To knock out sentence context, they changed word order (e.g.
“Contribute others. The of Reading measured”). To knock out whole word
recognition, they alternated capital and lower case (e.g. “ThIs tExT
AlTeRnAtEs iN CaSe”). And to knock out letter-by-letter decoding, they
substituted letters in such a way that word shape was maintained (e.g.
“Reading” becomes “Pcedirg”).
Letter decoding was found to
account for 62 per cent of reading speed; whole word recognition 16 per
cent; and sentence context 22 per cent.
I wasn’t there for the tests, but I believe that is measuring reading speed at margins other than what we find on the printed page. "Knowing what is coming" is in my view most important for reading fast. (I like to say "It took me 45 years to read that book." If you think you "just started" the book in your hands right now, you are failing to understand the proper marginal unit.)
I find that when I try to read graphic novels, I am not a very fast reader at all. My eyes get confused from not knowing where the next block of text will appear.
I am also struck by an incidental remark toward the end of the article; we are getting closer to the truth:
…among the faster readers, predicting words from sentence context made a
bigger contribution to reading speed than among the slower readers.
Addendum: Here is my previous post on reading speed.