Vision: A Resource for Writers
Lazette Gifford, Editor

To Write or Not to Write:

A Personal Narrative on Voice Recognition

By Margaret McGaffey Fisk
2004, Margaret McGaffey Fisk

If you are anything like me, you grew up on Science Fiction that touted the voice-activated computer.  Nobody did anything so foolish as to put fingers to a keyboard.  Scotty, transported into the past in Star Trek 4: The Voyage Home, picks up a mouse and speaks into it, declaring the computer so archaic as to need a microphone.  The question is: have we reached this point in the present day through voice recognition software?

When voice recognition first showed up, it was a dream becoming reality.  The earliest software available to the public largely targeted disabled users.  It opened worlds to people who had lost, or never had, the use of their hands.  The recognition was extremely rough and required discrete words, meaning an audible pause between them.  To use this software required the patience of a saint... or no better option.

Voice recognition changed from a curious development to watch into a critical necessity for me when, in 1994, I developed severe repetitive stress injuries in both hands, elbows, and shoulders.  My employer explored voice recognition because of the high rate of injury in our office.  The committee declared it inappropriate for general use in our environment because of noise levels where cubicles were the norm.

However, I had been involved in some of the testing and research so, when I joined those with injuries, I decided to do my own evaluation as a home worker.

I compared the available software and the various prices, finally choosing ViaVoice (  I really wanted the Dragon Systems product, the best on the market at the time with a history of success among quadriplegics, especially.  However, it carried the heftiest price tag.


The software took a long time to train and some words never came through on the first try.  I became very familiar with the old radio alphabet because, when all else failed, I could spell out the word I wanted to write.  I worked as an indexer and abstractor of health, academic and law journals.  The vocabulary provided with my software version was designed to aid secretaries in normal business situations.  Somehow, it failed to recognize the legal language (Latin) or the three-line-long chemical names.  Specialized vocabularies were available at an additional cost but since this came from my own pocket, I didn't choose to purchase them.

Being frugal by nature, and at the point where each keystroke sent searing pain through my arms, I determined I would use my two-week sabbatical to write a book I had been holding in the back of my head for years.  This was the true test of my voice recognition software and a validation of the time I had spent training it.

I learned to speak in a monotone voice (all those years imitating Spock coming to my advantage) and to pause between every word.  I did not achieve 120 words per minute (my then typing speed) and it produced some of the oddest errors, but at the end of two weeks, I had written a 60,000-word novel and created a serious edit project.

I gained one other thing from the experience.  I had a better grasp of dialogue.  While dialogue had always been my greatest weakness, speaking the lines aloud made them sound more realistic.  That lesson has stayed with me even on the keyboard.  Where before I would get requests for more dialogue and negative comments on what I put in, now, I get compliments.

Let us review.  I could barely type one word and yet, I managed to write a 60,000-word book in two weeks.  The software had some disadvantages: I developed chronic sore throats from talking so much and I had to learn to use the computer standing up so I could adopt good breathing practices.  However, these seemed minor in comparison to the looming permanent loss of both hands.

As soon as it came down in price and the software advanced to allow for natural language, I made the switch to Dragon Naturally Speaking.  This offered me the grand ability to use full sentences instead of the staccato style required by the earlier versions.

This breakthrough required that I complete a new training program, which involved reading a book excerpt aloud.  Natural language came as a struggle for me because I had been so carefully trained not to speak in sentences or with a natural voice.  Still, the ability to speak naturally came as something of a relief.  My problems with a dry, scratchy throat largely cleared up, though they never went away, while the pacing worked just as well with the new software.

I did find it difficult to avoid speaking in my characters' voices or slipping into any accents since that would undo the training.  Similarly, while I could force a throat clogged by allergies into a monotone, maintaining my "normal" voice became a struggle under those conditions.  The other big difficulty was not tensing my voice when the program repeatedly failed to recognize standard words.  Shouting did not assist recognition in any way.

Years later, my arms have improved to the point that the pain is tolerable most of the time.  Do I still use my voice recognition software (now Dragon Naturally Speaking 5: )?  Absolutely.  Do I use it all the time?  No.

My typing speed remains higher than my voice recognition speed, partially because my brain's connection to my fingers, having been trained very early and constantly reinforced, is much better and faster than my ability to speak what I am thinking.  Additionally, the editing of voice recognition passages is sometimes more than twice as extensive as typed passages, while correcting as I go is more complex, though possible.

That said, I continue to persist, not because it is fascinating technology (which it is) but because there are choices I have to make.  When my arms are flared, if I choose to work, I'm risking serious damage to the point that I won't be able to use my arms ever again.  In comparison, the difficulties with using voice recognition shrink to minor inconveniences.

In conclusion, based on my experience, I cannot say voice recognition has reached a point where it offers the equivalent value of a keyboard and mouse.  We have a ways to go before we achieve the Star Trek keyboardless computers that understand every word and can be used simultaneously by several people.  Current voice recognition requires a private space where the user will not distract others and others will not add random words into the text.  This means the software lacks a commercial drive to speed its development and so advances come much slower than might be hoped.

That said, going back to the original roots of the software, voice recognition clearly does provide an alternative, temporary or permanent, for those of us unable to use the more traditional methods of keyboarding and mousing.  The ability to transcribe from a voice recorder provides additional value to those who, while not injured, cannot type for other reasons, such as because they are driving a car.

If you are considering adopting voice recognition, be aware you will likely face an increased edit time and some errors will be more difficult to detect.  This is especially true of homonyms, because they will scan correctly.  When editing a novel written with both keyboard and voice recognition, I can easily identify where I switched because of the types of problems.  However, if you have a good reason, voice recognition, despite its limitations, can open a world that might otherwise seem lost whether because of time or injury.

Note 1: Voice recognition is dependent on memory and computing power.  Though high-end laptops can run it successfully, the instructions of most products recommend using a desktop, and performance is best with a high-end desktop as well.  I run on a 1.2 MHz machine with 512 MB memory.  Voice recognition slows my system significantly.  While Dragon Naturally Speaking can integrate with Word, remember that means both are running simultaneously and the size of the open document can impact performance.

Note 2: A more complete history of Dragon Systems is available here: