Science of Reading

reading wars

controversy of primary focus on teaching phonics vs whole language approaches

visual word recognition Rastle 2018

identify letters, process order, identify word

easier and faster than spoken word recognition

words read out and understood in 500-600ms

written word identified when present for <100ms

tend to look at each word for 250ms

Harley- eye movements

4-5 letters lie in fovea

fixations last 250ms

range of 15 characters to right, 3-4 to left

saccades jump 8 letter spaces

short words fixated 25% long words all

10-15% backwards saccades

preview info of words from outside fovea

factors affecting word recognition speed and accuracy Balota et al 2004

frequency, length, age of acquisition, semantic ambiguity, authographic similar words, concreteness vs abstractness

serial search model Forster 1976

check every word against every word we know one by one

parallel process models

Mortons Logogen model 1969- all possible words checked simultaneously

Connectionist models (IAC)- includes neuronal mechanisms

Interactive activation and competition AIC model- Rumelhart and McClelland 1982

info flows in all directions between three levels:

visual features, letter level (slot based units), word level

if a feature is detected it triggers excitatory connections for words starting with letters with that feature and inhibitory connections for letters without that feature

strengths of IAC model

demonstrates that parallel processing is computationally feasible

highly interactive- top down influence existing knowledge e.g. word superiority effect more common words

limitations of IAC model

slot based coding of letters- readers are not affected by letter switches

ignores sound information- faster mapping

mapping letters to words and meaning

letters to sound easier than letters to meaning (learned later)

lexical ambiguity Rodd 2017

when words have multiple meanings

Reordered access model Duffy et al

word meaning access is a competition

words biased to more common meanings

early context can predict word meaning

word dominance Rodd 2013/2016

hobbies change dominant meanings, recent experience with meaning increases dominance