Comprehension Deficits: Prior
Knowledge vs. Computational Deficits
Children with reading disability (RD) often show deficits not only in reading
comprehension but also in listening comprehension. The question we are
trying to answer is what underlies these listening comprehension deficits?
We know that children with RD often have less topic knowledge than non-affected
children because they usually do not read very much. So, one possibility is
that differences in listening comprehension skills reflect differences between
RD and non-RD children in the amount and quality of the knowledge that the
child brings to bear on interpretation of a passage. Alternatively, listening
comprehension differences in children with RD may reflect a general linguistic
deficit that affects not only their ability to learn to read but also to
process oral language.
The goal of our work is to determine if differences in knowledge, differences
in computational processes or both underlie differences between RD and non-RD
children in listening comprehension. Our approach is to assess listening
comprehension skills while controlling for knowledge differences between RD and
non-RD children by having all subjects learn a new knowledge base to criterion
before listening to passages that depend on that knowledge and having their
comprehension assessed.
Comprehension Monitoring
Comprehension monitoring refers to the ability to determine whether you
understand a passage or whether the passage makes sense to you. Deficits in
comprehension monitoring can be one of the main reasons for failing to comprehend.
If a person is unaware that a passage does not make sense, it suggests that
their attention is not directed to comprehension, perhaps because they are
focused on other matters such as word decoding (in the case of poor readers)
or their own thoughts.
We have developed a short test for assessing comprehension monitoring that
works well in children, even those with reading disability and attention
deficit disorder. We are examining the extent to which problems in monitoring
are responsible for comprehension failures in these children.
Computational Processes in
Word Recognition
A central question in reading research concerns the computational architecture
of the lexicon and whether it differs for those with reading disability or
dyslexia. We are using combinatorial priming in a variety of word
recognition tasks to investigate this. We have developed primes that are
related to their targets on multiple dimensions. Consider, for example, the
target word, BOAT. Combinatorial priming asks whether a word that combines
both phonological/graphemic and semantic relatedness, such as FLOAT, produces
more priming than a word related on only one of these dimensions, e.g., GOAT
or SHIP? The central question is how activation combines across dimensions
because the nature of this combination reveals whether the semantic and
phonological/graphemic processing units are functioning independently or
interactively.
We are examining these effects in both skilled and dyslexic readers to
understand the nature of computational processes of information integration
in the lexicon.
Genetic Factors in
Comprehension
Since 2001, we have been examining comprehension skills in twins, both
monozygotic and dizygotic, to examine the degree of similarity in their
performance on reading and listening tasks.
The project is funded by the National Institutes of Health and is being
conducted in collaboration with the Institute for Behavioral Genetics as
part of their Colorado Learning Disabilities Research Center. The goal
is to use quantitative genetic techniques for determining the etiology of
comprehension problems in terms of environmental and genetic factors.
Those aspects of comprehension that show high genetic loadings are then
used by other members of the Center to search the genome for genes
contributing to reading and language problems.
Individual Differences
in the Component Skills of Comprehension
There is a psychometric approach to predicting reading comprehension that
says that comprehension performance is determined by the person’s word
decoding skill and their IQ. Our twin study of comprehension skills has
shown, however, that there is more to it. Our analyses of both the
phenotype and the genotype have shown that there is independent variance
in reading comprehension accounted for by listening comprehension, even
after factoring out IQ and word decoding. This finding points to the
need to think about individual differences in comprehension in terms of
the component skills of comprehension.
We are assessing children ages 8 - 18 on a range of cognitive skills
from low-level lexical skills to higher-level processes such as
inferencing and comprehension monitoring to determine how profiles on
these skills predict comprehension performance.
Language Comprehension in
Fragile X
Fragile X is a developmental disorder caused by a mutation on the
X-chromosome that results in retardation in males and can lead to
learning disabilities or retardation in females. In patients with
Fragile X, a gene, a string of three coding molecules that repeat
six to 50 times, is over expanded. The misprint can include 200 to
more than 1,000 copies of this particular "triplet" sequence
(cytosine-guanine-guanine or CGG). The excess muddles the instructions
needed for the creation of a specific protein. The result? The level
of fragile X protein in a person with the syndrome is low or
nonexistent - the lower the level, the greater the mental impairment.
Recently, scientists discovered that the fragile X protein is vital
for developmental processes and may be involved with learning and memory.
Fragile X syndrome affects approximately one out of every 2,000 males
and one out of every 4,000 females.
Women with Fragile X can produce a range of the protein depending on
the extent to which their good X-chromosome is suppressing their
fragile X. Women with Fragile X also show a range of cognitive deficits.
We are attempting to relate molecular measures of the syndrome to their
cognitive deficits. We are focusing on higher-level discourse skills,
such as the ability to appreciate the coherence of discourse and the ability
to draw inferences. To date, we have found that one aspect of language
processing, affected by Fragile X, is the ability to revise an initial wrong
interpretation. This underlies problems in both coherence and inferencing.
We are also examining how the social or pragmatic aspects of language are
affected.
This research on Fragile X is representative of my general interest in
neuropsychological populations showing communicative deficits.
Latent Semantic Analysis
Latent Semantic Analysis (LSA) is a computer-based system for representing
the meanings of words as vectors in a high-dimension semantic space.
(Landauer, T.K., Foltz, P.W., Laham, D. (1998). Introduction to latent
semantic analysis. Discourse Processes, 25, 259-284.) The vector
values are derived from statistics of each word’s occurrence across many
different contexts. The meaning of a whole passage can also be represented
in this system; the meaning of a passage is represented by averaging the
vectors of its component words in this space.
We are exploring the validity of using LSA to score children’s retellings
of passages as a measure of comprehension. See the web site for LSA:
http://lsa.colorado.edu.
Reading vs. Listening in
Children with Reading Disability
The reason we examine both reading and listening comprehension is to
determine whether children with reading disability (RD) show deficits only
in reading comprehension, as would be expected from their difficulties in
word decoding, or if they also have difficulties in listening comprehension.
If they show deficits relative to non-RD children in listening comprehension,
this would indicate a broader language deficit than just reading problems.
Then the questions become what component processes are deficient in both
reading and listening, are there core deficits that are the same for all
RD children, or are there subtypes with different deficit profiles?
|