Presentation on theme: "The deep relations between speech and reading Mark S. Seidenberg Department of Psychology University of Wisconsin-Madison November 3, 2011 Society for."— Presentation transcript:
The deep relations between speech and reading Mark S. Seidenberg Department of Psychology University of Wisconsin-Madison November 3, 2011 Society for Language Development Boston
If the science is so good, why is reading achievement so poor? [preceding talks by Charles Perfetti and Rebecca Treiman]
OECD PISA Reading Results 2009 Wall St. Journal, 8/12/2010
NAEP Reading Results 4th grade
NAEP Reading Results 8th grade
National Assessment of Adult Literacy (2003)
2011 (downloadable from National Academies) (I am one of multiple authors)
Many reasons. What are the major ones? Ones about which we could do something. Here are some possibilities. Why?
1. Blame English English spelling is strange. Hard to learn.
OECD PISA Reading Results Singapore 6. Canada 7. New Zealand 9. Australia 17. United States
Still, people think learning to read English IS hard. English is an outlier even compared to other alphabets. Which have more predictable spelling-sound mappings.
English: DeepWelsh: Shallow
Italian Spanish German French Finnish Serbian Turkish others Handbook of Orthography and Literacy, Joshi & Aron (Eds.), Erlbaum 2006
In these studies, “learning to read” = read words (and nonwords) aloud Is this true? But reading aloud ≠ reading (comprehension)
I proved this some years ago. My Bar Mitzvah
Bar Mitzvah Languages It is also possible to mispronounce words and still know what they mean Non-oral Deaf individuals who read English; Words one knows but can’t pronounce correctly. Shallow: can be read without comprehension Hebrew (vowelled) Italian Welsh Finnish many others. So reading aloud is not a very good index of “reading”
Deeper simpler morphology Shallowermore complex morphology “Grapholinguistic Equilibrium Hypothesis” (Seidenberg, 2011; downloadable from Languages get the writing systems they deserve Languages and writing systems: There are tradeoffs Serbo-Croatian:couldn’t be like English English: couldn’t be like Serbo-Croatian The “outlier” is actually Italian.
2. How reading is taught NRC report on adult literacy: teaching adults to read is hard available methods are not effective Alternative: do a better job in the first place?
Barriers 1.What teachers are taught about reading For some children, what happens in classroom doesn’t matter. For most children, it definitely does.
Barriers 2.Failures to identify, help children with reading/language/learning disabilities No routine pediatric or educational screening for dyslexia “they’ll catch up” Some significant percentage of those 90 million low-literacy adults are undiagnosed, untreated dyslexics
We could do better. Change is needed, on the education side especially. Very hard to achieve.
3. Poverty Let’s look at the Achievement Gap
The "achievement gap" refers to disparities in academic performance between groups of students. Minority groups African American Hispanic Native American Income groups Reading, math, other areas Seen in grades, standardized-test scores, course selection, dropout rates, and college-completion rates.
For today, I will mostly focus on white-black gap. Why?large population extensive research societal concern different groups have different circumstances An important issue A complex issue A sensitive issue 30 minutes?
How to identify causes? (Does it make sense to even try?)
Maybe it is all poverty Poverty is associated with Higher infant mortality rate Atypical brain development Shorter life span Worse health and health care Higher crime rates Lower educational achievement Higher dropout rates Poorer schools with less experienced teachers Poor Reading!!
Economists’ Evidence Data: Early Childhood Longitudinal Study 20,000 children Interview data, 4 times during K and 1st grade Many kinds of data about child, family Black-white differences on reading readiness measures Their finding: a small number of measures account for gap at the start of kindergarten.
The factors are: Socioeconomic status Number of children’s books in home Age at kindergarten (months) Birth weight Age of mother at time of first birth WIC (welfare) participant
Fourth grade Eighth grade
But the gap is pretty consistent across groups. There is an effect of poverty on scores.
Earlier paper: 6 factors account for gap at start of kindergarten. Later paper: the gap gets larger over first 3 years of schooling. The 6 factors do not account for this decline. Nor do any others they could find in the data set.
Something else is going on. It contributes to increasing deficit in K-3 It affects kids from different SES It doesn’t show up in the econometric analyses What is it?
ECLS data set includes information about other languages spoken in home. But not characteristics of spoken language Which vary greatly. For example: vocabulary size (Risley & Hart, and others) lexical quality (Perfetti) Language
What about dialect? Languages have dialects. In the US, main distinction between “Standard” or “Mainstream” American English African American Vernacular English/AAE
African American English Spoken by most African Americans, regardless of SES 1970s: Research by Labov and then many others documenting properties of AAE Placed it in a broader linguistic context Dialect variation is not specific to English or to African Americans.
