improving comprehension online project, 2005-08 designing and testing a universally designed...
TRANSCRIPT
Improving Comprehension Online Project, 2005-08
Designing and testing a universally designed strategic digital literacy environment for diverse learners
Bridget Dalton, Vanderbilt University
Patrick Proctor, Boston College
IES Research Conference
Washington, DC ~ June 11, 2008
A goal 2 development award
to CAST, Inc.
Research team• Bridget Dalton (Co-PI), Vanderbilt
University, and Elaine Mo, Kristin Robinson, Ge Vue, Mary O’Malley, & Boris Goldowski, CAST, Inc.
• Patrick Proctor (Co-PI), Yi-Chien Li, & Kevin O’Connor, Boston College
• Catherine Snow (Co-PI), Paola Uccelli, Sabina Neugebauer, Lorena Landeo Schenone, Harvard Graduate School of Education
• School partners: 3 semi-urban and 1 urban school in northeastern Massachusetts
Project goal
• To develop and test a universally designed
(Rose & Meyer, 2002) strategic digital reading approach (Dalton & Proctor, 2007) to improving reading achievement of 5th grade students, including bilingual students and struggling readers
Multiple perspectives required
Vocabulary
Digital literacy
environmentsUniversal
Design for Learning
New literacies
Bilingualism2LA
Reading comprehension
icon
Universal design for learning (Rose & Meyer, 2002)
Design for the broadest range of learners from the beginning; avoid retrofitting
Provide multiple means of • Representation• Expression• Engagement
Assumptions
• Shift to universal design perspective – Attention to diversity and individual difference
benefits individual and society
• New literacies, while more complex, are more flexible and inclusive– potential to level the playing field for those
who have not fared well with print literacy
Rand Reading Study Group’s (2002) reading comprehension heuristic
text reader
activity
Sociocultural context
comprehension
text reader
activity
Sociocultural context
comprehension
Strategic Digital Reading (Dalton & Proctor, 2007)
textreader
activity
Sociocultural context
comprehension
Comprehension in a new literacies landscape: Strategic digital reading
text
reader
activity
Sociocultural context
comprehension
How does ICON support diverse learners in relation to…
• Representation?
• Expression?
• Engagement?
• What is unique for ELLs?
• What features/supports are essential for some; good for many/all?
Iterative design, formative feedback and testing
Y1. Develop
Vocabulary
Y2. Compare Vocabulary, Strategies & Combo
Y3. Compare Combo
Vs. Control
Y1. Develop
Vocabulary
Y2. Compare Vocabulary, Strategies & Combo Versions
Y3. Compare Combo
Vs. Control
ICON optimal prototype (Yr. 3)
Embedded Strategies
Spanish language support
Coaches
Level 1 coaches provide text-specific models and think alouds. As skill increases, students select strategies and coaches provide generic think-alouds.
Vocabulary: Connect It!
Vocabulary: Language Alert
More than 60% of the power words are Spanish English cognates
Vocabulary: Web It!
Vocabulary: Caption It!
All 3 years/studies: Feasibility, appeal & usability
• Teachers and students view ICON as a helpful reading tool, easy to use, & engaging
• Technical support required; bandwidth issue• Variation in teacher enactment of ICON suggests need for
additional study• English proficiency levels influence ways in which
students use ICON and extent to which additional support is needed– Peer collaboration one means of support
• Increased sensitivity to learner (needs, use of system, performance) is likely to benefit all
Y1 Study of Semantic Depth• 35 students, 24 bilinguals (Spanish and other low-
incidence languages), 11 English monolinguals
• Oral language skills (WJ picture vocab+listening comprehension)
• Reading skills (WJ passage comp + MCAS ELA score)
• Average semantic depth score for 8 target words (Anxiously, Bitter, Dense, Grasp, Ignore, Menacing, Powerless, Relieved)
Yr. 1 study of semantic depth(Proctor, Uccelli, Dalton, & Snow, in press)
Effective teaching and learning activities targeted for further analysis: Caption-It
Semantic depth was a significant predictor of reading performance, mediated by interaction
with English oral language proficiency
Figure 1. Effect of semantic depth predicting reading latent variable as a function of oral language proficiency
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
0 1 2 3 4 5
Semantic Depth Score
Re
ad
ing
La
ten
t S
co
re
10% Oral Lg
50% Oral Lg
90% Oral Lg
*No effect for language status (bilingual vs. monolingual)
What did we learn from Y1 vocabulary study?
Caption It: Encouraging but preliminary – As both an activity and an assessment, it appeared to reduce the monolingual/bilingual gap in students’ performance, though oral language was heavily implicated
Semantic Depth: Promising but far from final Semantic depth showed a positive association with reading
comprehension, beyond the contribution of decoding and oral language skills
Semantic depth seems to play a more prominent role as oral language skills improve
Semantic depth refers to a cluster of skills: associations among these and with other dimensions of vocabulary depth need to be explored.
