Studies challenge legitimacy of speed reading
- Thread starter Dustin Kesselberg
- Start date
-
- Tags
- speed reading
Who are 5 or so leading scientific experts studying speedreading, reading comprehension, skimming, etc. in 2007? Or where can I find this on the web? It seems like a better place for us to start than some of these somewhat shady anecdotes.
I'm disappointed by Just and Carpenter's research (way back in 1987). I expect reading speed and comprehension ability to be a bell curve. I'm curious how people perform at the highest end of that bell curve. For example, elite humanities professors probably have to read a ridiculous amount of material at a high comprehension level. Surely someone studies their reading speeds and comprehension levels at peak performance?
I'd think the legal community would invest big bucks in this type research.
I'd think the legal community would invest big bucks in this type research.
Silly Green Monkey
Cowardly Lurking in the Shadows of Greatness
Dustin, stop spamming. I don't see what posting repeats of someone's post can prove.
I took a speed-reading course as an undergrad, where we learned to read faster by focusing fewer times on each line. Very slow reading speeds refocus the eyes on every word, normal reading speeds focus the eyes five or six times along the line. I tend to focus my eyes about three times per line.
Do you still have your speedreading tool, Wolf? I lost my Visualizer years ago, and would love to find another.
I took a speed-reading course as an undergrad, where we learned to read faster by focusing fewer times on each line. Very slow reading speeds refocus the eyes on every word, normal reading speeds focus the eyes five or six times along the line. I tend to focus my eyes about three times per line.
Do you still have your speedreading tool, Wolf? I lost my Visualizer years ago, and would love to find another.
Boy oh, oh boy. Dustin with so many posts and still hasn't learned one thing; his avatar is suitable.Jeff Corey
New York Skeptic
- Joined
- Aug 2, 2001
- Messages
- 13,714
And yet he says, "In a highly unscientific survey (conducted by myself)..."
Dustin may unpopular for some posters in this thread, but some of the individual, anecdotal speed-reading claims posted here do seem to me to be a bit suspect, and Dustin may be on the side of the angels with that point. Dustin, it doesn't show much respect to the rest of us to post 10,000 words twice. It was clever the first time and made your point strongly. After that it becomes cumbersome and disrespectful to us other readers.
Dustin Kesselberg
Illuminator
- Joined
- Nov 30, 2004
- Messages
- 4,669
It's to remind people that free speech is important even when you disagree with what is being said.
Dustin Kesselberg
Illuminator
- Joined
- Nov 30, 2004
- Messages
- 4,669
I apologize.
Beerina
Sarcastic Conqueror of Notions
- Joined
- Mar 3, 2004
- Messages
- 34,934
That's complete BS. I can clear that easily. It's only 5 words a second. I can read out loud at 300 words per minute.
In all fairness I must point out that parsing perfectly valid English sentences is known to be an NP-complete problem. When you get to complicated ones, nobody and nothing can parse them easily.
Here is a trivial example, "buffalo buffalo buffalo buffalo buffalo." That's a valid English sentence, modulo my removal of capitalization. Don't believe me? Consider the English sentence "Boston cattle bewilder Boston cattle." You have no trouble parsing that. But the word buffalo can refer to the city, the animal, or can be a verb meaning "to bewilder or confuse". So the first sentence and the second sentence have identical parse structures - hence the first is grammatically correct. (Re-read it. In fact the word "buffalo" repeated any number of times is always a grammatically correct English sentence.)
Cheers,
Ben
fun fringe example, but i think it's of limited relevance to this discussion.
Wudang
BOFH
Not in the UK as we don't use "buffalo" as a verb.
Reminds me of a great PhD paper I read in the 80's on anaphoric resolution - how we determine what the referent is of pronouns aand such. The writer went through all the (then) theories and showed how they all failed to explain
"The boy stood on the burning deck,
picking his nose like mad.
He rolled it into little balls
and flung it at his dad".
