Skip to Content

Training Dyslexics First to Hear, Then to Read

For most of us, reading comes as naturally as falling off a log. But for many otherwise normal-and even gifted-individuals, the task is an arduous one. In fact, as many as one in five schoolchildren suffers from dyslexia, an unusual difficulty in reading.

Scientists theorize that although dyslexic children can hear normally, they can’t accurately interpret many language sounds and, therefore, can’t remember which symbols represent which sounds. This deficiency causes problems in both spelling (translating from sounds to symbols) and in reading (translating from symbols to sounds).

Now speech researchers Paula Tallal of Rutgers University and Michael Merzenich of the University of California at San Francisco have devised a novel computer program that addresses what they believe is the underlying cause of the deficiency: an inability to identify the correct sequence of acoustic cues that indicate language sounds, cues which in normal speech are presented so compactly that they overlap with each other.

Tallal bases her conclusions on 20 years of research on six- to nine-year-old children. In one study, she asked dyslexic subjects and normal readers of the same ages to listen to two complex tones of different pitches. She then asked both groups to repeat the order of the tones, using computer keys. As the time period between the tones became shorter, the performance of the dyslexic children deteriorated rapidly, while that of the normal readers was not affected. Normal readers reached the target score, 20 correct responses out of 24, with intervals as short as 8 milliseconds, but dyslexic readers needed at least 305 milliseconds to achieve the same score. When she repeated the experiments using three, four, and five tones, the differences in performance were virtually the same.

Tallal also asked the children to indicate whether the pair of tones they heard were the same or different. She found, similarly, that impaired subjects, unlike those who could read normally, could not differentiate among the tones at rapid rates of presentation. But their performance improved dramatically when she lengthened the duration of the tones and the spacing between them.

Tallal reasoned that slowing down speech sounds in the same way might help dyslexic subjects sort overlapping cues into discrete events-the task entailed in understanding language. So she processed normal recorded speech to make it 50 percent slower and boosted the volume of the transitions from consonants to vowels to make them more prominent. Played back, the processed speech resembles what a slightly testy computer might sound like: condescendingly slow and faintly metallic. To make the processed speech appealing to children, Tallal and Merzenich incorporated it into a computer game in which kids try to identify certain sounds and gradually work their way up from slow to normal speech.

Tallal and Merzenich-who have formed a company, Scientific Learning, to market the game-are encouraged by the kids’ reaction. Tallal reports that reading-disabled kids show a keen interest in the processed speech that their normal counterparts don’t. More important, though preliminary, are the results: One group of seven children aged five to ten years began the program two to three years behind their peers in language development. After four weeks of intensive training, six of the seven improved their language-comprehension skills to near-normal, normal, or above-normal levels. Another group of eleven children yielded ten success stories after similar training.

Other speech researchers, including Michael Studdert-Kennedy of the Haskins Speech Laboratories in New Haven, Conn., and Maria Mody of the Albert Einstein College of Medicine in New York, agree that dyslexia is often, if not always, associated with a phonological deficit. But in their view, it is an impoverished ability to differentiate among similar-sounding speech sounds that causes dyslexics to make errors-not a difficulty ordering speech sounds.

To show that impaired readers have difficulty distinguishing only among similar language sounds, as opposed to those that are distinctly different, Mody tested a group of 20 second-grade children selected expressly for their errors on one of Tallal’s tests. When she asked children to report the order of syllables “ba” and “da,” they made more errors when the syllables were presented closer together. But when the syllables they heard were not as similar, such as “ba”-“sa” or “da”-“sha,” the children made almost no errors on either task.

Tallal and Merzenich say the success of their program in helping kids learn to read argues for the validity of their approach. But they concede that more research is necessary to understand why impaired readers have trouble distinguishing among some sounds but not others. They also point out that it is still unclear why dyslexic readers err when trying to sort language sounds into the correct sequence of letters. One possibility is that because they find it much harder to tell the difference between speech sounds, they simply need more time to make decisions about them, and this leaves little attention left over to remember the sounds’ exact order. It’s like the “I Love Lucy” episode where Lucy and Ethel are struggling to inspect chocolates passing them on a conveyer belt. After a while they give up trying to select the right ones and simply grab whatever happens to be in front of them at the time.

Keep Reading

Most Popular

This new data poisoning tool lets artists fight back against generative AI

The tool, called Nightshade, messes up training data in ways that could cause serious damage to image-generating AI models. 

Rogue superintelligence and merging with machines: Inside the mind of OpenAI’s chief scientist

An exclusive conversation with Ilya Sutskever on his fears for the future of AI and why they’ve made him change the focus of his life’s work.

Data analytics reveal real business value

Sophisticated analytics tools mine insights from data, optimizing operational processes across the enterprise.

Driving companywide efficiencies with AI

Advanced AI and ML capabilities revolutionize how administrative and operations tasks are done.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at with a list of newsletters you’d like to receive.