In-person training improves assistive technology outcomes

In a recent research study, people who received in-person training from an occupational therapist had significantly better outcomes with their computer assistive technology, as compared to people who used a home-study program or those who received no training at all. Read on for a summary of this 2019 study from France.

In-person training improves assistive technology outcomes. Two images: one showing an occupational therapist and a person with a spinal cord injury working together. The other shows a close-up of a person typing using a typing splint.

In-person training improves assistive technology outcomes. Two images: one showing an occupational therapist and a person with a spinal cord injury working together. The other shows a close-up of a person typing using a typing splint.

As part of our AT-node project, we regularly search the literature on the use of alternative access interfaces by people with disabilities. This study popped up in our latest search: Influence of training protocols on text input speed on a computer in individuals with cervical spinal cord injury: a randomised controlled trial by Samuel Pouplin and colleagues, published in the Spinal Cord journal in 2019.


The focus of this study is on the use of word prediction as a typing accommodation for people with spinal cord injury, but the implications go well beyond that specific type of assistive technology. The idea of word prediction is that it saves keystrokes, and sometimes speeds up typing, because it tries to guess your word as you are typing. In this study, they used the Skippy word prediction software for Windows, set to display a 6-word prediction list horizontally.

A challenge with word prediction is that it sometimes does not help as much as you might expect. These researchers wondered whether a specific training protocol for word prediction might enhance performance.

Study design

The research team was interested in the effect of 2 factors on text entry rate of typists with cervical spinal cord injury (SCI): (1) training protocol, and (2) word prediction software. They hypothesized that the therapist-supervised protocol would help people use word prediction more effectively and lead to the greatest improvements in text entry rate.

This is a real randomized controlled study, an RCT, which is quite rare in assistive technology! Participants were randomly assigned to one of three training protocols:

  1. REHAB: 45-minute sessions with an occupational therapist, 3x each week for 4 weeks
  2. SELF: 15-minute sessions at home on your own, 5x each week for 4 weeks
  3. CONTROL: no training assigned

They were able to recruit 42 people participants with SCI between C6 and C8, sustained within the preceding 6 months (6 females and 36 males). 38 people completed the full 4-week study, making it the largest study we’ve seen on text entry rate with individuals who have physical impairments.

Typing speed was measured for all participants, with and without word prediction, at Day 0, Day 15, and Day 30. So, in addition to the between-subjects factor of training condition, there were within-subjects factors of word prediction (with and without) and time (D0, D15, D30). Between measurement sessions, participants received training as defined by their training group.

Main take-home points

By the end of the study, what factors actually helped enhance typing speed?

In-person training with a therapist helped

Results suggest that the REHAB training with an occupational therapist helped text entry rate (see table below). All groups started out at roughly the same typing speed and improved significantly with time. After 30 days, typing speed was significantly faster for the REHAB group (both with and without word prediction). We’re not talking about huge gains, but a meaningful gain of almost 20% faster. Satisfaction was significantly higher for the REHAB group as well.

Screenshot of Pouplin et al.'s Table 2, showing main results of their study. By Day 30, REHAB group with in-person training achieved 73 characters per minute vs. about 61 characters per minute for self-training group and control group.
Screenshot of Pouplin et al.’s Table 2, showing main results of their study. Note that speed is in characters/minute; divide by 5 to estimate words/minute.

Home-based training didn’t help

The SELF group, with a home-based self-training guide, did no better than the controls who received no training. The SELF training group did spend an average of 5 hours on training during the month, by self-report, so it appears that they did comply with the training. But this is lower than the 9 hours for the REHAB group, and of course, lacks the interactive in-person coaching that the REHAB group received.

Word prediction didn’t help typing speed

For this group of people with cervical spinal cord injury, use of word prediction did not enhance typing speed, regardless of training protocol. For example, for Day 30, the REHAB group averaged 13.5 words/minute with vs 14.7 wpm without word prediction. This result is not that surprising, given the injury level of the participants, and is consistent with past research in this area (see, for example, Koester & Levine 1996).

