RENEE BRADLEY
Welcome to the session on the classification approach. I'm Renee
Bradley for the US Office Of Special Education Programs, and I'll
be serving as your moderator for this session. During this session
you will hear a 30 minute presentation by the author of the issue
paper on this topic, followed by nine minute presentations from
each of the paper respondents. There will be a 30 minute period
for questions and answers at the end of this session. In order
to provide ample time for the presentation, please reserve your
questions until the end. This session is being video and audio
taped for later broadcast. It is therefore imperative that we
remain within the time limits assigned. In addition, if you would
like to address a question to the presenter during the Q and A
session, please use the microphone that will be circulated throughout
the audience. Out first speaker today will be Dr. Jack Fletcher,
who is a professor in The Department Of Pediatrics for the University
Of Texas Houston Health Science Center, and Associate Director
for the Center For Academic And Reading Skills. For the past
20 years, Dr. Fletcher, a child neuropsychologist, has completed
research on many aspects of the development of reading, language
and other cognitive skills in children. He has worked extensively
on issues related to learning and attention problems, including
definition, classification, neurobiological co-relates and most
recently, intervention. Jack.
JACK FLETCHER
Thank you. Thank you. It's important to recognize that I'm not
a pediatrician, I'm a psychologist, which is the first thing I'm
gonna clarify. And I'm gonna do something that's a little bit
unusual in a presentation like this, I'm going to tell you before
hand what the take home message is not. There are three mistakes
that people make, uh, when they hear these, uh, studies. What
happened? I'm in a hurry. The first is, that it's very important
to recognize that I work in schools, I evaluate children for learning
disabilities, I go to interdisciplinary team meetings and I go
to hearings, and I believe, uh, that identification is supposed
to rely on more than test scores. I know that Idea says that.
The problem is that it often doesn't. The second is that I believe
that learning disabilities is real, but the current definition,
the current classifications do not capture this reality. In particular,
I believe that it is possible to question the approaches that
we take to definition and classification without necessarily,
uh, questioning the reality of LD. LD is real. I am not saying
it's not by virtue towards the questions I'm going to raise today.
And then finally I do not recommend that children be identified
as LD based solely on low achievement, uh, definitions. In fact,
I am skeptical that any single shot psychometric approach to definition
will be effective or work for all children. Having said that,
let's talk about classification. To understand the Federal definition
of learning disabilities, we need to understand something about
classification. Because the Federal definition itself is more
than just a definition. It is a series of classification hypotheses.
Now classification itself is an area of behavioral research.
Classification is the process of forming groups from a large
set of entities based on their similarities and dissimilarities.
We take any of these that are heterogeneous and figure out ways
to make them more homogeneous.
JACK FLETCHER
Classifications are always hypotheses, and they occur at the level
of the independent variable. In classification research, we explicitly
evaluate the basis for grouping. (WORD?) classifications can
be differentiated according to variables not used to form the
groups. In LD this could be things like the cognitive course
of LD, response intervention or prognosis. But it's very important
to understand that classifications are always evaluated by using
things that aren't classification attributes. You are guaranteed
assuring that a classification is valid if all you study is the
process by which the group is formed. And then finally, classification
is not identification. Identification is the process of assigning
entities to an established classification. Identification procedures
may be valid, even though the classification itself is not valid.
How we define something may be valid, but the overall or over-arcing
classification may not be. Here are the levels of classification
for people with learning disabilities. We typically classify
learning disabilities along several different dimensions. Learning
disabled versus typically achieving, learning disabled versus
mentally deficient, reading versus math disabled, IQ discrepant
versus low achieving. Each level represents an implicit classification
hypothesis that can be evaluated. And if you look at these four
levels of classification and you simply ask a question, is LD
real? Well, the two levels that you would look at are these.
LD versus typically achieving, reading versus math disabled.
And if you look at these, you get ample evidence that LD is a
real disorder, part of a real classification. For example, here,
all I've done is I've graphed performance by groups of children
to find a typically achieving, reading disabled and math disabled.
The only disorder they had no, (UNINTELLIGIBLE)
JACK FLETCHER
I've got cognitive skills down here like sustained attention procedural
learning and concept formation, these are real important for math
disabilities, philological awareness, real important for reading
disability and then variables where the differentiation is not
as clear like visual motor skills. And you can see that I can
easily differentiate the two learning disabled groups on all these
dimensions here, based solely on level of performance. I can
also differentiate these groups here based on the quality, the
shape, of these profiles. These are not the same shapes that
you see here. And then I can also differentiate these two groups
based on the shape of their profile. Look a the strength that
this group has, the reading disabled group and procedural learning,
the weakest of philological awareness. Uh, the weakness over
here in procedural learning and concept formation in the math
disability group. This is very strong evidence that indicates
that learning disability classifications of this sort have discriminative
validity and represent a viable classification. How well is this
captured in the Federal definition of learning disabilities.
Well, there are essentially three elements of classification in
the Federal definition. The first is the discrepancy hypothesis,
that IQ discrepancy represents an inclusionary criterion that
differentiates children with learning disabilities from those
who are IQ consistent children. Their achievement is poor, but
it's consistent with their level of IQ. The second is the heterogeneity
hypothesis, and simply formulated it's the hypothesis that there
are seven different sub groups of lower, of, of academic deficits
in the LD category. And the third is the exclusion hypothesis,
which is that learning disabilities are different from achievement
definitions that are due to mental deficiency, emotional disturbance,
economic disadvantage, inadequate instruction.
JACK FLETCHER
Each of these elements is a hypothesis that can be evaluated in
classification and research, and what I'm gonna do is review the
evidence for these three hypotheses for you. First, IQ discrepancy.
The discrepancy hypothesis. Now if you hear the history, you'll
hear abut the Isle Of Wight, uh, studies. These are studies that
were done, uh, on the Isle Of Wight in the UK by Rudder And Uell
(SP?) and they were pretty exciting because they seem to support
the idea that IQ discrepancy was a valid marker for a specific
form of reading disability. Rudder And Uell studied all the kids
that were on the Isle Of Wight, uh, it's a large number of kids.
