So how did Simulscribe fare?

The messages in which I spoke in my normal voice were translated almost perfectly. Telephone numbers were transcribed correctly and the text was intelligible with, for the most part, only minor errors. Question marks appeared beside some words - usually proper names - indicating that Simulscribe was spelling them out phonetically. Some proper names it spelled correctly, however, including Lewinsky. (But not Monica.)

The only significant error was Simulscribe's complete inability to transcribe my standard North American pronunciation of "Kalamazoo," the city in Michigan.

I then tried leaving a message in which I played part of a recording made on a fairly good Olympus digital voice recorder. I held the recorder's speaker to the phone's microphone and hit the Play button. I thought this might simulate a bad cellular connection. The person speaking in the recording was also not consciously speaking clearly or slowly.

The results were, as I expected, inferior, with more question marks and enough words completely misinterpreted - but with no question marks - that the message was virtually unintelligible.

The real reason for this poor performance turns out to be somewhat different. Simulscribe founder James Siminoff explained that the company's software is as accurate as it is because it's tuned to listen just for the kinds of words, phrases (and numbers) people are apt to say when they leave typical 18 to 30-second voice messages. My recording was of an interview, which the Simulscribe software could not be expected to transcribe accurately.

For my final set of tests, I wanted to see what would happen if a person with a foreign accent left a message. Not having any confederates with foreign-sounding voices, I did a bad imitation of British comic actor Peter Sellers imitating a South Asian person. (Apologies both to the Sellers estate and the South Asian community.) The things we reporters won't do to get our story!

Whether it's a reflection on the quality of the Simulscribe speech-to-text technology or the badness of my fake foreign accent, I don't know, but these messages were transcribed at least as well as the messages left in my normal voice - including guessing right at the spelling of the fake name I used, Amir. (Or perhaps it's that the Simulscribe technology hasn't been tuned yet to interpret a Canadian accent.)

Bottom line: If speech-to-text transcription of voice mail sounds like something that would be useful to you, and it does to me, you only have to ask yourself whether it's worth the price of admission. As noted, though, Simulscribe is not the only service provider. So shop around.

Perhaps Rogers will let me try SpinVox.