Morse code for access: what do we know?

Morse code has been used in assistive technology since at least the 1970’s to support typing using one or two switches. This post summarizes what we know about typing performance for Morse code users with physical disabilities, and how it compares to other switch-based text entry methods.

The Morse code keyboard for the Gboard app, showing dot and dash keys.

The Morse code keyboard for the Gboard app, showing dot and dash keys.
Morse code can be an effective way to type using only one or two switches. It’s been around for decades as an assistive technology (AT) that can be used by people with high-level spinal cord injuries (often with a sip/puff switch), severe cerebral palsy, or other conditions that cause significant physical impairments.

This post was inspired by a question sent to the RESNA AT-FORUM listserv by Craig Wadsworth of the Illinois AT Program and Debra and Thomas King, long-time advocates of Morse code. They are trying to gather info from people who are using Morse or have helped someone use it, in order to build a firmer knowledge base about using Morse effectively.

Their question got me thinking about what we really know about the viability of Morse relative to other switch-based methods such as switch scanning.


The main thing I have to offer here is text entry data (typing speed) from KPR’s AT-node site. (For general info about AT-node, check out a recent blog post.) Briefly, Sajay Arthanat and I compiled all the performance data we could find in the literature, for people with physical disabilities entering text, and made these data available in a database called AT-node.

For this Morse code question, let’s see what data are available for Morse code text entry as well as other forms of text entry that might be usable by the same user population. These include switch scanning (with either one or two switches) as well as brain-computer interfaces. We’ll use AT-node to retrieve the available data and see if we can draw any conclusions from it.

We’ll cover each of the three interfaces below. Each section includes a link to see the full report produced by AT-node, and a screenshot that shows the summary portion of the report. Note that the report includes more than just the summary shown in the screenshots– AT-node also gives you a breakdown of performance by diagnosis and other dimensions, a table showing each data point and its characteristics, and a list of all the studies with their citations and abstracts.

You can also search AT-node yourself. All of these searches were done using the User Profile Search from the AT-node home page and specifying the interface of interest.

Morse code

AT-node retrieved four data points covering two individuals using Morse code, as can be seen in the full report on Morse code. Since AT-node is based on a systematic review of research on access interfaces, we can feel reasonably confident that this is all the info there is, at least in the published literature.

The screenshot below shows the top portion of the AT-node report for Morse code. In the basic stats table in the screenshot, we see that there are 4 cases, averaging 6.22 words per minute (wpm), and ranging from 3.39 to 12.39 wpm.

Top portion of AT-node's report on Morse code.  AT-node retrieved 4 cases, with a mean typing speed of 6.22 wpm, standard deviation of 4.16 wpm, and range from 3.39 to 12.39 wpm.

The histogram in the screenshot shows each data point: a cluster of three cases in the 3-5 wpm range, with the fourth case way up in the 12 wpm range. When we look at the data table in the report, we can see that those three clustered data points are from a single individual, a teenager with cerebral palsy, using one-switch Morse code. He achieved a high of almost 5 wpm after extended practice, using a switch with his left toe. The fourth data point, at 12.4 wpm, is from an individual with a high-level spinal cord injury who was an expert user of sip/puff Morse code.

And that’s it for Morse text entry: two studies, from 1986 and 1999, representing two individuals. Clearly we need more data! Yet even this sparse report gives us some frame of reference, demonstrating the possibility of 5 wpm for one-switch Morse by a person with athetoid cerebral palsy and 12 wpm for two-switch Morse when switch use is relatively unimpaired.

Our comparison interfaces have richer datasets. Let’s see what they have to say.

Switch scanning

For switch scanning, AT-node retrieved 80 cases, with an average of 1.72 wpm. Visit the full report on switch scanning to see all the info.

The screenshot below shows the top portion of the AT-node report for switch scanning. With 80 cases, we have a much fuller pattern in the histogram than we did for Morse. The vast majority (over 75%) of cases achieved a typing speed between 0 and 2 wpm using scanning. And there are only four cases above 4 wpm (don’t miss the amazing case way up there at 12.9 wpm — this must be a world record, and I’m not sure I would believe it if I hadn’t seen this individual with my own eyes).

Top portion of AT-node's report on switch scanning data.  AT-node retrieved 80 cases, with a mean typing speed of 1.72 wpm, standard deviation of 1.68 wpm, and range from 0.15 to 12.94 wpm.

The published data on text entry with switch scanning suggest that the most likely speed is below 2 wpm — not a very productive speed! To put that speed in perspective, check out our infographic on text entry rate for people with physical disabilities.

Brain-computer interface

For brain-computer interface (BCI), AT-node retrieved 28 cases, with an average of 0.81 wpm. Visit the full report on BCI to see all the info.

When interpreting these BCI data, note that an interface must be available for public use in order to be included in AT-node. So these retrieved results are for BCI systems that use P300 technology, and do not include study results for systems that are limited to investigational use only.

Top portion of AT-node's report on brain-computer interface data.  AT-node retrieved 28 cases, with a mean typing speed of 0.81 wpm, standard deviation of .33 wpm, and range from 0.4 to 1.74 wpm.

As shown in the histogram in the screenshot, every one of the 28 cases typed below 2 wpm. So BCI systems may well become competitive in the future, but these data suggest that at this point they are still slower than alternatives for this user population.

Any conclusions from these data?

Clearly we need more data regarding the typing performance that Morse code provides for switch users. If we are willing to use the existing-but-sparse Morse data as a basis for comparison, these data suggest that Morse code really is worth a more committed look. At least when looking at data from experienced Morse users, switch scanning has a hard time providing the same typing speed as Morse code.

So Craig’s effort to build a firmer understanding of Morse and its effectiveness makes total sense. We need to know if it truly is more effective for at least some subset of people with severe physical impairments, and if it is, it should be used more often.

Another take-home point from these data is that, unsurprisingly, studies report a wide range of typing performance. For switch scanning, for example, the range is a low of 0.15 wpm up to a high of 12.9 wpm. This almost-100x difference may be extreme but highlights the need to complement research data with direct measurements when evaluating what access method may work best for a given individual. In other words, your mileage may vary.

Measuring an individual’s typing speed with different interfaces can help answer important questions. Where is this person’s current performance falling on the histogram? If performance is lower than that reported for similar individuals, are there opportunities to move toward the higher range? I plan to address the issue of measuring performance with Morse code in a future post.

Experiences of actual Morse code users

Recall that Craig’s original listserv post was looking for the perspective of actual Morse code users with physical disabilities, not necessarily data published in research studies. To that end, I wanted to mention a couple of Morse users who have written about their Morse code use on the web.

Tania Finlayson is a woman with cerebral palsy and long-time Morse user. She has recently been instrumental in Google’s development of a Morse code keyboard for their Gboard app. She offers this perspective on learning Morse code:

It took about a month for me to learn Morse code; yes it was hard, boring, and tiring but it was worth it. It has giving me independence in every aspect of my life and will continue to do so until the end of my life. If you are considering using Morse code as an input method whether it is through my device or another device, I urge you to keep with it even though you might feel overwhelmed and or discouraged at first. It is so worth the work and time; I can’t stress this enough!

Jim Lubin uses sip-puff Morse code, due to a C2 spinal cord injury. His Morse code page has a mix of historical and current information and provides some updated resources on devices and software that support the use of Morse code across a variety of platforms.

Please leave a comment below (or contact me directly) if you’d like to help Craig in his quest to gather info about using Morse code. Also let me know if you’d like a copy of any of the articles retrieved for Morse code, scanning, or BCI.

Leave a Reply

Your email address will not be published. Required fields are marked *