Year 2 study: Strategies vs. vocabulary vs. combo
• Very hard to find research that compares effects of vocabulary versus comprehension instruction
– Likely because the two are so strongly intertwined
• Given that vocabulary is a primary focus of instruction
for ELL students, we found this question intriguing, and asked 2 basic questions:
1. Does assignment to condition (Vocabulary-only, Strategy-only, Combo) affect students’ performance on standardized and researcher-developed measures of vocabulary and comprehension?
2. Do the effects vary by language status (monolingual, Spanish-English bilingual, other bilingual)?
Our hypotheses
For standardized and researcher-developed
vocabulary:
Combo > Vocabulary > Strategy
For standardized and researcher-developed
comprehension:
Combo > Strategy > Vocabulary
Y2 study of vocab vs strategies: What matters and for whom?
(Dalton, Proctor, Uccelli, Mo, & Snow, in preparation)
• 106 students, 21 Spanish-English bilinguals, 17 other-English bilinguals, 68 monolinguals in 6 classrooms, 3 districts
• Random assignment to condition (vocabulary, strategy, combination)
• 14-week intervention• Pre-post standardized testing,
embedded vocabulary and comprehension quizzes
Y2 effect size overview
What did we learn from Y2 study of vocabulary vs. reading
strategies vs. combined?
• Overall, hypotheses held, and combination version showed strongest results across standardized and researcher measures
• Did the small sample size mask learner by treatment interactions?
• Theoretically, an interaction between student characteristics (reader type and/or language status) would make sense
– Sample size may be too small– Student controls access to support and may not be
making good decisions about when and how to use support
– Thresholds of language proficiency
Year 3 study: Combo vs controlHaving established general effectiveness, time to move toa comparison between treatment and control using optimalversion of ICON
• Quasi-experimental study
• 12 classrooms, classrooms assigned randomly to treatment or control condition, n = 227 (108 control, 119 intervention; 10.5% other bilinguals, 48.5% Spanish-English bilinguals, 41% English monolinguals)
• For intervention group, 2 x 50 minutes per week, for 16 weeks
• For control group, across the three districts, standard literacy curriculum included reading strategies focus, but limited vocabulary instruction
• Initial training of teachers and students by research team, gradual release of ICON prototype teaching responsibility
• Analyses conducted at student level, randomization at teacher level
Y3 measures
• Gates-MacGinitie reading vocabulary and comprehension subtests pre- and post-intervention
• Aprenda reading vocabulary - pre• Researcher-designed breadth of vocabulary
(targeted words) post-intervention– 20-item multiple choice assessment
• Reseacher-designed depth of vocabulary, post-intervention– 5-item definition, drawing, & captioning assessment
Y3 results
General results:• No effect of condition on standardized
measures; significant voc and comp gain for both groups
• Strong effect of condition on researcher developed measures
Condition Depth/BreadthStandard
Vocab and comp
Condition effects on ICON vocabulary breadth
0
10
20
30
40
50
60
70
80
Mono BilSp BilOth
Control
Experimental
Significant effect of condition on ICON voc. Breadth F(1,205) = 56.62, p < .001
Significant difference between Spanish bilinguals and English monolinguals (t = 5.1, p < .001)
Strong readers significantly outperform average (t = 5.1, p < .001) and struggling (t = 12.6, p < .001)
No interactions by language or reader status and condition
Condition effects on ICON vocabulary depth
Exp. significantly outperform Control on vocabulary depth
F(1,224) = 101.4, p < .001
English monolinguals significantly outperform Spanish bilinguals (t = 5.3, p < .001) and non-Spanish bilinguals (t = 2.2, p < .05)
Strong readers significantly outperform average (t = 5.0, p < .001) and struggling (t = 9.4, p < .001) readers
No interactions by language status, BUT average-reader X condition interaction (p = .048)
Spanish-English bilinguals and ICON depth of vocabulary
0
1
2
3
4
5
6
7
8
400 500 600 700
Aprenda Vocabulary Score
Su
m E
ng
lish
Vo
ca
bu
lary
De
pth
S
co
re control vocdepth
intervention vocdepth
For intervention Spanish-English bilinguals, Spanish Vocabulary scores explained English Vocabulary Depth performance, after controlling for condition and prior English proficiency.
Conclusions and next steps
• Continue to analyze Year 3 data– Worklog responses, multimodal retellings, student
feature use, teacher use of feedback support.
• For whom does this intervention work best? – Goal 3: effects for Spanish-English bilinguals and
struggling readers are intriguing– Goal 2: Work for transfer. Design for increased
sensitivity to learner characteristics, especially language proficiency.
• Distal effects on standardized measures