How to Get Used to Dictation: Weeks, Not Hours

The research we found points to weeks of regular use rather than a single session. The clearest measured curve, for radiologists, leveled off after about 10 weeks ¹. The research does not support a fixed number of hours after which dictation feels as natural as typing, and we found no study that measures how long the awkward feeling lasts. What the evidence does show is that navigation and error correction take a big share of early dictation ², that a shift to combined voice and keyboard correction appears after about 20 hours of use by one secondhand estimate ³, and that short, consistent daily practice is what the habit research we found suggests, although that research is not about dictation. This guide covers the timescales, why the first sessions feel slow, what to dictate first, a daily routine, how to measure progress, and how to keep the habit when the keyboard is within reach.

Dictation skill builds over weeks, and the timescale varies widely

The clearest measured learning curve comes from a 2013 hospital study of radiologists who moved to speech recognition for their reports. At the start, new users corrected 15 characters in a report of 433 characters ⁴. Across their reports, new users corrected 11 characters per report ⁵, against 6 for experienced users ⁶. That is about 1.8 times as many characters corrected per report ⁷.

The rate of words the recognizer did not know leveled off at 0.36 per report, after about 10 weeks ¹. That tells you when the software stopped changing for these users, not when they felt fluent. Not everyone followed the same curve. The authors report that some users had a very unfavorable learning curve, with error rates that did not fall as expected ⁸.

The three studies below measure different things, so their timescales are not one answer to "how long".

The chart compares the three character counts from the radiology study.

The practical reading is that the curve is measured in weeks, not days. For a daily habit, the Lally study found wide variation between people ¹⁰.

Early on, correction takes a big share of dictation, and novices keep fixing errors by voice

In one longitudinal study, navigation and error correction took a big share of the dictation task in the early stages ². Users became more efficient at dictation, navigation and error correction as they gained experience ¹¹. The largest productivity gains in that study came from better dictation quality and from using navigation commands ¹². On this reading, learning to navigate and correct is part of learning dictation itself.

Novices tend to keep correcting by voice. Kraal and colleagues say that this shows an incorrect mental model of dictation ¹³. In the same paper, citing an earlier study, users who keep dictating for about 20 hours start to combine voice correction with keyboard and mouse correction ³. Both the timescale and the mechanism matter here. On this reading, the first hours feel slow because you are learning a new correction habit, not just a new way to produce text.

For a practical rule, speak a sentence, read it back, and fix the obvious problem with the keyboard or by voice, whichever is faster for that fix. You do not need to force voice-only correction. Our reading of the research is that the mixed approach is the normal path: the 20-hour finding describes users moving from voice-only correction to combined correction.

No study we found measures how long the awkward feeling lasts

We searched for research on how long people feel self-conscious when they dictate, and we did not find a study that measures it. This guide therefore gives no number of days for the feeling to fade. Whether the feeling is common, and how long it lasts, is not something the research we found answers.

Some of the difficulty is measurable, though. Speech is fast. In a controlled study on an iPhone, English entered by speech reached 153 words per minute, against 52 words per minute on the built-in keyboard, 2.93 times as fast ¹⁴. The same study found that speech left more errors in the final text: 1.30% uncorrected against 0.79% for the keyboard ¹⁵. The speed is real, and so is the need to read and fix what you produced. That study used short phrases and an older recognizer, so treat the numbers as a rough comparison, not a guarantee for your own speech.

Feeling awkward and making more uncorrected errors are two different problems. The second one has a fix you can practice, which the next sections cover.

Start with short, low-stakes text before long documents

Because correction takes a big share of early dictation, start with text where a mistake costs little. Short messages and quick notes are better first targets than a long report, since each error in a long document has to be found and fixed later. This ordering is our suggestion from the correction evidence above, not a tested curriculum.

Choose a topic you already know. Composing and speaking at once is hard enough without also looking up facts. The guide to how to dictate covers the speaking technique in more detail, and dictation mistakes covers the errors you will meet at each stage.

A daily ten-minute routine suits how habits form

The most direct evidence on daily habits that we found comes from Lally and colleagues. They recruited 96 volunteers, who each chose a daily behavior and repeated it in the same context ⁹. Of these, 82 provided enough data for analysis ¹⁶. Two findings matter for a dictation routine. Performing the behavior more consistently was associated with a better model fit ¹⁷. Missing one opportunity to perform the behavior did not materially affect the habit formation process ¹⁸.

The same study also warns about the limits. Only 39 of 62 individual curves fitted well ¹⁹, so the timing for any one person is uncertain. Its volunteers chose eating, drinking or activity behaviors, not dictation ⁹.

Spacing research points the same way, for memory tasks. A 2006 meta-analysis covered 839 assessments of distributed practice across 317 experiments ²⁰. It found that the gap between study sessions that gave the best retention grew as the period the memory had to last grew ²¹. Dictation is not a memory task, so we apply this by analogy only. Spreading short sessions over days is a sensible default, and the Lally finding suggests a missed day is not a reason to start again.