What they found was a bi-level distribution reading scores.
So they were able to differentiate general backwards reading from
what they call specific reading retardation. So that there was
a hop in the bottom, in the bottom of the distribution with a
long tail represented, representing general backwards reading.
In other words the distribution was not normal, and there was
a long hop, but the hop itself were backwards readers, IQ consistent
children. they found differences between backwards readers and
specific reading retardates in things like prevalence, uh, gender,
neurological finds, reading and spelling characteristics and prognosis.
Note, however, that they were not different. These two groups
were not different in most speech and language characteristics.
They had similar impairments in the language area. And then
keep in mind that the Isle Of Wight is the Isle Of Wight. It
maybe an epidemiological sample, but it's not necessarily a representative
sample. And just to illustrate, 36 percent of children were general
backwards readers had a known or suspected central nervous system
disorder, and many were mentally deficient. There were no exclusionary
criteria applied in the Isle Of Wight studies because the question
that Rudder And Uell were asking is whether there is a syndrome
of specific reading retardation that stands out against the backdrop
of all other forms of reading retardation.
JACK FLETCHER
And they basically concluded that there was evidence, uh, for
this. Well, this has been very controversial, uh, both in terms
of the issue of the bi-mobile distribution, and the issue of whether
there are differences in the co-relates. In terms of whether
there's a hop, uh, in the distribution of, uh, reading scores,
there are now six different, five different epidemiological studies,
I'm sure, in here. There are also three others that I haven't
listed because they weren't actually published. Uh, across different
regions of the country, going back to 1983, different international
studies, and virtually all of these studies failed to find evidence
for bimodality in the distribution of reading skills. Uh, in
fact, the study that was concluded in 1992 by the Schiowitzes
(SP?) at Yale, concluded that reading ability occurs along a continuum
that blends imperceptibly with normal reading ability. It's a
lower portion of a continuum of reading capabilities. Now this
is a very important finding, because it says in essence that reading
is a dimension that is normally distributed in our population.
It says that we only need one theory to explain how kids learn
to read an why some fail, and most importantly it raises the problem
of where do you define disability. At what point in this continuum
do you demarcate it, in order to show that some people are reading
disabled and some people are not, and the problem is that any
psychometric approach to this, which I'm not going to show you
today, but it's in our paper, is inherently unstable because once
you set that cut point, there are all these border line cases
that hover around it that are unstable that will be discrepant
on one day, not another, uh, and so on. And that's a really big
problem.
JACK FLETCHER
Well, Rudder said, that in essence, in 1989, well, there's no
hop but the hypothesis is still viable. And the question is whether
we can identify characteristics that differentiate backwards readers
from reading retardates. Well, let's look at some of this evidence.
Take (WORD?) cognitive skills, uh, for example. Are there cognitive
differences between children who are IQ discrepant, and IQ consistent?
Well, there are three meta-analyses now that address this. One
of them is being presented here, uh, by Doug And Lynn Fuchs and
their colleagues, and the research questions they ask is not really
a classification hypothesis. It's, is the reading performance
of under achieving children with and without the learning disabilities
labeled the same, or different. The Fuchs code is 79 studies
comparing children labeled as LD with children who presumably
had low reading achievement but no label. They had no specific
criteria for reading disabilities and low achievement. They relied
primarily on the label which is entirely appropriate because with
the research question, uh, here, and they did not, this is the
one that I think is important, did not differentiate variables
used to form the groups from those that served as dependent variables.
They were mostly interested in reading. They compared reading
scores, uh, in a meta-analysis of children defined as LD versus
children not defined as LD who presumably met some definition
of low achievement. And this is what they found. In meta-analysis,
the (WORD?) between point two, uh, around, around, between minus
point two and point two are considered small and negligible, not
meaningful. Not meaningful differences between the groups, and
then those that are larger are considered to be more meaningful.
And what was reported is an effect size in the moderate range
of point six one for reading, but look at here. (UNINTELLIGIBLE)
awareness, rapid naming, not used to define the groups, most reliable
co-relates for reading that we know, uh, philological awareness
in the negligible lane, rapid naming is in the small range. A
second meta-analysis published in the year 2000, the Journal Of
School Psychologists by Hopkins And Lee Swanson. They quoted
19 studies with stringent IQ and achievement criteria, uh, and
this is what they, uh, found. Uh, look at (WORD?) reading, takes
out a negligible effect size of point oh two, pseudo-word reading,
small effect size of point two nine, philological processing was
small (WORD?) was negligible.
JACK FLETCHER
Notice how much smaller the effect sizes are in reading than those
that were reported in Fuchs. And then our meta-analysis, uh,
which is not published but is under review, where we quoted 46
studies I which groups could be clearly identified as discrepant
or low-achieving, but separated variables usually define the groups
from those used to evaluate differences between the groups. We
tested IQ discrepancy as a classification hypothesis. Now this
is a lot of data, but what it boils down to is right over here
in charge of these descriptions. Uh, for example, in the behavioral
domain, we found negligible effects by differentiating groups.
The overall effect for achievement as negligible, but it was
heterogeneous and in essence we find small effects for some skills,
like pseudo word reading and real word reading, but negligible
effects for things like reading comprehension and math. Why would
two groups that differ so much in IQ not differ in math? It's
an interesting question. Any cognitive ability, if you look at
these core skills which Joe Torgesen will tell you later at this
meeting, are most closely related to reading. We had negligible
effects for all four. No evidence for difference. In other analyses,
we found that the overall difference in cognitive ability was
about three tenths of a standard deviation, uh, which means that
there is substantial overlap of IQ discrepant and lower achievement
groups and perhaps most importantly, found that effect size variation
of 12 different studies can be predicted simply by knowing the
IQ and achievement criteria used to define, uh, the groups. In
other words, it's sampling variation. And this is what you're
talking about. This is the same slide I showed you earlier.