Typing speed for people with C6-8 SCI

This study provides some solid additional data on text entry rates achieved by people with cervical spinal cord injury. As noted above, participants had injuries between level C6 and C8, sustained within the preceding 6 months. So they were all fairly new to learning a new way of typing after their injury, but were all experienced computer users prior to their injury. All used upper extremity typing. The paper doesn’t specify whether unilateral or bilateral, but describes the typing method this way: “All participants accessed the computer using a standard keyboard and either pressed keys using the meta-carpophalangeal joints of their fifth digits bilaterally with the forearms supinated, or used typing splints.”

I made this little line graph using Pouplin’s data, to try to give a more visual perspective on the results. These typing speeds are for letters-only typing (without word prediction), since that was the faster typing condition for all groups, regardless of training. The groups without in-person training (SELF and CONTROL) stayed at around 12 wpm for the entire study, while the in-person training group (REHAB) got up to almost 15 wpm by Day 30. This suggests that typists with C6-8 SCI can type about 15 wpm, within the first few months of injury.

A graph I made from Pouplin et al.'s data. By Day 30, the Rehab group that received in-person training achieved faster typing speed (almost 15 words per minute) than the other 2 groups (about 12 words per minute).
A graph that I made from Pouplin et al.’s data. By Day 30, the Rehab group that received in-person training achieved faster typing speed than the other 2 groups.

Use AT-node to compare to other research data

We can compare this to data from other studies by using KPR’s AT-node for Access website. We’ll search AT-node for all data points involving typists with cervical SCI using a physical keyboard.

The pattern is pretty similar to the 12-15 wpm seen in Pouplin’s study. With AT-node, though, we get to see individual variation, in this case across 29 individuals. There is a pretty huge range in typing speed: from 3.7 to 35.5 wpm. So don’t assume that a given person will automatically type 15 wpm: be sure to measure to see how things are going.


To sum up the key points:

  • In-person training with a therapist improved typing speed.
  • Home-based training with a non-interactive instructional document didn’t.
  • No matter what training method participants received, word prediction didn’t help typing speed for these typists with C6-8 SCI.
  • These typists averaged 13.2 wpm, typing letter-by-letter with their pinky finger knuckle(s) or typing splints, with the fastest group achieving 15 wpm after 30 days of in-person training.

These results suggest that supervised training should be provided to all individuals who are prescribed with alternative access methods to get users off to a good start and facilitate better performance.

Consider using these results as a resource the next time you are asked to justify in-person training for access technology.

4 thoughts on “In-person training improves assistive technology outcomes”

  1. Great find, thanks! This confirms what I have found anecdotally for years: You can’t just mail AT that takes some skill to someone and hope it works out-they need someone to work with them. I see this often when I see how poorly folks operate Dragon when going solo. It also confirms that word prediction isn’t a speed tool, but perhaps an anti-fatigue tool for some and a spelling tool for others. It’s also good to know that 15 wpm is a good benchmark for people in this group.

    1. Thanks, Paul! Always good when research results align with real-world experiences!
      It’s interesting that participants in the in-person training group improved in regular letters-only typing, too (not just with word prediction). I think there’s just something about that one-on-one connection and accountability, in addition to the specific content of the training. Maybe similar to piano lessons: some people can get good on their own, but usually things go much better when you have a teacher and regular lessons.

  2. Interesting study! Thanks for sharing this information. With the increasing prevalence of online learning and remote meetings, it is good to be reminded that in-person training is more effective than other methods. I would assume this is true for other groups and topics, but as you mentioned, studies of AT usage are difficult to find.
    I am curious whether the study participants found word prediction helpful for reasons other than typing speed.

    1. Thanks for reading, Kimberley! I agree that it would be interesting to know whether participants found word prediction helpful for reasons other than typing speed. They may have, but looking back at the paper just now, it’s not really presented in this paper. There were significantly fewer errors per 10 minutes of typing when using word prediction vs without (about 11 errors with vs 14 errors without). They also measured participants’ ratings of speed, cognitive load, and satisfaction, but they didn’t compare these ratings with vs without word prediction (in this paper, at least). Their focus was on the 3 training groups — satisfaction was significantly higher for the in-person training group; no significant differences found for speed or cognitive load ratings.

Leave a Reply

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