Voice recognition technology first went into action in the cockpits of US fighter planes more than three decades ago. Since then, the technology has made its way into more accessible locations - such as offices and homes. But the habit of talking to computers is only just beginning to take off in British schools - and an on-screen cauliflower may hold the key to its success.

The cauliflower made its debut at this year’s BETT educational technology show, in a demonstration to mark the launch of the most wide-ranging investigation ever conducted into the use of voice recognition in schools. Managed by BECTA (the British Educational Communications and Teschnology Agency), the Voice Recognition (VR) Project is studying the potential the technology holds for learners with special educational needs. But it could have spin-offs for all children.

Voice recognition has come a long way since pilots used it to launch missiles without taking their hands off the controls. Walk into any superstore, invest as little as pound;50 on a box of software and, with some preparation and a little coaxing of the system along the way, you can dictate fluent prose to your computer.

The attractions are irresistible. It is so easy to imagine the boost this technology could give a child who has never been physically able to put pencil to paper, and the sheer exhilaration it could bring students who find writing a mystifying or even terrifying experience.

BECTA’s Chris Stevens says this instant appeal causes problems. “People immediately see the potential, so they go out and buy the system. But they fail because they don’t do the groundwork, or because they are not using the software in an appropriate way. There is a huge range of ways voice recognition could be applied - what is needed is some guidance on how it can be best used.”

The DFEE-funded project sets out to answer a host of questions posed by teachers, and raises fascinating issues that few will even have considered.

From the 40 schools and specialist learning centres which submitted research topics, 12 were selected for the project, each receiving pound;3,500 to spend as they wished. Topics range from opportunities for developing family literacy (Bilston Community College, West Midlands) to the possible benefits for visually impaired learners (Exhall Grange School, Coventry).

With many programs, the user must first train the software to recognise individual voices before it can be put into action. Each user trains the system by reading aloud a special script. Present a complicated script to a poor reader or a blind learner, however, and motivation could plummet at the outset.

One possible solution is for a helper to read out the script, with the student repeating it to the computer. But the VR project will go one stage further, looking at whether training is necessary at all. And experience with the cauliflower suggests the results could be surprising.

Chris Stevens explains: “At BETT, we did a demonstration in which people had to label on-screen pictures - a cauliflower, red apples, green apples, pineapples - by talking to the computer.

The software was untrained and we achieved almost 100 per cent accuracy on an open exhibition stand, competing against an amazingly high level of background noise.”

Could a dialogue with a computer help a child with autism to communicate? Might the need to sustain accurate dictation help hearing-impaired students develop speech consistency? And how does the software respond as pupils grow up and their voices change? The researchers are putting those, and more questions, to the test.

Although state-of-the-art, “continuous speech” software allows the speaker to talk quite naturally, the project will also look at the use of “discrete” systems, which require enforced pauses between words. Some dyslexic learners may benefit from a clear association between saying a word and seeing it on the screen.

Participants have been given free rein over their choice of technology, although they will be encouraged to test their findings with more than one software package. BECTA expects to publish the results in Spring 2000. Stevens says: “Although we are focusing on special needs, we hope to pick up issues of good practice for all children. But we will be talking closely to people from the National Literacy Strategy before we make any broad statements.

“Mainstream children still need to be able to read and write - this is just another tool to help demonstrate these skills.”

The technology is now finding its way into mainstream education. Emmeline Johnston, of IBM, says: “We believe there is a place for this in the classroom and the staffroom.” IBM says its ViaVoice software may not work well with the voices of under-13s. But older pupils are using it to dictate essays and staff are now dictating reports into administration systems.

Johnston advises that as long as the software is trained in the environment where it will be used, background noise should not be a problem. “The worst mistake is to give a pupil a laptop to take home and train. It will not work in class. But as long as you have done the training properly, this software does save time.”

Dragon Systems 01242 678575 www.dragonsystems.com

IBM 01705 492249 www.software.ibm.comspech

Lernout amp; Hauspie 0800 973 365 www.lhs.comvoicexpress

Each supplier produces several grades of voice recognition software. They vary in their sophistication. A top-of-the-range version might take dictation, accept spoken commands for controlling the computer, and even read back text to its user.