All I've done is I've taken the kids in that slide and divided
them into IQ discrepant and low achievement, uh, groups, and you
can see that these groups are not very, uh, different. These
are not statistically significant differences, uh, for example.
Uh, and if I adjust for their vocabulary level right here, I
eliminate every bit of difference between these two groups. What
co-relates most highly with IQ? Vocabulary.
JACK FLETCHER
What about other areas? Prognosis, uh, for example. This is
data from The Connecticut Longitudinal Study looking at kids all
the way from kindergarten to adulthood. Uh, these kids are now
about 22 years of age. We defined them, uh, into groups of kids
who met low achievement definitions, uh, normal, statistically
achieving kids, and kids who met discrepancy definitions. Uh,
notice that there is not a whit of difference in the long term
growth of reading skills between these two groups of kids despite
the fact that they have IQ differences of about 18, uh, points.
No difference in prognosis, in fact, that's probably an artifact
of regression to the mean. Treatment validity. Does IQ or IQ
discrepancy predict, uh, response to outcome? Here are six studies
looking at relationships with word recognition outcome, you know,
mostly since 1998 or so and you see that there's essentially no
relationship in virtually all of these studies except for the
study by Barbara, where only one of the three outcome measures,
uh, uh, yielded an effect and explained only five percent of the
variance. Only one study looked at IQ discrepancy, but there's
substantial overlap here, and that was Frank Vellutino's study.
And he concluded that the IQ achievement discrepancy does not
reliably distinguish between disabled and non-disabled readers,
children who were found to be difficult and easy to remediate,
and it does not predict response to remediation. Now everybody's
question immediately is what happens if you study comprehension?
Well, there are much smaller relation ships of IQ in some studies
with comprehension outcomes. But think about what's driving that
IQ score? It's vocabulary, and vocabulary's a very robust predictor
of reading comprehension skill.
JACK FLETCHER
What about other forms of learning disability? For example, here
are our same variables here, cognitive variables here, in terms
of a mass disability group defined as being IQ discrepant or low
achievement. The shape of these profiles is not different. The
level of performance is different, because of this big difference
here that you see in vocabulary. Well, actually vocabulary has
nothing to do with math performance in these kids. It's all these
variables over here, problem solving, concept formation and things
of that sort. In fact, they got the vocabulary, uh, difference
which is a product of the definition itself. Because IQ tests
contain vocabulary that I eliminate this difference right here.
Speaking language, uh, disorders. This is has been studied extensively
in the large-scale, uh, studies by (SOUNDS LIKE) Bruce Tomlin.
And this is what he concluded. The children with general language
delay closely paralleled the specifically language impaired group,
current diagnostic methods and standards for specific language
impairment do not result in a group of children who's profiles
of language achievement are unique. And a working group from
The National Institute Of Deafness And Communication Disorders
recently published a paper in which they concluded that there
was little validity to what's called cognitive referencing, or
the comparison of language scores, uh, to language but to, to
some sort of index of aptitude, like IQ, for example. This is
where the field, where many in the field are, those who have studied
this in a very systematic sort of way and summarized actually
seven years ago by Keith Stanovitch and Linda Segal, neither the
(WORD?) nor the (WORD?) indicators of poor reading are co-related
in a reliable way with IQ discrepancy. It is a classification
hypothesis that has very weak validity at best. Heterogeneity,
disorders of oral expression, listening comprehension, written
expression, basic reading, reading comprehension, mathematics
calculation or mathematics reasoning. The question is, are there
seven sub-groups of learning disabilities? Well, think about,
for example, oral expression and listening comprehension. First
of all, these are problems that are essentially assumed under
speech and language impalement. And when I talk about this, I'm
only questioning whether they belong in a classification of children
with learning disabilities.
JACK FLETCHER
I recognize that there's overlap. I'm simply pointing out that
disorders of listening and speaking are essentially speech and
language disorders, and there's another classification, uh, for
these kids. But, uh, more importantly listening comprehension
usually parallels reading comprehension. So why would you talk
simultaneously about problems with listening comprehension or
listening, and reading comprehension? This needs to be looked
at. Basic reading and reading comprehension. This is probably
the strongest evidence that we have, uh, for, for the heterogeneity
hypothesis. These are essentially different sorts of problems.
We're talking about word level versus text level disorders, and
there's good evidence to talk about differentiating these two
sorts of problems. The difficulty is that this does not address
fluency. And we have good evidence now for three forms of disability
in reading that co-occur and occur in isolation, disorders involving
word recognition, comprehension and fluency. What are we gonna
do about fluency? It's important that disorders in a classification
occur in isolation so that it's not simply co-morbidity and so
that you can identify what the prototype, uh, would be, that would
drive a classification. Math. Talks about math calculation and
reasoning. The big problem with classification of the math disability
is that we don't know what the marker variable is. With word
level disorders it's word recognition. With reading comprehension,
it's comprehension of text, although that's difficult to assess.
What component of math tells you that somebody is disabled?
Is it, for example, computational difficulty? Is computational
disability a marker for math? That's always the way I've approached
it, and the way you would do classification research is you would
systematically test this, as a, as a hypothesis, which we've been
trying to do.
JACK FLETCHER
In terms of the disorder of math reasoning, what part of math
does not involve reasoning or concepts? How do you measure math
reading or math concepts? This is a very difficult, uh, construct
and it's questionable as to whether it belongs. There is evidence
for types of math disabilities that you can see most clearly if
you evaluate math disability in isolation versus reading and math
disability together. Reading and math disability together is
a more pervasive disturbance of language and working memory.
