Individual Differences in Children`s Comprehension Skill

advertisement
Individual Differences in Children’s Comprehension Skill: Toward an Integrated Model
Oakhill, J., Cain, K. & Yuill, N.
1) Purpose of This Study:
 To determine what causes poor reading comprehension?
 To determine what causes poor comprehension in readers with good
decoding skills?
Test Group: Pure Poor Comprehenders Wanted: In this study, students had equal
skill at the word level including decoding and naming speed and vocabulary. They
also controlled for phonological processing ability. Yet, they still found
comprehension differences. Children in this study had specific comprehension
problems.
2) Processes
3) Lexical and Synthetic Processes:
4) Perfetti (1988): The Verbal Efficiency Theory. Chart (p 27 / Carlisle)
5) Lower level processes, such as lexical access and parsing, become
automatic in proficient readers. When they are not automatic or efficient,
lower level process use up attentional resources that are needed for higher
level (comprehension) processes.
5) Slower naming speed correlated with poor comprehension in 3rd and 5th
grade students (Perfetti and Hogaboam, 1975) (p 345). Poor
comprehenders had longer naming times for even very common words. (p.
346)
4) Sentence Level Issues:
5) Poor comprehenders may have difficulties at the sentence level by failing to
understand certain syntactic constraints. (346)
5) Cromer et. al (1971): Poor comprehenders tend to read word by word in
list-like fashion. They did not use the sentence and text structure to guide
their comprehension.
3) Inference Making:
4) One persistent finding is that poor comprehenders are not skilled at making
inferences. (347) Good comprehenders are more likely to make integrative
inferences and to bring together the ideas in a text as a whole. (348)
4) Lack of background and relevant knowledge is challenged as a significant
factor affecting inferencing ability in younger readers.(see Domain K).
4) PC may not recognize that an inference is needed or permitted.
4) Limited processing ability in Working Memory may interfere with the ability
to access relevant knowledge, and to store and integrate text.
3) Understanding Text Structure
4) Measured By: Conventional Story Structure: 4 Story conventions; openings,
endings, character/setting, scene setting. Quality of Event Structure: 3 types
“to capture the extent to which causality between events was signaled and the
story as a whole cohered.
4) Results:
5) PC produced less well internally structured stories, especially with topic
rather than picture prompts. (Table 18.3, p. 352), even though all groups
demonstrated use of ‘conventional story structure’ as defined above..
5) The ability to produce a well structured story is not just due to reading
experience. The older PC told stories with weaker internal structures than
younger, comprehension matched students (who being younger probably
have had less exposure to books). Therefore, knowledge of internal story
structure may be causally linked to comprehension skill (352).
3) Comprehension Strategies / Monitoring for Comprehension
4) Anomaly Resolution Task: PC were equally good as GC in resolving
neighboring anomalies but significantly worse when the anomaly was
separated in the text (Table 18.4, p. 354).
4) Error Detection Task: PC were more affected by the distance manipulation than
were GC (Table 18.6, p 355).
3) Working Memory Limitations
4) WM as a Cause of PC: PC poorly integrate information from different parts of
the text (disconnected text base) possibly due to deficient working memory.
GC “are better at making inferences, monitoring their understanding,
interpreting anaphors, and deriving the structure of stories because they
have more efficient working memories…PC’s more limited WM are
sufficient to support some local integration of text, but that they cannot
support the integration of the text as a whole.” (357).
5) Weak WM might be due to weak phonological coding. Test results are
mixed and incomplete. Authors found PC had weaker working
memories than GC, but their test may not have measured WM as they
said it did (P. 357).
4) Problems with this Theory:
5) Poor Wm may be caused by poor comprehension rather than being the
cause of poor comprehension.
5) WM can’t be the only cause of PC as training in the need to make
inferences were successful, and it’s doubtful that they were successful
because such instruction improved WM capacity. An alternative
explanation is that inference making training allowed PC to bypass
weak WM problems, or to use their limited WM more efficiently.
2) Knowledge
3) Word Meanings
4) Link between Vocabulary Knowledge and Comprehension
5) In most cases vocabulary size in children is a good indicator of reading
comprehension skill, perhaps because both knowledge of word meanings
and comprehension depend to some extent on general linguistic skills.
(358)
5) Texts are easier to recall if children know the meanings of the words in
them. Beck, Perfetti, Mckeown (1982).
4) Learning-from-context Hypothesis: Daneman (1988). Vocabulary and
comprehension are correlated because vocabulary tests and
comprehension tests both reflect the individual’s ability to learn or acquire
new information from context.
5) In this study the poor comprehenders had normal vocabulary ability
(though the assessment they used may have been flawed). It’s hard to
control for this factor since vocabulary and comprehension are intimately
interrelated.
4) Additional Findings: Other research has shown limited effects of vocabulary
instruction. Weak phonological processing (storage) may account for
limited vocabulary.
4) Reading Experience and Vocabulary:
5) There is some convincing evidence that those who read avidly tend to
improve their vocabularies. For 5th graders: time spent reading at home X
reading rate = # of word read per. Year. A child at the 90%ile would come
across nearly 6.5 times as many words per. year as a child at the 50%ile
(1,823,000 vs. 282,000 words). (Anderson et al.)
5) The “Title Recognition Test” accounted for a significant proportion of the
unique vocabulary knowledge in 4th and 5th grades, even when age, and
Performance IQ (Rave’s matrices) were factored out. (Stanovich, 1991)
5) HiPrint/LoComp vs. LoPrint/HiComp Comparison: The correlation
between comprehension skill and print exposure is not perfect. (Stanovich,
1991) HiPrint/LoComp still performed better than the LoPrint/HiComp
group on the PPVT.
This study provides striking evidence that it might be possible
for print exposure to compensate for modest levels of
cognitive ability. (360)
3) Domain Knowledge
4) Evidence to Support the Importance of Domain Knowledge in Comprehension:
Spider reading: “The extent and quality of their knowledge determined
how well the texts were understood.” (p. 361) BG knowledge instruction
facilitated better comprehension for intermediate and middle school
students in studies cited here from the early 80s (ex. McKeown (1982)).
4) Evidence to Show Limits of Domain Knowledge as a factor: Barnes and
Dennis (1996): Devised a make believe world and taught a set of “facts”
to mastery. PC still were weaker at drawing inferences from their
knowledge base than were GC. Also/ Perfetti: General language
comprehension apart from any domain knowledge.
4) Opposite Link/ GC leads to Domain Knowledge: GC read more and get more
meaning from what they read by building a solid text base, and by making
more associations, inferences, and elaborations.
4) Text Structure: Knowledge of text structure improves comprehension. “Story
grammar rules may be used to guide the construction of a text
macrostructure: An incomplete grammar would therefore restrict both
planning of and processing of stories” Stein and Trabasso (1981) (p. 363)
2) Conclusions
3) Working Memory constrains the comprehension of children by limiting their
ability to build mental models of texts due to:
 Too much attentional energy on lower-level decoding processes, and not
enough capacity.
 Generally inefficient WM (possibly due to poor phonological coding), that
can be addressed using strategies to help WM operate more efficiently.
 Deficient WM is not likely to be a full cause for poor comprehension
3) Need For Later Studies
 Longitudinal studies are needed to determine 1) if measures at an earlier
age will serve as predictors for RC issues at an older age, and 2) to
determine causal relations between factors.
Download