o

University of Pittsburgh

LEARNING RESEARCH AND DEVELOPMENT CENTER

17 October 1977

Prof. Joshua Lederberg

Dept. of Genetics, $331

Stanford University Medical Center
Stanford, CA 94305

Dear Prof. Lederberg:

This letter is an amplification of certain aspects of the proposal
that James Greeno and I made to SUMEX-AIM on 29 July 1977. Its purpose is
to indicate the extent to which we are proposing research that will involve
NIH target populations. I will briefly comment on four aspects of the
proposed work that relate to this issue: (a) the overall nature of the
research projects in which the proposed work is embedded; (b) the general
adaptability of the simulation medium we are proposing to use (ACT) for
studying learning deficits; (c)specific aspects of the proposed reading
research that relate to impaired learners and children who are not learning
well in school; and (d) similar aspects of the proposed work in
mathematics.

LRDC research goals. The work at LRDC of which our efforts are a part
is directed at improvement of elementary school instruction. The work has
been aimed, for the most part, at developing the understanding necessary
for the construction of a truly adaptive curriculum in the basic skills
areas. In principle, much of the work in which we are engaged is not
specifically addressed to the problems of those who have learning deficits.
However, the nature and cost of the approaches that have been developed at
LRDC have made them most often used with students who are hard to teach,
i.e., those in urban public schools and, in recent years, those in
learning-disability classes. In particular, the adaptive or individualized
approach is well suited to current demands for mainstreaming of children
who are, in one sense or another, handicapped. Thus, the research we are
proposing will be embedded in an overall research and development effort
which has above-average promise of affecting children with learning
deficits.

 

suitability of ACT simulation for this work. Another general point to
be made is that the simulation language we will be using, Anderson's ACT
system, is well suited to the exploration of the effects of certain types
of cognitive deficits on academic performances. The parameters that
control overall function of ACT are directly tied to concepts that are
often discussed in theories of learning disability. For example, it is
possible in ACT to directly represent differences in short-term memory by
adjusting a single parameter. Through a combination of adjustments in the
system parameters and in weightings given to individual productions (the

 

LRDC BLDG., 3939 O’HARA ST., PITTSBURGH, PA. 15260
Prof. Joshua Lederberg Page 2

micro-units of the simulation), it is possible to simulate differing
conceptions of the attentional deficits that are often claimed to be
present in learning-disabled children. Further, the use of ACT allows a
clear separation between system dysfunctions and lack of cognitive skills.

The proposed reading work. As discussed in the proposal, the
simulation work I propose to carry out on reading processes involves three
basic steps. First, the language coding deficits of slow and otherwise
poor readers will be isolated and specified in terms of an ACT model of
word recognition. Second, a model will be built of the processes whereby
sentences that are read are related to what has already been understood
from earlier parts of a discourse. Finally, both the word recognition and
the understanding processes will be combined into a single simulation. In
this way, we hope to generate predictions of some of the specific
understanding problems that can arise when language code processing
functions are inadequate. the primary group to be modeled is the pool of
children who, while showing no hard or soft signs of medical impairment,
still are not doing well in school. However, we are also interested in the
learning-disability population. Indeed, as part of the non-simulation
component of my research, I am currently comparing, with Isabel Beck,
another LRDC researcher, LD children's oral reading skills to those of
"normal" children of different ability levels. Thus, by the time the
simulations have been designed, we will have reading performance data on
both LD and unlabelled slow readers. It will be interesting to see whether
adjustment of ACT's parameters is necessary to capture the reading
behaviors of the LD kids or whether they are modelled sufficiently by
variation in the productions they are thought to have learned, or how well
they have learned them.

The proposed arithmetic work. We are still at the stage of developing
models of the basic processes of comprehending and using arithmetic
operations. When we have made some progress with that task, it will become
possible to begin to differentiate levels of arithmetic proficiency and
their causes. This is a goal with considerable appeal to Greeno and his
research group, and we expect that comparative studies to provide analyses
of sources of learning difficulty will be included in the program as
quickly as they become theoretically feasible. One relatively imminent
possibility involves a model of comprehension states in the understanding
of word problems that Greeno is developing in collaboration with Mary
Riley, a graduate assistant on the project. It is anticipated that this
model will identify processes associated with syntactic and semantic
components of the comprehension processes. If preliminary work with this
model is successful, we anticipate developing a study in which parameters
of the model would be related to students' skills in various dfomains,
including reading and arithmetic achievement.

I hope the above has clarified the issue that your advisory committee
raised. If you need further information of any sort, please feel free to
contact me by phone, mail, or via the GREENO SUMEX-AIM account.

Sincerely,

(le. Wi

Alan M. Lesgold