Slashdot Mirror
MS Psychologist on How We Read
RenderMonkey writes "In another follow-up to Can You Raed Tihs? Microsoft's Kevin Larson, a cognitive psychologist, dissected the main hypotheses on how we read at ATypI's Vancouver Typography conference. "Kevin supports the 'parallel letter recognition' model. People don't he says, recognise whole-word shapes. Instead the recognise each of the letter components and then make a series of best-guesses on the information returned to assemble, first, phonemes and then words." So what about the case of patterned re-ordering, aka the counter example to Can You Raed Tihs?"
Microsoft has a psychologist?
Personal observation and various readings in the topic make me pretty confident that context is critical for letter recognition. Whether that means words are recognized as "whole words" or not, the fact is, it very clearly is not a simple, straightforward bottom up "letters then phonemes then syllables then words" recognition process. Recognizing the letters is partly a feedback loop with the words and other parts, as demonstrated by experiments where parts of letters are blacked out. In a recognizable context (i.e. a word) they're still identifiable. Standing alone, they are not.
7 November 2006: The day Americans realized corruption and incompetence weren't addressing 11 September 2001
Great....a justification for Leet speak...
After reading the article, it seems rather lacking in explanation. Okay, so Larson says that there are three main models for word recognition and presented evidence for and against each one; parallel letter recognition being the one supported by his evidence. The article then goes on to present none of the evidence, which is a shame as it could have been enlightening for us masses.
So, we have our counter-example here but what about the rest of the rules to flesh this out? What rules do we need to follow to still allow comprehension of otherwise obfuscated text, and what rules produce unintelligible rubbish?
Incidentally, could this be used as our next method for determining a human user versus a program, rather than using images? How well could this survive being decrypted by a well crafted perl script? Maybe some research is in order...
I showed this type of paragraph to several of my Japanese co-workers, who are very good at English but not quite native level yet. They had an extremely difficult time making out the words and couldn't grasp the meaning of the whole paragraph at all.
A lot of reading comprehension comes from how you learned the language in the first place. Your ability to understand a given second language depends on how similar it is to your native language.
I think in this case its mostly a vocabulary problem. Native speakers know that "wlohe" and "raed" are not English words, and our minds can easily search for possible alternatives, but non-native speakers would need a dictionary to confirm that those aren't actually words they didn't know.
Yes, there are some cool examples. However, if a person jumbles up the letters of a word, knowing what the original word is, they may be subconscieously keeping a pattern which denotes the original word. This pattern is how we read. Changing the letters' order in a more mechanical way (as was done by the researchers at British Columbia) seems to produce less readable text.
What the research by Cambridge Uni may show is that it is not the exact ordering of letters that we recognise.
Anyone who is dyslexic would be able to tell them that for nothing!
If I told you that the human brain was an amazing pattern-recognition machine, would you give me a nobel prize? I think not.
Pattern recognition is how we make decisions every day. Our brain does not compute every possible outcome of a situation, it merely takes previous experiences and extrapolates on them.
This is the same reason that brain activity drops off after two years of age. The brain has developed and stored enough patterns to make "informed" decisions. We do not have to re-learn these patterns, only refer back to them, so brain development slows down.
Your paragraph only reinforces this. We see each word in the paragraph, and based on the context in which we see the word, we make educated guesses at what the next word should be. We check back to the patterns which we have already created, and verify that we have chosen the correct action.
This is the reason why you can look at your e-mail and see what is spam and what is proper better than your computer. This is the same principle for face recognition. We equate somebody's face with our previous experiences, people we know, and make immediate judgements of that person based on skin colour, eye placement, hair colour, hair style, face shape, etc. That's why people have an "Honest" face. In fact, most people that you consider to be honest, look more like you than people you consider dishonest. For me, this is why I would sooner believe Bill Clinton then I would have Marin Luthor King. (and before I get crucified on this one, my true opinion is that Bill Clinton was a slimy weasel used car salesman and M.L.K. was perhaps one of the greatest non-manufactured heroes of the twentieth century)
This is not startling news, this is only a pattern which we have put a name to and examined.
