gain scores are inappropriately employed for interpreta-
tary motions when musculature is under voluntary tion where groups were not equal initially on the
contraction... [They] reflect a source of interference dependent measure.
with fine motor coordination" (p. 291). It was concluded
that eye-hand coordination, as measured by the Motor Effect of Therapy on Language Development
Accuracy Test (MAC) of the SCSIT, of children with
minimal brain dysfunction improved with SI therapy A recent study (Ayres and Mailloux, 1981) grew out
administered individually or in pairs one half hour per of the Ayres study (1972), already discussed, which
school day for six months. The population for this study contributed to an interest in the association between
was drawn from the larger sample used in the previously the body senses and language development. This later
discussed study (Ayres, 1978). study hypothesized that some aspects of language devel-
While the study concludes that a positive effect oc- opment depend on the motivation and processing of
curred, i.e., the null hypothesis was rejected, in fact the somatosensory and vestibular input and the generalized
data indicate that the levels of significance of difference capacity to make adaptive responses. Therefore, SI
between the mean gain scores of the experimental and therapy will produce a change in patterns of language
control groups were .058 and .061, i.e.. they did not reach development in children with language disorder. (Note:
significance at the conventional < .05 level. Thus, while The study has been included in this review, even though
Downloaded from cjo.sagepub.com at UNIV CALIFORNIA SAN DIEGO on June 8, 2015
APRIL/AVRIL 1984
the sample is too young to conform to the major criteria appears to this writer that the Carrow Elicited Language
for definition as LD relating to academic achievement; Inventory (CELI) is more a test of verbal memory than
inclusion is based on the sample having been found expressive language. Finally, no statistical analYsis is
to demonstrate SI dysfunction and on the fact that the provided to support rejection of the null hy-pothesis, i.e.,
diagnosis of developmental aphasia is frequently among to demonstrate that differences in pre-treatment and
the terms, such as minimal brain dysfunction and dys- post-treatment scores are significant. Thus. the conclu-
lexia, which are included under the broad umbrella of sion that SI therapy resulted in a consistent increase
'learning disabilities'). Four aphasic children with SI in the rate of growth in language comprehension and
dysfunction received SI therapy as part of a single case expression is untenable due to the strong possibility of
experimental study of each child. The research design Type I error.
is typical of the elemental A-B methodology for con-
trolled single subject research described by Hacker Conclusion
(1980) where A represents baseline data gathering of
target behaviours and B represents intervention. The Based on the foregoing review of five key outcome
study concludes that, with SI therapy, there was a studies, one is forced to conclude that SI therapy remains
consistent increase in the rate of growth in language a perhaps promising, but virtually unresearched, tool.
comprehension compared to the previous growth rate These five papers are fraught with, serious methodol-
because auditory processing, enhanced by vestibular ogical errors which destroy the internal validity of each
input, may have improved. Secondly, the two children study so that external validity does not yet warrant
with HPN were deemed to have made notable gains examination. The major flaws center on lack of adequate
controls to control systematic variance, such as non-
on expressive language.
equivalence of groups at pre-test, as well as absence
These two conclusions are subject to a possible Type of control for the Hawthorne effect, and for tester and
I error. Inadequate definition of the population is evi- subject bias. Lack of clear definition of subject popula-
dent. The four children were diagnosed as aphasic "at tion exists as well as inadequate presentation of data,
other agencies"; however, no criteria for this diagnosis such as the lack of data on changes in SI function, and
are presented. Also, it is stated that one child "was relevant information regarding reliability and validity
considered to be low in development in all other areas" of measures and levels of significance. In short, the need
(p. 384); the possibility of mental retardation is suggest- for systematic research, at a time when the public is
ed but not fully explored. Because of the single subject becoming increasingly sensitized to the problems of LD
research design, this issue does not have as great impli- children and the profession of occupational therapy has
cations for internal validity as it does for external validity so overwhelmingly and uncritically jumped on the
and the generalizability of conclusions. bandwagon of SI, has never been clearer.
The major methodological flaw of this study contri-
buting of Type I error lies in the absence of systematic
variance to maximize the effect of the independent REFERENCES
variable, SI therapy. Extraneous variance is a result of AcLardo, P. J. (1980) neurodevelopmental perspective on spec'
the individual speech therapy, which was started for learning disabilities Biltimore: University Park Press.
3 children during the baseline period and for one child Angelo, J.K. (1980) Eflects of sensory integration treatment on the
during the intervention phase; the special language low-achieving college student. American Journal of Occupational
Therapy, .34. 671-675.
instruction, started in the aphasic class for 2 children Ayres, A.J. (1972). Improving academic scores through sensory nte-
during or before intervention; and the developmental gration. Journal of Learning Disabilities, 5, 338-343.
therapy, started for one child during intervention. In Ayres, A.J. (1974). Sensory integration and learning disorders. Los
addition, the effect of maturation on language devel- Angeles: Western Psychological Services.
opment has not been controlled for — an effect which Ayres, A.J. (1977). Effect of sensory integrative therapy on the coordin-
ation of children with choreoathetoid movements. American Journal
is of particular concern for the age range of the subjects. of Occupational Therapy, 31, 291-293.
This period from four to five and a half years is Ayres, A.J. (1978). Learning disabilities and the vestibular system.
frequently a time for much natural development of Journal of Learning Disabilities, 11, 18-29.
language ability in children. The provision of baseline Ayres, A.J. and Mailloux, Z. (1981). Influence of sensory integration
periods equal in duration to the intervention periods procedures on language development. American Journal of Occupa-
tional Therapy, 35, 383-390.
might at least have constituted an attempt to control Chalfant, J.C. and Scheffelin, M.D. (1969). Central processing dys-
for this effect, although no procedure can fully control functions in children. HINDS Mongraph 9 Institute for Research
for the effect of maturation in single subject research. with Exceptional Children, U.S. Department of Health, Education
No mention is made of test and treatment conditions & Welfare.
Cohen, S.A. (1976). The fuzziness and the flab: Some solutions to
and whether blind procedures were implemented to research problems in learning disabilities. Journal of Learning
control for tester bias. In the area of assessment of the Disabilities, 10, 129-139.
dependent variable, one must consider the question of Coles, G.S. (1978). The learning disabilities test battery: Empirical
comparability between different language comprehen- and social issues. Harvard Educational Review, 48, 313-341.
Faas, L.A. (1980). Children with learning problems. Boston: Houghton
sion tests since, for two children, different measures were
used to collect baseline and post-treatment data. No Hacker, R. (1980). Single subject research strategies in occupational
information is provided regarding reliability and validity therapy: part 1. Atnerican Journal of Occupational Therapy, 34,
for any of the measures of the dependent variable; it 103-108.