AAE and SAE overlap but also differ Phonology Morphology Syntax Discourse conventions Among others
The depth of immersion in AAE varies Washington-Craig: “dialect density” proportion of utterances that exhibit AAE features Familiarity with standard dialect varies As does degree of code switching
Could dialect differences have an impact on learning to read?
Extent of impact under-recognized
Early empirical studies reported non effects Socio-cultural controversies about intrinsic value, status Emphasis on validity of dialect overshadowed investigation of impact in the educational context Attention diverted elsewhere
Our thinking Need to establish more direct links between dialect and specific aspects of learning to read Drawing on reading research, theory And relevant research on language learning, plasticity, age-related changes
Learning spelling-sound relations Child learns to relate spellings to words in spoken vocabulary What “phonics” is for Strongly related to early reading achievement Difficult for many children Impaired in dyslexia Downstream effects on comprehension Reviews: NRP, Snow “Preventing reading difficulties”
Pronunciations in AAE and SAE Many words pronounced the same (at phonemic level) Many words pronounced differently Percentage varies with dialect density. Estimates: 30% and higher
Major Differences in Pronunciation GOLD, FLOOR, and LOW rhyme in AAE 1. Postvocalic consonant reduction MORE /mo/ 2. Substitutions for /th/ WITH /wif/ 3. Devoicing final consonants HIS /hiss/ 4. Consonant cluster reduction WORLD / werl / 5. Consonant transposition ASK /aks / Many others. Craig & Washington, 2004
SAE learner’s task Learn to map spellings onto existing pronunciations AAE learner’s task Learn correspondences between SAE-AAE pronunciations Learn to map spellings onto SAE pronunciations
Teacher: G-O-L-D, that’s “gold” [child searches spoken language vocabulary for “gold”] Child: Ohhh, “gole”
The main point AAE learner’s task is more complex But children are assessed against same achievement milestones Not a problem with the dialect Not a problem with dialect users A problem with the situation Other factors aside
Our research Behavioral experiments Pronouncing words aloud Children and adults Computational models Effects experiential differences Entrenchment effects Ways to overcome them
Behavioral experiments African American children, adults Read words aloud Contrastive: different pronunciations bound old toast Non-contrastive: same pronunciation brush air stage Latencies do not differ in ELP data base.
Children (N =22, M age =11.4 years old)
Adults: N = 32, M age = 35.5 y.o.
Summary Both children and adults mainly produce SAE pronunciations Contrastive words longer latencies (also more errors) For Ss with greater dialect density (more AAE)
Computational Model Similar to Harm & Seidenberg (1999) dyslexia model
Details Corpus: 1700 words from second grade norms AAE pronunciations created using 5 most common features of AAE About half the words are the same in AAE-SAE, and half differ
Training Pre-reading: Model learns phonological forms, to 90% correct Then introduce reading task: Given spelling, generate pronunciation
Why are these findings potentially controversial? Language Arts, journal of the National Council of Teachers of English
What could be done?
A. Nothing One group cannot legislate how another group should speak But AAE is not the dialect of instruction, business, science etc.
B. Accommodation Level the playing field! Recognize dialect differences Modify school practices to take it into account Provide pre-school activities to mitigate effects Do what they do in other countries that have major dialects?
An Indian comparison Sonali Nag (York) Chamarajanagar, Karnataka Mainstream language is Kannada Many tribal dialects Low status Low SES Mainly oral “Bridge year” pre-K Emphasis on exposure to Kannada
What Form Could This Take in US? Preschool activities To promote knowledge of major features of SAE, awareness of differences between SAE and AAE To promote ability to code switch School activities Continue activities focused on improving code-switching ability Increase awareness of problem: teachers and students Provide additional time for initial reading instruction, which takes into account language the child brings to school.
Would This Activity Eliminate Effects of Poverty? No.
Would This Increase the Probability of School Success? It could. Have to do more research!
Would This Approach Be Acceptable? I don’t know. That is a different kind of question.
Summary Dialect differences matter They affect early learning, later performance There are ways that impact could be mitigated “AAE itself is not the cause of reading difficulties so much as lack of knowledge of SAE, given educational and society expectations”
Reading is deeply dependent on spoken language Writing systems vs. languages they represent Basic Skills -- decoding Comprehension Language background Of course, literacy also changes spoken language But that is a different talk. Conclusions