I can read this sentence almost instantly and i am not english native, neither is my home language directly related to english language (Latin based). But on the other hand I read very easily english (although I write it with a lot errors).
I think this has partly to do with *how* you learnt english , but not whether it is your home language or not. (Heck I understand english humor perfectly like Discworld tongue-in-cheek humor whereas some Australian friend do not understand it). By "how" I mean how you read any word even in your own language !
Believe me or not I know I read by "grasping" what the phonem of a word are, and not necessaraly in a linear order. For example when i read a word which i do not know at all, I realize I read 1st phonem , then 3rd and 4th then 2nd etc... And not 1st , then 2nd, then 3rd. I also read book very quick with a full comprehension of what is written.
This seems to me to be pointing that "reading" might be far more complicated than most people describe it,might be education and cultural related, and depend on other factor. Such as training, whether you find pleasure in it or not, and (tadam) whether you learnt the language on your own without using somebody else method (as in my case with english : self taught).
It might be interressant to compare how people learn foreign language and then compare how they read *jumblewd* word out of those foreign language. it might give better conclusion than using native reader recognition of words.
C. Sagan : A demon haunted world:
http://www.amazon.com/gp/product/0345409469/
visit randi.org
"(a typical psychology study uses 10 college students for its sample)."
Really? Are you talking a psychological study from an undergrads perspective, a graduate perspective or a postgraduate?
My department does a LOT of psychological study at Indiana University...unless the study actively involves college students, we don't seek them out. Heck, I don't remember last time I did a study with 10 folks in our PILOT group (i.e., to work out bugs in administration and otherwise), let alone a full study. Heck, we were a little pissed a few weeks ago as we were contracted to do a pilot study and work up the methods and all that crap...unfortunately we WANTED to get 200 folks just for that pilot...we only got 186. Thats 186 pieces of information that will ultimately be discarded before the REAL study with at LEAST 10x that amount goes public.
Heck, for the last master's thesis I was helping with (the kiddies ask for my help since I do this daily) 200 students was a small amount...yeah...the ultimate goal of that is to get it published, but its more of a "Look What We Found -- Give Us Some Money To Prove This Is *REALLY* What My Hypothesis Says".
I always have trouble with folks that aren't involved with a field for their willingness to draw sweeping conclusions from their tiny, and generally imaginary, homogenious life experience.
I agree, context is critically important in human pattern recognition. Context appears to constrain the possible choices at each junction of reasoning. To put this in context ;) imagine a short story: you enter a room, you turn on the light, you sit down in your chair. At the "you turn on the light" action, one of the possible branches in that story is not getting a drink from the water cooler. Your choices at that junction are limited by the context, or an analogy of "this leads to that".
Shh.
Ever notice how you can read chicken scratch handwriting usually pretty easily? I know when I take notes in class its usually the first letter, then some sort of semblance of letters in the middle and the last letter is usually right. Given the context, almost anybody can read these notes even though they are usually no more than some lines with an ocassional random dot above them...
Nullum magnum ingenium sine mixtura dementia (There is no great genius without a mixture of madness) - Aristotle
I don't think we really need to go any farther than Chinese to deduce that people recognize words from their shape more than from individual components of the shape.
Not that I've read the article or anything . . .
Someone you trust is one of us.
Look people, enough grumbling about Microsoft and their psychology department... as a corporation who's main product is a human-machine interface, it is in their best interest to understand and maximize everything that eases these tasks.
...
They studied eye strain, and whipped up an improved font display system called ClearType. Windows XP has a Speech module in the control panel that's getting pretty good at speaking random text. Word and their Spelling modules are pretty good, but English isn't the only language.
Microsoft is obviously positioning itself for something big. Is this a new phase for improving Spell Checking - mimic the brain's methods for decoding scrambled text into a word? Is it time for Microsoft to take on Babelfish's language conversion -- on-the-fly language converting instant messaging with better results. New OCR technology for converting text embedded in images? Whatever it is, there's money to be made.
Finally, don't you find it ironic that an article on word recognition contains spelling errors?
2: The reader recognbises each letter in turn
foldplay your photos won't know what hit them.
Well, I am sure. Every time you changed the last letter in a word i had to stop and think about what the word was.