76 Downloaded from cjo.sagepub.com at UNIV CALIFORNIA SAN DIEGO on June 8, 2015

CJOT — VOL. 51 — NO. 2
Johnson, D.J. and Myklebust, H.R. (1967). Learning disabilities: Rourke, B.P. (1978). Reading, spelling, arithmetic disabilities: A
educational principles and practises. New York: Grune and Stratton. neuropsychological perspective in H.R. Myklebust, Ed. Learning
Kershner, J. and Cooke. W. (1982). A jiffy guide to evaluating research disabilities, Vol. 4. New York: Grune and Stratton.
design and methodology type 1 and type II errors, unpublished paper. Sapir. S. and Wilson, B. (1978). A professional's guide to working with
Kinsbourne. M. and Caplan, P.J. (1979). Children's learning and the learning-disabled child. New York: Brunner/Mazel.
attention problems. Boston: Little Brown. Wiig, E.H. and Semel, E.M. (1976). Language disabilities in children
Polatajko, H.J. (1983). The Southern California Postrotary Nystagmus and adolescents. Columbus: Merril.
Test: A validity study. Canadian Journal of Occupational Therapy, Wong, B. (1979a). The role of theory in learning disabilities research:
50, 119-123. part I. Journal of Learning Disabilities, 12, 585-595.
Ross. A.O. (1976). Psychological aspects of learning disabilities and Wong, B. (1979b). The role of theory in learning disabilities research:
reading disorders. New York: McGraw-Hill. part II. Journal of Learning Disabilities, 12, 649-658.

Résumé
Cet article est en réponse l'effet marqué que la théorie de l'intégration sensorielle a produit sur
la théorie et l'exercice de l'ergothérapie. Une revue critique est faite de cinq études majeures utilisant
l'intégration sensorielle avec des enfants ayant des troubles d'apprentissage. Dans quatre études, la
possibilité d'erreur de Type I, qui détruit leur validité interne, est fortement suggérée suite des méthodes
inadéquates d'échantillonnage et d'appariement des sujets: un effet Hawthorne potentiel: au manque
de définition de la population: l'emploi de scores bruts: la possibilité d'un parti pris des investigateurs;
l'absence d'information pertinente en ce qui concerne la fiabilité et la validité des méthodes et les
niveaux de signification. Dans une de ces études. la possibilité d'erreur de Type II est évidente cause
des groupes inégaux et de la variabilité considérable entre les sujets indiquée par les résultats. On conclut
que certaines des recherches démontrent des erreurs méthodologiques sérieuses et qu'il y a un besoin
pour des études contrôlées et systématiques en ce qui concerne l'intégration sensorielle.

manufacturing inc.
63 Cournand Ave., East
Kitcheaer, Ontario
N2G 2T6
(519) 742-5041
*Aida For Daily Living
People working and learning: a non-profit organization

-nom•
c.4

WHEEL CHAIR TRAY: These Trays are custom-

built to the specifications of your wheelchair. A custom-
SHOWER BOARD / BATH BOARD: This sized hole for a motor box is optional. The height and
adjustable bath board is our most popular item. Made of
special 1/2" five-ply birch plywood, it was tested by the
placement of the edges are made to conform to your
requirements. Specify wood or clear plastic.
University of Waterloo at a breaking strength which ex-
ceeds 1000 lbs. Special raised boards for hip ailments etc. -wooden (give dimensions required) 30.00
can also be made to order.
20.00 -plastic 45.00

APRIL/AVRIL 1984 Downloaded from cjo.sagepub.com at UNIV CALIFORNIA SAN DIEGO on June 8, 2015
77