It picks up components of both. This represents a problem learning,
representing and retrieving math facts as opposed to math without
reading disability which represents difficulties in the acquisition
and use of strategies and procedures. We are very close to having
an accurate resource base to talk about criteria for these sorts
of disorders. Written expression is a real problem. Are we talking
about problems with spelling, problems at a textual level or problems
with hand writing. And the major issue here is the independence
from children with reading disability, math disability, and attention
deficit disorder. Many of these kids have problems with written
expression, and do these problems occur in isolation? People
like Steve Graham and (SOUNDS LIKE) Virginia Bernard who are actively
researching, uh, this area, I think, will have evidence about
this fairly soon. Uh, bit this is certainly a viable hypothesis.
There is evidence for heterogeneity, is the bottom line. What
about the exclusion criteria? These are factors that cannot be
the primary cause of LD, sensory problems, mental deficiency,
emotional disturbance, environmental and cultural economic disadvantage
or inadequate instruction. The most important thing to understand
about these criteria, is that they cannot be the primary cause,
as they're evaluated in an interdisciplinary team. I'm not talking
about that. I'm talking about a classification question, which
is, essentially, are reading and math disorders different in kids
who might fit these exclusionary criteria. Do we need different
programs, for example, for children who are served in Title One
programs versus children who are served in LD programs, versus
children who are not served and are called slow learners, for
example. Well, the evidence does not support that. Essentially
the cognitive correlates and the instructional needs of academic
difficulties with children who have underachievement that's attributable
to emotional disturbance. It's usually co-morbidity to have both
a learning disability and an emotional problem, to economic disadvantage,
it looks the same, or to inadequate instruction, do not appear
to be different than those who are defined with LD. Ed Kavale
said in 1990, 1988, since culturally disadvantaged children have
been shown to exhibit the behavioral characteristics of LD, it
is difficult to determine why the culturally disadvantaged group
is categorically excluded from the LD classification.
JACK FLETCHER
Children from lower SDS levels with LD type behaviors have little
chance of receiving LD diagnoses and treatment with an increased
likelihood of being labeled retarded. They're not clearly identifiable
as separate entities. We make certain decisions about policy
in terms of how we serve kids, but we need to be careful about
what this says about the cause of learning disabilities. Inadequate
instruction is the one that's the most puzzling because inadequate
instruction causes learning disabilities. It is a primary environmental
cause. But it's rarely used in eligibility decisions. Response
to good instruction may be the best indicator of a true disability,
and early intervention prior to, to identification may be essential
if you take this question here seriously. So, in terms of exclusion,
intervention needs, response to intervention or mechanisms where
by interventions work do not appear to vary according to the classic
exclusionary factors. And as a classification hypothesis, exclusion
appears to lack validity. Now this gets even more complicated
when we think about what causes learning disabilities. And the
point here is that the boundaries of causation are blurred. It
is not nature versus nurture, genetics versus the environment,
it's some of both in most people. Historically, learning disabilities
have been attributed to intrinsic neurobiological factors within
the child largely because of exclusionary factors. The known
causes have been eliminated from consideration. But think about
this. Think about the brain imaging. Think about the genetic
studies first. We know from the genetic studies that have been
done in Colorado that only about 50 percent of the variants in
reading outcomes can be explained according to, to genetic factors.
JACK FLETCHER
That means that the other 50 percent is environmental. And what
are the environmental factors, this illiteracy environment in
which the child develops, and it's instruction. And instruction's
really very important. This, for example, is from some of the
functional imaging studies that, that people enjoy looking at,
uh, quite a bit. This is from magnetic source imaging, which
is a modality that we have at The University Of Texas, where we've
been systematically evaluate (SIC), uh, kids, uh, with, uh, learning
disabilities. And here we see a very characteristic profile that
you'll see in all of any study using any modality of functional
neuro-imaging. And that is, if you look at you're, uh, your good
readers here, you notice the last hemisphere, you notice activation
here in the basal temporal area, superior temporal jar, Wernicke's
Area and the angular jar. This is neural net work that mediates
word recognition and it's laterlized, mostly in the left hemisphere
and not the right hemisphere. But look what happens in kids that
have reading problems. It's over here in the other hemisphere,
and we're not gonna talk a lot about why that is or why you might
see these big, lateralized effects. What we're gonna talk about
is what happens when you teach kids, uh, to read and the answer
is that this pattern changes. It's reversible. Uh, these, and,
we have a group of kids, there are now about ten of them, who
went to a very intense, 8-week intervention, about 80 hours of
instruction. We imaged them before and after, uh, intervention.
Before you see the very characteristic, uh, reading disability
profile, a little activation over here in the left hemisphere.
Afterwards, lots of activation over here. Before, after. One
of our kids is 17 years of age. And this, this is, uh, same sort
of method from kids who are at risk for reading disability who
are imaged at the end of kindergarten, doing a letter sound task.
And you'll notice that we see the same sort of pattern of more
activation over here on the right, in the at risk child, not in
the not at risk child. We believe that with instruction, as these
children learn to read, that this pattern will change to this
one. So to conclude, the Federal classification of LD is outdated
and not in line with research.
JACK FLETCHER
There is little evidence for discrepancy as an inclusionary criterion.
There is support for hetero, for the heterogeneity hypothesis,
but domains need revision and a single definition does not appear
feasible. We need different definitions for reading disability,
for math disability and so on. The exclusionary criteria is not
valid from a classification perspective. We need domain specific
inclusionary treatments oriented definition. Here's a very quick
example of a definition of word-level reading disability, uh,
developed by a research committee from the IDA. Dyslexia is a
specific language based disorder characterized by difficulties
in the development of accurate, influent, single word decoding
skills, usually associated with insufficient biological processing
abilities. You met, how do you assess this? You measure reading,
you measure biological processing, you look at this and you develop
a treatment plan. It's a treatment oriented definition. the
concept of intelligence and learning disability is outmoded.
IQ tests lead to a focus on eligibility by which children are
sorted by IQ, not achievement. It's been called milk and jug,
uh, thinking. Children, uh, do learn at levels that are higher
than those predicted by IQ tests. IQ is not a measure of aptitude
or an indicator of a response to intervention, and they will be
an outcome of the same processes that lead to LD. Most importantly
is we do not need IQ tests to identify children with learning
disabilities and we could probably change the definition and change
how children are assess in schools in a way that's much more efficient
and cost effective. We need to develop a consensus process to
formulate a new classification of children with learning disabilities.