Also remember that whatever heuristic models the brain develops only need to be good enough - they don't need to be able to parse every valid english sentence. They have to be just good enough - for suitable values of enough. And despite being a programmer I think any current computer models should be looked as nothing more than vaguely useful analogies for the purposes of discussion.
Although to continue "in all fairness," NP-completeness doesn't mean much when applied to a problem of this sort. All NP-completeness means is that a Turing machine can't solve the problem both quickly and without error. Since human beings are well-known to make errors, this factoid earns a great big "so what" from the psycholinguists.
So the next question to ask, of course, is what strategy humans use that is quick but potentially error-ridden. And, of course, this gets back into probabilistic parsing....
Around about Christmas time, I was thinking about putting together some code that could this handle this sort of ambiguity, in the context of lisp style coding.
This poem could be rephrased as;
Code:
(assert (stood boy burning-deck))
(pick he nose)
(roll it little-balls)
(flung it dad)Now if we create a stack where we push on, in sequence the arguments of each function, and then its output as we work through the code, when we encounter any it type statement, we can just roll back until we find the last element on the stack which has sufficient properties to satisfy the current context.
Now obviously, this is very simple, it doesn't double guess itself or backtrack in anyway if it makes mistakes. It just gives up on the coder as being incoherent, but it seems to me that it would handle this example OK.
The first ambiguity encountered is 'he', the only value satisfying it is 'boy'.
(pick he nose) would then return both a modified boy and snot.
Snot would satisfy (roll it little-balls) and the output from this would satisfy (flung it dad).
At this point I gave up with the implementation, because I didn't fancy making endless property lists for all objects and outputs. But it seemed very implementable to me, in principle.
Now obviously, I don't know the field, and I am cheating by using context-free grammars, but what's going on that stops some variant of this from solving the poem?
Except, um, it doesn't. You're pulling "snot" out of.... um,.... thin air?
The word "snot" -- for that matter, the concept, SNOT -- doesn't appear in the discourse prior to the pronoun. How does the computer know that the operation (pick boy nose) yields (snot)? You're essentially punting to world-knowledge, which is the traditional failing of all parsing systems. Yes, if you understand what the writing says, then it becomes easy to parse --- but if you already understand what the writing says, why are you bothering to parse it in the first place?
I believe that Doug Lenat has made similar statements, and has spent the past thirty or so years attempting without success to demonstrate that "principle" by making those property lists. Everyone else in the field regards Lenat's lack of success as a conceptual proof of infeasibility.
Our inability to represent world knowledge. Look up the term "AI-complete" for a good summary.
Last edited:
Ok, that makes sense. But presumably understanding the poem is dependent on your priors. I don't see how any person could understand the poem without the world knowledge that snot comes out your nose.
Well mainly, because I was concerned with resolving ambiguities within a well defined framework.
I can well believe that.
Thanks, I was considering parsing as a problem separate from, but dependant
from world knowledge and not worrying about how to bootstrap myself into the problem.
Yes. But you've stumbled upon a key issue in your word "priors."
The most common theoretical position -- we can thank Chomsky for this -- is that syntax is autonomous of meaning, and prior to it. Hence the development of creatures like context-free grammars (the name says it all, really), lexical-functional grammars, and so forth. The general idea is that we identify the syntactic components of a sentence and then use those syntactic components to drive meaning; the adjective "Polish" means something different from the verb "(to) polish."
Syntax, then, is "prior" to semantics. I'm generally an anti-Chomskian myself, but I find myself in complete agreement with him there. There are very few theories that posit that semantic understanding is necessary for parsing -- and pseudo-sentences like *"Your frammis pilks" are extremely problematic for such theories. I may not know what a frammis is, but I know it's a (common) noun, and I know that you've only got one of them, and I know that pilking is an action that frammises -- frammisi -- framiztim? -- can take.
Ok but don't you so still need semantic understanding to resolve the anaphoric cases? I can't see how you could do this in general without checking for matching properties.