kitchen table math, the sequel: the "triple dissociation" and slow and fast readers

Tuesday, June 25, 2013

the "triple dissociation" and slow and fast readers

re: We can see just 7 or 8 letters at a time

The NYU reading-speed study, which has come up a couple of times on ktm, found a "triple dissociation" amongst three contributors to reading speed:
  • Decoding by phonics (62% of reading speed)
  • Whole-word recognition (16% of reading speed)
  • Word-prediction via knowledge of grammar and context (22%)
"Triple dissociation":
Surprisingly, the effects of the knockouts on reading rate reveal a triple dissociation. Each reading process always contributes the same number of words per minute, regardless of whether the other processes are operating.
I've just this moment realized the implications of the triple dissociation they found:
  1. If students don't know phonics, they can never read quickly enough to do college-level reading,  which requires at least 200 wpm. 
  2. Students who don't know the grammar of written English well, will be significantly slower readers than students who do. ("Know" in the procedural sense of being able instantly to comprehend grammatical structures such as appositives and relative clauses and anaphora and the like.)
  3. Until you have substantial background knowledge in a subject, you will be a slow(ish) reader in that content area. 
I'd love to know how that 22% is distributed between grammar and background knowledge. The researchers did not test to see how grammar and background knowledge

Parts, Wholes, and Context in Reading: A Triple Dissociation by Denis G. Pelli mail, Katharine A. Tillman

1 comment:

Anonymous said...

I looked at the triple dissociation paper you pointed to, and am not very impressed with the connection between their experiment and their conclusions.

I'm not convinced that their letter substitution method really gets as the letter-at-a-time vs word-at-a-time recognition question, as their definition of word recognition relied on substitutions that preserve low spatial frequency, but not chunks like consonant clusters.

I'm also not impressed with their decision to use constant center-to-center spacing on a font not designed for that. Distorting the word shapes in that way certainly slows down my reading, so they artificially favored letter-at-a-time reading.

I think that there is good reason this is a PLoS One paper, and not in a journal with stricter peer review.