| Harvard University Web Accessibility Standards |

Findings from a survey of Harvard University web sites

In June 2000, David Heitmeyer (Lead Web Technician, Faculty of Arts and Sciences) and Elaine Benfatto (Harvard University Homepage) used diagnostic software* to analyse the accessibility of 717 Harvard web sites.


Given available resources, this was an abbreviated survey intended to assess trends among Harvard web sites, rather than to gather exhaustive data. Only the top index page of each site was analysed during this sampling, and the pages were evaluated by computer only. The recommended visual evaluations of features such as use of color and frames were not performed, again due to resource constraints.


Priority One failures**

# of errors found*** # of sites % of sites
0 173 24%
1 417 58%
2 121 17%
3 6 1%

In these findings, only 173 sites (25%) were found to be free of Priority One errors.

Of the sites that contained just one error, 351 failed only because they did not provide alternative text for images. Thus, by fixing "ALT" attributes alone, we estimate that a large portion of Harvard web sites could be considered Priority One compliant (173 + 351 = 524 sites, 73% of those surveyed).


While we recognize that our survey could not be considered statistically valid, the trends shown in the data are encouraging. Adding "alt" information to <img> tags is a simple (although tedious!) job and could be accomplished relatively painlessly on most web sites. The benefits to our users would be significant. It must be noted, however, that for some sites, meeting Level One checkpoints could involve a serious overhaul to the core HTML coding and/or site design.

* Bobby, from the Center from Applied Special Technology (http://www.cast.org/bobby/)

** Priority One refers to a list of checkpoints defined by the World Wide Web as mandatory for web accessibility. Satisfying these checkpoints will provide basic access to all groups. Noncompliance will prevent access to some information contained on a site.

***Note that any Priority One error would only be counted once per page, no matter how many occurences were found. For example, consider the common error, "Failure to provide alternative text for all images." A page may fail this check dozens of times because it contains dozens of unlabeled images, but it would still be scored for only one error.


Copyright 2001 by the President and Fellows of Harvard College
Email the web staff