Recently, we've reviewed a couple of speech recognition packages (you can read the reviews here). They're good; with a bit of training you can get far higher word speeds than you could get by typing, even if you were a professional ten-fingered typist. And yet speech recognition is still not massively popular. It is used, but in nowhere near the quantities that were predicted when this product category was first created.

Good ideas sometimes take a while to reach fruition, but speech recognition has been around for a long time; good quality, usable software has been available for under £100 for more than five years. Even so, few people are making use of it, either at home or in offices.

Why? I have a few ideas. First, there's the self-consciousness associated with talking to your computer. This isn't Star Trek, and while some people will undoubtedly get a kick out of having an inanimate beige box doing what they tell it to, that probably just means they need to make more friends.

Second, speech is disruptive in an office environment. It's bad enough having someone next to you talking on the telephone while you're trying to work, but imagine an office full of people saying "Dear Sir comma father correction further to hour correction our letter of the twenty first...". You'd go nuts. Or you'd all have to wear ear plugs.

Then there's the security/privacy issue. Do you really want everyone else around you to hear what you're typing? Half the time it's likely to be non-work-related e-mails and other mildly dodgy stuff, so speaking out loud would just reveal to your boss and colleagues exactly how little real work you were doing.

And what about the strain on your voice? That was always a problem with the old 'discrete word' speech recognition packages, which required you to leave a gap between words. Repetitive strain injury of the larynx, anyone? And even with 'continuous speech' packages, there's a limit to how long most people can talk for without losing their voice or feeling hoarse.

But all these points have been made before. The one that I believe will really crush any attempts at making speech recognition widespread is that different areas of the brain seem to be involved depending on whether you're speaking or writing. Despite currently studying psychology, I have no real evidence for this other than a couple of recent examples, but obviously I won't let that stop me from making sweeping generalisations.

Two journalists wrote reviews for this site which were significantly below their usual quality, and had to be edited much more than usual. When I contacted the journalists concerned, asking them if the reviews had actually been written by someone else (as they were so very different), they both said that the reviews had been written using speech recognition software.

Interestingly, they couldn't tell that the quality of the reviews was any worse than normal, but it was obvious to anyone familiar with their work. There was no flow or style to the reviews, just a punctuated list of facts. In effect, it was written like conversational speech, which also has little flow or style unless it's prepared beforehand.

Further anecdotal evidence comes from the fact that even when I've been out with friends and am so drunk that I can barely speak, I can still write. I've tried this experiment out a few times (just for your benefit, dear reader) and as long as I can focus on the keyboard, the reviews or articles I write are the same quality (according to independent readers) as the ones I write when I'm sober. Which should probably concern me.

I expect that somebody somewhere is studying these effects using brain imaging scanners, in which case I'd be interested to know the results. But if my hunch is correct, we'll be using keyboards for a long time yet.