JACK FLETCHER
We need to focus on results because kids should improve, uh, when
they're, uh, placed in special education. We have little evidence
that they do. Low achievement definitions will not be adequate.
There is no basis for establishing cut points, and they are inherently
unstable and LD is real. The changes in how we think about LD
are needed. When the discrepancy formula disappears from the
educational scene, so will the concept of LD. Ha. The notion
of discrepancy has led to a confounding, most clearly seen in
the suggestion that there are more similarities than differences
between LD and low achieving students. Such a suggestion calls
into question the very notion of LD. Wrong. We can question
classification without questioning the viability of the construct.
And definitions may be valid when classifications are not. IQ
discrepancy validly identifies children with a learning disability.
The classification is what's not valid. And we should rejoice
that, that LD as a construct holds up independently of different
approaches to definition. The issue is how do we take what we
know and develop model classifications? I think we need a consensus
process to do this. Under achievement and learning disabilities
is expected, not unexpected, and is expected because of what we
know from research. It's like talking about yellow fever before
we knew that mosquitos spread the disease, or inflation without
knowing its relationship with spending, for example. It reflects
the accumulation of research over the past three years. Let's
line up the classification with research. Otherwise, why do it?
Thank you very much. (APPLAUSE)
RENEE BRADLEY
Thanks, Jack. Our first respondent this afternoon is Dr. Deborah
Speece, who currently holds the rank of Professor in The Department
Of Special Education at The University Of Maryland. Dr. Speece's
research interests include classification issues in learning disabilities,
the development of children at risk for school failure, and contextual
features that influence school performance. Our next respondent
is Dr. Linda K. Elksin, who is the immediate past president of
The Council For Learning Disabilities, and co-editor of Assessment
For Effective Intervention. She is a Professor Of Education at
The Citadel in Charleston South Carolina, and serves on the editorial
boards of Learning Disabilities Quarterly, and Exceptionality
and Intervention In School And Clinic. Our final respondent
is Dr. Nancy Larson, who has been the Minnesota State Learning
Disabilities Coordinator for the past eight years. Nancy is responsible
for overseeing the development in implementation of state guidelines
and rules, as well as for insuring field input into learning disabilities.
DEBORAH SPEECE
Good afternoon. Uh, did you get all that? There are three things
that I wish to emphasize in my response to the paper by Fletcher
and his colleagues. Convergence with other perspectives, expansion
of points raise, and caution with respect to the implications
of their position. First, convergence. It is serendipitous that
I've received the Fletcher paper as I finished data analysis for
a study on experts opinion on the definition of reading disability.
Hedges And Washington noted that convergence of evidence between
research and expert opinion provides validation for both, uh,
sources. (WORD?), it is instructed to compare the conclusions
of Fletcher and his colleagues and the opinions of experts who
are members of editorial boards for journals and learning disabilities
and reading. And I hope some of you were my subjects in that
study. We've asked these professionals among other things to
select components of a definition of reading disability for practice.
I selected the most important component, and to indicate whether
exclusion criteria should be part of the definition. Fletcher
and his colleagues offered three conclusions, first, IQ achievement
discrepancy is not a valid criteria, criterion for LD classification.
DEBORAH SPEECE
Second, exclusionary causes are not positive indicators of LD.
Third, the classification of LD will likely include responsiveness
to instruction, with respect to the validity of IQ achievement
that's, um, with respect to the validity of IQ achievement, uh,
validity, 70 percent of our experts, and this first line, 70 percent
of our experts, uh, believe that IQ achievement discrepancies
should not be part of the definition there by supporting pressured
positions. Because in the second point, on exclusion clauses,
75 percent of our respondents believe exclusion criteria should
be used, which contradicts the Fletcher position. On closer inspection
of the expert data, it was clear that our experts rejected the
criteria, (UNINTELLIGIBLE) from his colleagues found to be least
supportable. Specifically, emotional behavioral disability, cultural
differences and economic disadvantage. These criteria were selected
by fewer than 30 percent of the experts, there by lending some
support to the view that exclusion criteria are not useful. With
respect to the third conclusion in response to instruction, two
thirds of our experts agreed that treatment validity defined in
part as responsiveness to instruction, should be a component of
the reading, uh, disability definition. This is surprising finding
given the recent appearance of treatment validity in discussions
of identification. I will return to this point in consideration
of expanded approaches to identifying LD. The other point of
the claimant that responsiveness to instruction can play an important
role in the identification of LD, leads to my second theme, expansion.
And could you cover up all but the first line, please?
DEBORAH SPEECE
Responsiveness to treatment represents a critically important
ingredient to definitions of LD. Definitions in classification
of LD focus exclusively on (UNINTELLIGIBLE) of exclusion, to the
exclusion of the role of contextual factors including instruction.
That it took decades to begin to acknowledge the role of context
in LD, puts our field in the of must social sciences. Duncan
And Ryan first commented that social science research is far from
definitive about whether context matters, despite the fact that
context such as families never (WORD?) in school are essential
to making the child fully human. Fletcher and his colleagues
provide a coherent explanation of the problems associated with
a view of learning disabilities that neglects the effects of context
generally, and instruction specifically. What is missing in the
discussion is reference to an existing model developed by Beck
And Lynn Fuchs, that provides conceptual specificity to the link
between LD and the lack of responsiveness to instruction. Undoubtedly
this map, model, and I know now that it was, was talked about
in Frank Gresham's paper, however it requires mention here because
of its potential to address several points raised by Fletcher,
including that responsiveness to instruction be considered a classification
criteria for LD. Briefly, the treatment solidity model requires
continual monitoring of academic progress of all children in a
classroom, identifying children who failed to respond given a
generally effective instructional program and intervening within
the general education classroom with the identified children.