A routine that fits this evidence might look like this.

1. Minutes 1 to 2: dictate one or two sentences of a message or reminder, speaking at a normal pace.
2. Minutes 3 to 7: dictate a short note on a topic you know. Do not stop to edit while you speak.
3. Minutes 8 to 10: read the note back, fix the errors with the fastest method, and count how many corrections you made.

Use the same time and place each day, as the consistency finding suggests. A fixed slot, such as the first coffee break, is an easy way to apply that finding.

The built-in tools we checked describe no voice profile, unlike older Dragon versions

Some older dictation software, such as Dragon for Mac version 6, asked you to train it on your voice. Nuance's help for Dragon for Mac version 6 says that Voice Training helps Dragon learn how you speak, and asks you to read aloud one of the available stories ²². That is a dated release, and current Dragon products may work differently.

The built-in tools we checked describe no comparable step. Microsoft's voice typing for Windows lists three requirements: a working microphone, an internet connection, and the cursor in a text box ²³. Microsoft says that voice typing uses online speech recognition, powered by Azure Speech services ²⁴. Apple's Mac Help lets you choose the key combination that starts and stops Dictation ²⁵, and the page we read describes no voice-training step.

This matters for practice. On this reading, the learning happens in your speech and correction habits, because the built-in tools we checked describe no profile to train. Practice means repeating the routine and changing how you speak and fix text, not waiting for the software to adapt. For a full answer on training, see do you need to train dictation software to your voice.

Three habits to avoid in the first month

Three habits tend to slow the learning down, and each one follows from a finding above.

The first is forcing voice-only correction. Kraal and colleagues describe persistent voice correction as an incorrect mental model of dictation ¹³, so treat it as a habit to change. If a fix takes a long spoken detour, switch to the keyboard, then go back to dictating the next sentence.

The second is starting over after a missed day. The Lally finding that one missed opportunity did not materially affect habit formation ¹⁸ suggests that one missed day is not a reason to start over. Pick up the next session instead of trying to make up the missed one.

The third is moving to long documents before short text feels routine. Navigation and error correction take a big share of the task early on ², so a long report at the start means many slow fixes in one sitting. Stay with the stages in the table above until the short text no longer needs much fixing.

Track words per session and corrections to see progress

Progress shows up in two numbers: how much you dictate, and how much you have to fix. We did not find a consumer study that prescribes a metric, so the method below is our suggestion. It borrows the measure the radiology study used, which was the amount corrected before a report was finalized ⁵.

• Words dictated per session. Count the words in the text you kept. Over a few weeks, this should rise while you keep the same session length.
• Corrections per 100 words. Count the fixes you made in each session, divide by the words, and multiply by 100. We expect this to fall as you get used to speaking complete sentences.
• Minutes spent correcting. Note how long the fixing takes. Tracking it shows whether the fixing is getting quicker.

Log the numbers in a notes file once a week, so you can see the trend rather than one bad day.

Tie dictation to one task you already do every day

One common approach is to attach a new habit to something you already do. Pick one routine task, such as writing a daily log, replying to messages after lunch, or drafting the first paragraph of a report, and dictate that task in your short daily session. The routine above works best when it sits next to an existing habit.

Keep the keyboard close. The research describes users who combine voice and keyboard correction after extended use, rather than giving up the keyboard ³, so the keyboard is part of the workflow, not a failure of it. If a fix is quicker with the keyboard, make it with the keyboard and move on.

Set a shortcut you can reach without thinking. On a Mac, the key that starts Dictation can be changed in the Dictation settings ²⁵. On Windows, voice typing starts with Windows key plus H ( Microsoft's voice typing page), and voice typing needs the cursor in a text box ²⁶. A shortcut you can press without thinking removes one decision from the routine.

If you want an on-device dictation app for Mac and Windows rather than the built-in tools, Yap, which we make, is one option. The built-in tools above are enough to start, and a free browser option lets you try dictation without installing anything. You can compare prices and versions on the pricing page and the download page.

Next steps

If you have ten minutes today, start the routine with one message and count your corrections. Then come back in a week and compare the numbers.

Each of these answers a question you may already have:

• How long does it take to get used to dictation?
• Why does dictating feel so awkward at first?
• Do you need to train dictation software to your voice?
• How do you practice dictation?

For the wider subject, go back to the dictation overview. Related guides on dictation include how to dictate, dictation punctuation, dictation mistakes and dictation vs typing. If you want to know how fast you can speak compared with type, see speaking vs typing speed. Mac users can start with how to use dictation on a Mac, and Windows users with Windows voice typing.

Terms used in this guide are defined in the glossary: dictation and voice typing.