The initial results of our screen reader survey are now available – http://webaim.org/projects/screenreadersurvey/
With over 1100 responses (thank you to all that participated!), the data provides a load of useful information about screen reader user demographics and preferences. Some of the results were about what we expected, but some of them very much surprised us. We are already re-thinking some of our long-standing philosophies about alternative text – particularly for decorative-type images.
We understand that the survey design was not perfect and that this may not truly represent all screen reader users, but we believe the results can be very informative for developers wanting to create accessible web content.
Here’s a small sampling of a few findings:
- The most common screen readers used are JAWS (74%), Window-Eyes (23%), NVDA (8%), and VoiceOver (6%).
- 74.6% of screen reader users upgrade to the newest version within one year.
- 12% of respondents use a screen reader on a mobile phone.
- 76% of users always or often navigating by headings.
- 36% never or seldom use text-only versions of web pages.
- 71.5% of screen reader users reported that Flash is very or somewhat difficult.
There is much, MUCH more to be learned about this diverse group of screen reader users at our survey results page.
If you have feedback or questions about the survey, please post them here. We will be posting more in-depth analysis and details on the free-form text responses in the near future. We anticipate doing more of these types of surveys – if you have recommendations or questions you would like answers to, please let us know.
not many big surprises in there although a bit shocked there there were no Hal or supernova uses in the survey, the former being a pretty popular screenreader here in the UK, second to jaws I’d say and certainly in wider use than window-eyes, supernova is the same screenreading technology but with a magnififier built in and is almost certainly teh most common reader/magnifier combo in the UK much more so than combining jaws with magic etc. The flash question I’d like to see explored further, in my experience most flash is either easy or impossible but rarely anything much in between. sometimes on a content rich page you just have to try a bit harder or take little longer to get to where you want to be but it isn’t usually like tht with flash. and with pdf files I suspect that many difficulties are to do with whether the pdf file opens in your pdf reader app of choice or in the browser. I’ve found most people cope just fine with pdfs in teh appropriate app but get frustrated if for example they can’t get out of or close a pdf window within the browser. and pdf still suffers from a poor accessibility reputation quite some years after it deserves to, it’s amazong how many screenreader users won’t attempt to open pdf files because of a problem they had 5 years ago – although quite possily those arent’ the users who’d have made it to and through this survey.
Adrian-
There were quite a few Hal and Supernova users. We just haven’t added them up yet. The numbers were not in the top 4 of screen readers for respondents.
Thanks for your comments.
I was initially skeptical because of it not being a controlled sample. I would have expected the survey to elicit a higher than average response among more advanced users and for the results to be skewed accordingly. So it was interesting to read how some of the results break down by screen reader proficiency. Though often not the way I would have expected.
I think this is a key issue because you have the following self reported proficiency levels:
Expert 17%
Advanced 41%
Intermediate 32%
Beginner 9%
I work for the National Council for the Blind of Ireland and I asked our main user technical support worker how he thinks it breaks down among Irish screen reader users. Totally unscientific of course, but his estimates are:
Expert 5%
Advanced 10%
Intermediate 20%
Beginner 65%
The differences here are quite striking. Only 41% of your respondents stated that they are beginners or intermediates, compared with our estimate of 85% among the Irish screen reading population as a whole.
Following on from your conclusions that screen reader users are a diverse bunch with diverse needs, preferences and strategies, this opens up questions concerning who should we most cater for? For example, I’m reminded about the debate around site-specific text resizing widgets. Opponents say that users should be taught or expected to learn to use their browsers’ built in controls. ‘ve even read the view that users should be expected to use their own stylesheets. But if 65% of users are ‘beginners’…?
Mark-
Thank you very much for your insight. We fully understand that the survey respondents are likely to be more technically and screen reader proficient. And certainly beginners will not be well represented, though if I were to make an estimate, I think much fewer than 65% of the screen reader user population is more advanced than beginner.
Despite these potential (and totally uncontrollable?) issues with the online survey, the fact that so many responses are what they are is perhaps even more insightful. For instance, if 71% of respondents, who are admittedly more proficient, find Flash difficult, consider what the number would be if the respondents were truly representative of the less technical audience.
I am fond of PDFs when done well (I use several screen readers). The vast majority of PDFs I read are untagged. The majority of them read ok, but some are very messy. It’s a similar issue with Flash. While Flash done well has much potential for screen reader users, far too often buttons are not labelled.
Andrew
Andrew-
Well said. Flash and PDF and arguably any other technology is only as good as the accessibility techniques implemented into them. Andrew Kirkpatrick from Adobe has some perspective on this at http://blogs.adobe.com/accessibility/2009/02/webaim_screen_reader_survey_a.html
I would like to see the question of phone numbers asked in a future survey – what format do they prefer phone numbers to be? (555) 555-5555? 555-555-5555? 555.555.5555?
I found the preference for “Photo of White House” interesting. Could it be that more experienced users elect not to have the tag read to them? So instead of getting the redundant “Image: Photo of White House” they just get “Photo of White House” which would obviously be preferred over simply “White House”?
I have to disagree on the point re similarity between PDFs and Flash – it rather depends on how the PDF files are generated but in the main if a pdf is built from an electronic source (EG if it isn’t a physical item that is scanned in) then it does contain true text and thus, like html, no matter how little tagging or structure is applied you can still at the most basic level read the words where as with Flash the content is not readable unless accessibility is pro actively supported. you can read an untagged pdf, you can not work with untagged Flash.
I still place more responsibility on the authoring tools than I do on the authors. Making it “easy” to generate accessible content via prompts and validation should be the norm.
Training in accessible authoring is almost unheard of from most vendors. Attending a VS.Net course solidified this opinion because the instructor knew nothing about accessibiliy design principles. The same with Adobe LiveCycle “info-demos”. They all stress how easy it was to post content rather than posting accessible content.
They should focus more on what you should do than how easy you can do it.
Jared,
Are you aware of any translations of the survey? This was discussed on the mailing list, but eventually nobody came forward saying: “Hey, we have a translation at…”
There are no translations of the survey that I’m aware of. For the next survey we provide, we hope to provide translations into a few different languages. There are several translations of the survey results that will be available in the near future.
Excellent survey — I hope you will do more of them, and with even more detail. I will not echo the comments of others, but I will reinforce the value of representative sampling when you consider the policy implications. It’s easier to convince a company or public agency to change its behavior if you can extrapolate results to a large population.
Perhaps the self-reporting of expertise could be refined to ask about years of use, frequency of use, etc. Another simple technique is to ask “Which happens more often: you getting tips from people you consider screen reader experts, or other people getting tips from you?”
Regarding PDF and Flash, you’ve uncovered the “accessibility value chain” issue — users only get the benefit of accessibility work if both the tool making and the content creating are done right. We’ve got this same problem in learning management systems, IVRs, video production, voting machines — everywhere.
There are differences of opinions about people’s experiences within these 2 media formats. It might be worthwhile to do what you did with alt text, frames, etc., and create some examples of each format with certain characteristics, loose users on them, and collect the experiences. You might find that there are patterns to the resulting success and failure that could inform tool makers, trainers, content developers, screen reader companies, etc.
But bravo for now — you’ve moved the field forward with this high-volume study!
As Steve, I wonder about phone numbers. I wonder if formats such as “Phone area code 555 number 5555555″ or ” Phone with area code 5555555555″ would help in any way.