Children who fail to demonstrate progress after well-designed
and implemented general education intervention, would be candidates
for more intensive instruction in special education, and may be
considered learning disabled or at risk for learning disabilities.
DEBORAH SPEECE
Fuchs And Fuchs provided a complete analysis of this process including
preliminary data. What is of interest to the present discussion
is how this model supports pints raised by Fletcher. These points
include reunion of research and practice, focus on behaviors relevant
to instruction, acknowledgement that LD may be best indexed within
academic domains, and focus on inclusion criteria rather than
exclusion criteria. This seems to me to represent a substantial
convergence and expansion of ideas in an ares that has created
more anxieties than agreement. Other methods of conceptualizing
response to treatment, I mean, are possible and should be studied.
The primary issue, however, is that we can use the points of
agreement to guide an expanded field of LD, that recognizes that
classification is a dynamic process and that LD is a developmental
phenomenon. The second point under expansion, and if you could
just pull that down, Renee, is the need to consider a more comprehensive
view of validity in a second potential of any classification approach.
In the discussion of validity, Fletcher and his colleagues (WORD?)
their arguments in the realm of construct validity. This is appropriate
because context, uh, construct validity is the cornerstone of
any discussion on validity. However, analysis of classification
validity needs to be expanded to include other facets of validity
as defined by Messick (SP?), and I'm not going to go through all
four, uh, cells there. But briefly, Messick proposed that assessment
of validity depends on, put this out from the computer cells,
that you.
DEBORAH SPEECE
So Messick proposed that assessment of validity depends not only
on traditional psychometric evidence, but also on the consequences
of test use. In a present context, this means the consequences
of applying a particular method of LD classification. In Messick's
terms this terms this would mean an analysis of value complications
of LD classification, and the social consequences of LD classification.
These factors are located in the second row of the matrix. For
example, though validity of the IQ achievement discrepancy formula
as a classification criterion, should be assessed not only with
hypothesis related to low achievement, that is, construct validity,
but also in terms of the consequences of applying the criterion.
Specifically, we know that it is difficult for young children
to qualify for special education under the discrepancy criterion.
This option violates what many report as important value, the
ability of a classification system to identify children for services
before they experience repeated failure. In terms of social consequences,
the discrepancy formula as applied in schools, results in over
representation of males and minority children. However, steady
set use researcher defined samples did not find these biases suggesting
that over representation of males and minorities is an unintended,
innate consequence of school identification. These examples illustrate
that the (WORD?) criterion may not pass muster on several facets
of validation and provide important evidence with which to evaluate
the validity of classification criteria. The examples also emphasize
Messick's requirement that both intended and unintended consequences
be evaluated. I suggest the adoption of an expanded view because
Messick's views emphasize that both data and logic are required.
Because consequences of a classification need careful consideration,
and because he held the enterprise brings with it a host of value
and beliefs that go often unstated and unexamined. Analysis of
validity via the (WORD?) requires that values and beliefs be examined
as carefully as data.
DEBORAH SPEECE
Hypothesis testing is critical, but a view of validity that encompasses
and expands traditional approaches will be from the classification
systems that is more comprehensive and acceptable to the many
intended audiences. Admittedly, there is little work on construct
validity. But if we are to begin anew with consensus committees,
let's be sure to incorporate a comprehensive and contemporary
view of what is required to establish validity. My time is up,
but we made eight points. Thank you. Uh, my final theme sounds
a call for caution, and there are two points I want to make.
Uh, uh, there is action in Jack's paper, and I don't think he
should be, uh, solving all of our problems, but what is absent
is what should we do, but how should we replace discrepancy.
And I think we need to seriously consider that if we drop discrepancy
practitioners may then conclude that LD is not real. If we don't
have criteria, it must not exist. Now, they very clearly said
that they believe LD is a real phenomenon. But we have to consider
if we drop it, we need to replace it with something. And finally,
what we don't know, uh, what we don't know about the classification
of LD is, um, let me say this to the camera, what we know is stopped
by what we don't know. Okay? Most of our information, unfortunately,
within the area of reading with young children. Okay, so we
need to look at all the children as well. Thank you.
LINDA ELKSNIN
We respondents have to talk quickly. First of all, I think that
we really do need to thank, uh, Jack Fletcher and Reed Lion (SP?)
for re-energizing the field of learning disabilities, and forcing
us, uh, to really reconsider some of our practice. And in that
say, I will also say that, uh, it was difficult to respond to
this paper, because it kind of incorporated everything that was
included in the other eight white papers. So I've been somewhat
selective relative to what I want to talk about today. But I
think in the simplest terms, uh, Fletcher et. al. based on the
results of the NICHD research that focuses on early reading, have
concluded several things.
LINDA ELKSNIN
First of which, Jack talked about at great length earlier, and
that is that there's little difference between our low achieving
students and our learning disabled students. Uh, based upon that
they propose that we eliminate discrepancy and exclusion components
from the Federal definition identification criteria, and then
I think something that was not addressed early on that I'd like
to address and that is that school districts begin to offer services
using special education as a mechanism for so many students that
we consider to be low achievers as opposed to being learning disabled.
The reality is, at the present time we serve almost three million
students with learning disabilities. Many of these students represent
the lowest performing of low achievers, not necessarily of basic
reading skill, which has been addressed by the NICHD sponsored
studies, but also reading comprehension, mathematics calculation
and written expression. The other factor is that these students
are probably what Joe Torgesen refers to as treatment resisters.
They are students who require intensive intervention, even with
the NICHD research, Torgesen admits that there are between two
and six percent of students who will remain poor readers despite
NICHD fund of prevention intervention. That's something that
we need to keep in mind. The interventions required by the lowest
of the low achievers are not restricted to beginning reading instruction,
and in fact the greatest increases, I think, were, uh, mentioned
today by, uh, Pasternak, and that is that students within the
age ranges of 12 to 17 account for the largest number of students.
These are students that we need to focus our attention on, content
with (WORD?) to the use of longer strategies and through teacher
to provide advanced organization and explicit practice opportunities.
