In our efforts to make knowledge accessible to everyone, if we’re not careful, good intentions can cause us to blunder into useless attempts that benefit no one. I was painfully reminded of this recently when I received a request from a university for an electronic version of my book Show Me the Numbers to accommodate the needs of a student who is blind.

By providing a student who is blind with an electronic version of my book, these well-intentioned folks hoped to make it accessible through the use of “reading software”—software that reads text aloud. I explained to them that books about data visualization cannot be converted into a form that is accessible to someone who is blind because much of the content—indeed, the most essential content—is contained in images that must be seen. They responded by arguing that the Chaffee Amendment gave them the right to convert my book into an accessible form, whether I granted them permission or not, so they would remove the book’s pages from its binding and scan them individually to create an electronic version. When I pointed out that the Chaffee Amendment did not apply in this case because the version of my book that they would create could not possibly be accessible to someone who is blind, they chose to ignore my concern.

I wish I could make the content of my books about data visualization accessible to people who are blind, but I can’t, and technology can’t either. Even if technology existed that could convert a data visualization—an image—into a verbal description, that still wouldn’t solve the problem, for a verbal account of quantitative values in a graph is not a substitute for visual perception. The patterns that are revealed in a data visualization and the operations that are enabled by it (e.g., comparing values and patterns) are not revealed or enabled by words.

Good intentions cease to be good when they produce ineffective results. Sometimes accessibility isn’t possible.

8 Comments on “Inaccessible”

By Andrew. September 19th, 2019 at 11:43 am

Recently I read an article about making video games more accessible to people with varying degrees of blindness. Methods include reducing detail or field-of-view, using higher-contrast palettes, even just making the game easier. Obviously a person with no sight would still not be able to enjoy these games, but there are gamers who, while legally blind, can still make out lights and shapes well enough that they may sufficiently enjoy such a game. In some cases these gamers have to sit with their noses right up to the screen, but it works for them.

I could see electronic versions of your books being potentially useful for people with partial blindness. Maybe reading is difficult and so they may use reading software – obviously that won’t help with visualizations. But it seems that with an electronic version one could blow pictures up and modify the contrast, even put their faces right up to a backlit screen. That could be extremely useful to some people.

By Stephen Few. September 19th, 2019 at 12:46 pm


I understand and agree with the rationale that you just explained for making PDF versions of my books available to people whose sight is only partially impaired. If an enlarged version of a visualization makes it possible for them to see the image clearly enough, then a PDF is useful. I’ve always happily provided PDFs of my books for students with such a disability. I support what’s useful but resist nonsense.

By Kenneth Tyler. December 2nd, 2019 at 2:35 pm

Braille graphs? I know work has been done on translating visual images to a grid of sensors implanted in a person’s back, and this allows the blind to “see” the visual imput to the grid. I think it would be an interesting experiment to get some blind people to participate in.

By Stephen Few. December 2nd, 2019 at 5:53 pm

Hello Kenneth,

I’ve never heard of this research. If you have a link to information about this, please share it with us. I’m finding it hard to imagine how a grid of sensors implanted in a person’s back could simulate the perception of anything but an extremely crude image.

By Jessica. January 28th, 2020 at 6:03 pm

I am also familiar with the research Kenneth mentioned; these are some examples of the research I was able to find:

– A wearable haptic navigation guidance system (Ertan, Lee, Willets, Tan, & Pentland, 1998)

– Human Spatial Navigation via a Visuo-Tactile Sensory Substitution System (Segond, Weiss, & Sampaio, 2005)

– Tactile Sensory Substitution Studies (Bach y Rita, 2004)
[This last one is more of a review / overview]

A lot of visuo-tactile substitution research seems to be focused on wayfinding/navigation, so signalling direction and obstacles or objects (dynamic and static) has been a large part of what these devices seek to achieve.

While there are hard limits with fidelity of haptic input, I could imagine some form of “augmented reality” style of tactile visualisation, where a user could navigate a digitised visualisation through gesture/mouse-over/etc. and receive a stream of haptic feedback corresponding to the location that they are ‘touching’ on the viz. Vibration/pulse amplitude, duration, and frequency can be used to signal different variables, and you could theoretically ‘trace’ lines and such. Auditory input (sonification) could further augment such experiences.

It would never be able to fully replace visual input for data visualisation (especially the more complex the viz), and it would take a lot of work to make it viable for widespread use. But I find the possibilities interesting/exciting to consider.

By Stephen Few. January 30th, 2020 at 11:04 am


I find it difficult to imagine haptic devices functioning effectively for anything but simple data visualizations. The fact that none of the research studies that you cited have been done in the last 15 years, paired with the fact that I’m not aware of anyone actually using haptic devices as a substitute for data visualization outside of a research study, suggest that this research hasn’t produced a viable alternative.

As an example, imagine a relatively simple line graph with four lines, each showing how a series of 12 monthly values changed through the course of a year. Now imagine a haptic device located on someone’s back that provides tactile stimuli for each of the lines. Could the user simultaneously sense all four lines in a manner that allowed those lines to be compared to one another? I doubt it. I can imagine the device tracing each of the lines serially (i.e., one value at a time) across the back, one line at a time, but this would not allow the lines to be compared due to the limitations of working memory.

I’m certainly open to the possiblity of haptic devices serving as a substitute for data visualization, but it seems unlikely in all but simple cases. Every primary benefit of data visualization is enabled by the fact that we perceive visual information in parallel rather than serially. I can’t imagine haptic devices providing stimuli that can be perceived in parallel to the degree that would be required to serve as a substitute for data visualization.

By Valentin. February 5th, 2020 at 3:19 am

I think that you underestimate the ability of blind people to read braille graphs. Even without the use of haptic devices blinds can read complex data visualisations properly converted into braille graphs.
A friend of mine uses his laser cutter/engraver to print Excel charts for his wife. He replaces colors by texture.
The more difficult chart for his wife to read are the one with a lot of colour (area or stacked area chart) or the one with a large number of lines. so not so much the complexity than the volume of different set of data.
For him, the part that he cannot automate is writing the legend and the axes (titles, unit and numbers). But now he tends to engrave the type of graph, the graph title, the legend on a different sheet of plastic or leather that his wife can use as a reference.

By Stephen Few. February 5th, 2020 at 9:04 am


I’m actually not underestimating the abilities of blind people to read braille graphs. Unfortunately, braille cannot serve as a viable substitute for data visualization because it allows the reader to take in information serially, one piece of information at a time. The power of data visualization stems from the fact that we can see the entire graph at once. In other words, a graph printed in braille can reproduce the data that appears in a data visualization, but it cannot reproduce the perceptual experience. The parallel processing the occurs in the visual cortex when looking at a graph, with all the data available at once, cannot be matched by any form of serial processing. I could describe every single value in a graph by speaking words and numbers, one at a time, but that serial process would not duplicate the perceptual experience of a properly designed data visualization.

Leave a Reply