Brain researchers have long studied how students hear and read individual words, but it has proven difficult to parse out what happens when a reader understands a long and difficult passage of text.
Now, a team of researchers at Northwestern University in Evanston,Ill., have developed a new way to view reading comprehension in the brain, and in the process highlighted a new angle for testing comprehension that isn't skewed by a student's background knowledge.
In the study, published in the November issue of the journal Frontiers in Neuroscience, Julia A. Mossbridge, a research associate at Northwestern, and her colleagues there monitored college-aged readers using electroencephalography, in which electrodes placed on a person's scalp measure electrical activity in the brain in response to specific actions, such as hearing a sound or reading a word.
"It's almost counterintuitive that you could use [brain electrical signals] to study something as complex as comprehension. It makes sense that you could use them to study something like [phonics] because you're saying, let's look at the difference in response to 'cat' versus 'dog,' or nouns versus verbs in word identification," Mossbridge said. "In a certain sense, we wanted to subtract out decoding and word identification because we wanted to look strictly at comprehension."
Random or Order?
The researchers presented each student with nearly 1,200 words of challenging text, part of the first chapter of Emile Zola's 1893 classic, Doctor Pascal. (Mossbridge said it was chosen because "no one had heard of it.") In one trial, participants read the text word by word in random order, and were asked to press a button every time they read the word "and." In the second trial, the participants performed the same task on the words presented in order, and in addition to marking the word "and," they were given a comprehension test on the text as a whole.
When participants read and understood the text well, there was a significant difference in their brain activity between when the words were read in proper order (shown in red) versus when they were read randomly (shown in blue):
By contrast, those who did not comprehend what they read had similar patterns of brain activity regardless of whether they read words one by one in order (red) or randomly (blue):
(The brain activity on the y-axis in these charts is measured on a current-source-density index, or CSD, which indicates the position and strength of the electrical information coming from the electrodes positioned around the scalp.)
By analyzing the difference in these two patterns, Mossbridge and her colleagues could correctly predict which readers comprehended the text 88 percent of the time. Among readers who saw the random words first (and thus were familiar with all of the words), the measure was 90 percent accurate.
"An important piece to me is that someone can read a whole [text], and they can even be at the college level, and seem like they've read it well--and they can do horribly on a comprehension test," Mossbridge said.
"Why would that be? It's because there is something that's not sticking--it's in working memory, or that's my guess" based on a critical difference found in electrodes placed over a part of the brain associated with working memory, she said. "Something is not carrying the information along, binding together the past with the future in what's just happened and what's coming up. That needs to be worked out. There's so much research out there showing the importance of working memory to reading comprehension, so this is another reminder."
Targeted Comprehension Test
Mossbridge and her colleagues are planning to refine the measure by repeating the experiment with younger students and those of different reading abilities. But already, Mossbridge said the study could be helpful for educators looking for more targeted ways to test reading comprehension, particularly in complex and difficult texts.
The Northwestern researchers developed their own comprehension test, using questions with multiple answers. A student was given credit for a given question only if he or she chose all of the correct answers and none of the incorrect ones.
"One of the problems with common reading comprehension tests is, if you happen to know a lot about physics, you don't have to understand the text about physics in order to answer the questions correctly," Mossbridge said. In the study's comprehension test, "If you didn't understand the text, your performance by chance would be not getting any questions correct."
The format was intended to weed out students guessing at answers, and the use of an older piece of fiction helped prevent students from relying on their own background knowledge rather than information from the text.
Photo: The charts show differences in brain activity while reading words in random order or in their proper order for a text. Note that readers who comprehend the text have a greater difference in activity between random and ordered words than readers who did not comprehend the text. Source: Northwestern University.