LINDA ELKSNIN
The other issue is I think we need to separate out, and I think
this was done earlier as well, the concept of LD versus how we've
operationalized the definition. Uh, some of the work that Lee
Swanson has done clearly indicates that even novices as well as
experts, have an implicit understanding of what a learning disability
is and what the characteristics are. The problems with the LD
classification stem from not how we've conceptualized LD, but
how it's been operationalized by practitioners. There are a couple
of implications, I think, that are, uh, important for practice.
The first has to be that we do, we do some reliance and discrepancy
formulas per se. That those are not going to be productive in
terms of getting us a great understanding of how to classify and
understand the population of one of the disabled students that
we come and see, serve. That we need to get eligibility teams
in schools, the flexibility to use proactive professional judgement.
I know that's another topic of one of the Wight papers during
this LD summit. Thirdly, we need to focus on effective cognitive
skills, which may be a difficult undertaking in how these skills
contribute to how students learn to read right and to do mathematics.
not just our students with LD, but also our low achieving students
as well.
LINDA ELKSNIN
And then finally, when it come to the critical judgement piece,
I think we've kind of lost track of the fact that if we look at
other professionals for people in the area of mental health, we
need to acknowledge the importance of clinical judgement in the
classification process just as psychiatrists do when you look
at a DSM four. In other words, if you look at the DSM it talks
about the idea that the diagnostic criteria that are included
are meant to serve as sidelines to the, to be informed by clinical
judgement and not meant to be used in a cookbook fashion, which
is what happens when we rely exclusively on discrepancy formula,
and IQ involved in discrepancy. Fletcher and his colleagues also
talk a lot about response to treatment, and I think this is probably
one of the more positive aspects of the paper. But the issue
of developing a definition that's treatment oriented I think it's
not one that we're gonna easily be able to accommodate. They
argue that alternative definitions of learning disabilities quote
should provide specific criteria indicating that the child has
a particular type of LD, and that they point towards a set of
potential interventions. However, the fact that the Federal definition
of LD fails to prescribe treatment is irrelevant. Uh, very few
definitions or (WORD?) patient's systems include treatment recommendations,
and that might include all the other IDA disability definitions,
as well as none educational classification systems, such as the
DSM Four. Rather than treatment intervention of begin the focus
of definitions, our definition, I know Nancy Larson will address
this, are used to really make entitlement decisions including
whether or not there's a presence of, of disability and whether
or not the student is eligible for special education services.
Obviously, we're gonna need different assessment information
for post entitlement decisions, such as intervention planning.
And the cost of a learning disability, for example, the individuals
inability to store, organize, acquire, retrieve, express or manipulate
information will go a long way in helping us to determine the
intervention approach. Several implications for research. The
first thing is that, again, we need to reduce the efforts to validate
the discrepancy formulas as this is unlikely to lead to specification
of recognizable, learning this disability's behavior. The second
thing is that, in my mind, when you can increase our efforts to
identify the cogent processes that mediate discrepancies and non-discrepancies
among poorly performing students as a whole, including students
with learning disability.
LINDA ELKSNIN
And finally, probably the most exciting thing, and that's was
mentioned by Deb earlier, is that we need to focus on how students
respond to intervention, and have that be part of the assessment
process. We do have an arsenal of powerful strategies and methods
to teach students with learning disabilities. Some of the work
that Lee Swanson has done clearly suggests that direct instruction,
cogitative strategies instruction work with our students. But
there is a big gap between what we know works and what is used
out in the schools. So another purpose would be to examine the
ways that we can support and sustain effective instructional practices
in the schools, which is something that (SOUNDS LIKE) Wes Gerston
has, um, examined a recent issue of LD Research And Practice.
And finally, one of the problems we had in interpreting LD research
is that the subject's are inconsistently and inaccurately described.
If we select LD subjects strictly on the basis of public school
identification criteria, that is just a sample of convenience,
we really get a post hoc interpretation of subject characteristics
and past performance. And Federal agencies could go a long way
in directing this issue by requiring minimum standard, uh, universal
protocol for describing participants in Federally funded LD research.
And this has been suggested by some of Jack's colleagues, (UNINTELLIGIBLE)
Council For Learning Disability and goes as far back as Barbara
Keos (SP?), (UNINTELLIGIBLE) marker variables when we present
our research.
LINDA ELKSNIN
In some ways, radical changes in ways in which we identify and
classify students with learning disabilities appears to be unwarranted
at this time. The problems that we have, really, are due to the
operationalization of the definition, not to the definition itself.
The results, uh, Jack talked a lot about the unexpected underachievement
and IQ achievement discrepancy. It really depends upon which
body of research you look at. There are some studies that do
support that concept. So it is equivocal at this point. It only
focuses on reading for the most part, and you had mentioned some
of the other areas, by the emphasis has been on reading, particularly
early reading. And then our students with LD are going to require
intensive intervention because they often are the treatment resistors
that, um, we need to find better ways of serving in either special
or general education. Uh, and, Renee says I can do one more,
so that's not to say that we can't improve practice by discouraging
reliance and over reliance on formulas and cut off scores, by
trying to determine the cognitive processes that mediate performance
and to, uh, examine response to intervention and offer ways to
insure that the good work that NICHD and others are doing are
getting into the schools and teachers are actually using (WORD?)
intervention strategy. (APPLAUSE)
NANCY LARSON
Hi. I've had a lot of people with a very technical orientation
precede me, and I'm want you to know that I'm very honored to
be here and humbled by the presence of some of the people who
have done so much research over the course of time. But I want
to talk to you about remodeling. I'm from Minnesota, can you
tell? I have spent the better portion of this last year remodeling
and I got to thinking about this presentation, wait a minute,
uh-oh. And after thinking about the remodeling of LD, something
that has gone on ever since I entered the process, and I got to
thinking about building (UNINTELLIGIBLE) an analogy, and certainly
the preponderance of the discussion has occurred around researchers
and policy makers, but largely researchers. The policy, actually,
has remained the same since 94-142, and the two iterations of
idea, but the researchers constantly send us new questions and
of course that's very healthy. The, the team members that are
not present ever consistently in this practice seems to me to
be the teachers. So the practitioners are left out of this process
all together. And I think that any kind of, uh, discussion about
reinventing LD in any capacity really needs to include the, the
field. The field is suffering, the field needs to be part of
this, a part of this discussion. Of course, we have stages of
building I discovered, in my little remodeling session. Over
the course of the year we have the foundation in rock, carpentry,
plumbing and electrical, finished carpentry, paying the contractor
and moving in. We have yet to move in, by the way. The foundation
for this discussion is entitlement. And any change in classification,
any tweaking in definition automatically will result in the field
into changing entitlement.
