John H. Flowers, Department of Psychology, University of Nebraska, Lincoln, jflowers1@unl.edu
A scatterplot is a commonly used visualization format for illustrating the relationship between two numeric measurements or properties of a set of “cases.” Each case is plotted as a point on chart in which distance along the horizontal axis represents the value of one of the measurements, and distance along the vertical axis represents the second measurement. The “shape” of the “cloud” of points illustrates the strength (amount of correlation) and type (e.g., postive or negative, linear or curvilinear) of relationship between the two measurements. The visual inspection of a scatterplot can also permit detection of "outliers", indicative of potential data entry errors or other anomalies in the data that require attention prior to further analysis and interpretation.
How might such information be provided to users who are blind or have low vision?
Could the use of other sensory modalities provide tools for exploring bivariate data that are useful to normally sighted users, as well?
A scatterplot can be designed to be “felt” by plotting the axes a raised lines or edges, and the points as raised “bumps” in a page. A variety of commercial devices are available for producing "rasied line drawing" -- and these can be adapted to producing tactile scatterplots. However such equipment (and usually the special paper or material used to produce the tactile surfaces) is somewhat expensive. Production of permanent displays on such devices is a slow printing process compared with the "click and view" visual displays of modern statistical and graphical software. A more important consideration for blind or low vision users, is that tactile scanning of a scatterplot is a somewhat slow and laborious process (even for users experienced with use of other tactile displays). Our informal experience with introducing undergraduates in statistics classes to tactile scatterplots, also suggests that features such as outlying data points may be missed. Some of these drawbacks may be eliminated by extensive practice and experience.
But ... are there other display formats that might be useful to visually impaired or blind (as well as normally sighted) users?
|
| This example represents a strong positive relationship, indicating that both tests are likely measuring the same underlying skills and abilities. (Pearson’s r = 0.87). Also note that the "gap" in the point cloud is probably more noticeable by listening to the auditory version than by viewing the graph. |
| Click here to LISTEN to these data! |
 |
| A scatterplot for two variables that have a correlation of r = -0.43. Note the outliers. |
| Click here to LISTEN to these data! |
 |
| A scatterplot for two variables that are essentially unrelated to each other (r = -0.03). Note the lack of any trend in pitch change across the cloud of notes. |
| Click here to LISTEN to these data! |
Flowers, J. H., Buhman, D. C. & Turnage, K. D. (1997). Cross-modal equivalence of visual and auditory scatterplots for exploring bivariate data samples. Human Factors, 39 , 341-351.