NANCY LARSON
That most directly affects parents, students and teachers. It
very little affects the people who are in this room or myself
as a policy person. But it directly affects large groups of students.
I think that that's the essential thing to keep in mind. That's
the thing that was not really mentioned. That's an element that
wasn't really mentioned in this lecture study. And it is of tremendous
concern. The other accusation over the course of time is that
LD, and everyone's mentioned this, is not real, that it's a fabrication
and a construct in the minds of researchers or theorists, and
that always brings us to the issue of when is a disability not
a, not a disability. In other words, when is something funded
as a disability and that's where the field as a whole goes is
directly to pragmatic matters. So if LD is in fact watered down,
altered or changed, and we have this resultant change in entitlement,
we will have quite an uproar in the field and we need to be prepared,
as my colleagues pointed out, for the, for the implicit changes
that will occur if there are changes in definition or classification.
For all other disability areas, um, criteria is a state function.
For learning disabilities it's a Federal function. Why? I'm
not too sure. But when we talk about a classification, we are
talking about two elements, one definitional and the other is
criterion oriented. And actually, there is a third element that
pops into most state interpretations of Federal rules and that
would be evaluation methodology. So we really have three pieces
in the continuum. Rough carpentry, and I know why they call them
rough. They immediately came to my house, ripped off all of the
siding and then we had a monsoon in Minnesota, so it was a very
exciting. (UNINTELLIGIBLE) as a court issue in LD, and it is,
um, very much part of our definitional consideration. And the
other piece that Fletcher et. al. alluded to is that kids identified
under the discrepancy model and formula, are too, kids with disabilities.
And it is not a mild disability. And we do discuss this constantly
as if it were a mild disability.
NANCY LARSON
And the issue around determining characteristics is to determine
what those characteristics are throughout the life span of the
individuals with a learning disability. I am bothered, this is
third point, by curriculum based disability definitions, meaning
that for all other disabilities we have a notion or a core notion
of what that disability is. For learning disabilities we want
to break it out into reading disability and math disability and
so on disability. It is a problem for the field. You cannot
both have a curricular definition and a philosophical definition
that is at a core concept. It is very confusing as it plays out
in application in the field. Is a disability to do with reading?
Well, no. Because 13 other categories tell us differently.
They tell us that we have to have a sever effect on curriculum
in order to provide special education and related services. So
that is another consideration. And the next, are we going to
continue, now, just a second. I'm changing to my final in electrical.
Uh, to continue with our roof use of terminology, which is, it
is very important to have precise terminology when, when moving
the rule or law or notion to the field. The application is all
involved in the terminology. So, if we use, I'm gonna talk just
a second about disempowering language, but if we use terms like,
reading disability to mean disability, and what is autism? And
what is MMI? And what is anything else that is a disability under
Federal rule. Okay, that seems to mean more of a curricular disability.
The disempowering language piece is all about, uh, calling things
reading retardation and I realize that the studies in Fletcher
et. al. use these, these pieces, these terms.
NANCY LARSON
Mental deficiency. Those are terms that will instantaneously
alienate a large segment of the population that we serve. So
be careful about disempowering language, I think, also is, is
very essential. And the interchangeable, uh, terminology piece
is really around reading disability, dyslexia and learning disability.
And they are used in this paper interchangeably and it does confuse
the field when we have more than one term for the same classification.
I'm going to skip underachievement since we already had that
discussion. Never mind. And I'm going to talk about The Matthew
Effect. The Matthew Effect goes something like this, for teachers
in the field. If teachers right this minute cannot implement,
uh, idea 976, consider what any tweaking to the definition will
do, will have a cumulative effect in terms of how they are able
to operationalize what ever it is that has developed at the Federal
level. So eventually they'll have a cumulative negative effect
field application. This is the meat and potatoes of my existence.
It is, it takes an incredibly long period of time to trickle
down to the masses whatever changes are going on, especially in
research. There is no time out there, I'm here to tell you, there
is no time, no one has time to read the research, and if they
do, they give it a cursory effect. It's not that they're not
interested, it is that they have no time. And the next piece
that anyone will ask you is, gee, to whose benefit is it to tweak
classification systems to reconceptualize, to draw new charts
in learning disabilities. Does it help the teacher who's actually
doing the teaching? Does it help clarify pieces for parents as
they operationalize a learning disability. And the last question
is just a question I've wondered about for a long, long time.
Is uniformity in definition useful, or not useful? In other
words, is it useful to have a uniform definition across states,
when in reality in lots of other policy pieces we, uh, are very
much promoting states individual, uh, individual rights to determine
their own fate, if you will. So that would be another question.
Does that actually add to our field or subtract from our field?
NANCY LARSON
The gist of this issue for me is the third bullet there, which
is give me the money. If we are shifting resources, general ed
wants those resources. Uh, and so does everybody else. They
are, uh, we're in an age where there's lessening public support
for public education. It is a real problem when you change entitlements,
you change the money flow. And everybody's going to want to be
in the give me the money business. So I think that's the final
consideration I've got for today. Um, I have a few more slides,
but we'll just stop here so everyone can have their, their question
and answer session. Thank you.
[END OF TAPE]