Skimming and scanning behaviors in reading digital information with screen readers 

Sighted web users rely on visual skimming and scanning to consume text heavy information without reading complete text. We wanted to understand how non-sighted web users use the screen reader to consume text heavy information on the web, especially in the absence of semantically designed accessible webpages

NOTE: Classroom based research for course: Accessibility and design for diverse users at DePaul University, conducted by a group of 2 students. My role: Literature review and research questions, research protocols, data collection, qualitative analyses of data for insights

 what quick reading methods do screen reader users use? 

Do our web interfaces afford effective non-visual information consumption?

  • Sighted individuals consume information using visual quick reading methods such as skimming and scanning to "enhance the rate of reading without duly compromising comprehension or retention of information"[source]

  • Research has established that people who are blind, and use screen reading technology, navigate web-based content sequentially, often first through headings and links (scanning navigation), following relevant information, skipping irrelevant information [source].

  • Much of the web is not designed and built semantically, such that meta information like <h1>Heading or <p>Paragraph tags are often not available to screen readers. A study of 7.1 million index pages showed that a minor percentage of webpages use semantic tags such as section, nav, header, footer in their HTML markup[source]

  • There is a need to separate non-sighted users' adaptive information consumption strategies from the strategies to escape inaccessible content.

 scenario based observations and inquiry 

We conducted scenario based observations to understand how screen reader users skim and scan the contents of a webpage

4 screen-reader users demonstrated reading tasks based on timed hypothetical scenarios, which allowed us to observe motivations and behaviors. Additionally, some test considerations included: 

  • Test webpage had missing semantic tags and varied type of content. This allowed us to observe adaptive browsing strategies which may not be needed to explore semantically organized websites. 

  • Tasks were demonstrated on computer devices, softwares and speech rate that participants typically use, such that comfort of web access is not a variable. 3 participants used JAWS (Job Access With Speech), and 1 used NVDA (NonVisual Desktop Access) on their personal or work computers

3 scenarios & tasks to simulate quick reading

graphic illustration shows website with varied type of content, such as wikipedia. Task 1 includes participant exploring whole webpage. Task 2 includes participant exploring a subsection. Task 3 includes participant finding something previously explored on this page.

TASKS in limited timeframes

1

scanning body of information

Participants explored the given webpage to understand what type of content is available

2

skimming a smaller section of text

Participants explored the contents of a section of the page

3

scanning previously explored sections to find desired content

Participants looked for 2 instances of content familiarized in the previous tasks

caption: webpage with varied TYPE OF content - headings, subheadings, paragraphs, sidebars, links in text/lists/navigation, tables, banners, figure captions.

Data was collected in the form of observation notes, followed by inquiries after each task. Participants explained why they used specific browsing methods in context of the intent of the task. We used coding and affinity mapping techniques for analysis.

 findings | Adaptive quick reading strategies 

We observed screen reader users employing various adaptive browsing strategies to consume information selectively. These techniques helped parse content quickly, without reading entire text.

1. Scanning body of information to gain a high-level picture of the content 

Participants accessed page headings and links, similar to the "scanning and gambling navigation" reported by Takagi et al. [source].

  • One participant mentioned that Heading navigation helped "quickly go through all the content". Another mentioned using Headings to understand "high-level layout of the page". 

  • One participant described using Link navigation, "links provide context, are important especially when I am trying to get to a particular part of the page”. 

  • In the event where these techniques did not provide enough information, participants mentioned that they would scan tables, figure captions, demonstrating an "item-level" scanning as reported by Takagi et al.[source].

On illustration, content heavy webpage shows a red border on parts of text selected by user and read by screen reader. Participant used Heading navigation to select headings, and  down arrow key to read by line when heading was relevant. Another shows borders on text selected using link navigation, including links in sidebar menu, global menu on top, in text links, listed links.

caption: while scanning a page using heading, link or sequential navigation, participants remembered meta information such as relevant keywords, order of relevant heading or link, first letter or word of relevant link or sentence

2. Skimming the contents of a relevant section

Participants used keywords, or "information scent"[source] gathered during page navigation, to stop the scan and explore more content by skimming that area. Within a chunk of information, such as a paragraph, participants demonstrated selective reading by actively skipping undesired content.

Illustration shows red border on parts of text selected by user to read on page. Some words in lines are skipped when deemed irrelevant, and complete line is read when relevant to information needs.

caption: participants read selectively by skipping complete sentences. for relevant information, they remembered meta information such as order of line within section, first word of sentence.

3. Jumping to relevant sections using memorized meta information, and skipping navigation

Participants used the meta information they gained during initial scanning, to directly jump to desired content. 

Illustration shows that after previous link navigation on page, user conducts another link navigation. Then they skip irrelevant links, and select relevant one using remembered meta information such as first word, first letter, or order of line within section, or # lines away from heading.

caption: using item navigation or sequential navigation, memorized meta information, and skipping to reach desired content

How might we leverage these non-visual skimming and scanning methods in navigation?
Users' information intent could be leveraged to present relevant transcoding solutions based on the quick reading methods they may use.  

Some low-tech design considerations could include:

  • Bookmarking for scanning navigation to mark relevant items which were used as landmarks (landmark navigation as defined by Takagi et al. in research).

  • Bookmarking relevant keywords in text, to retrieve those later.

  • Allowing selection of relevant text while skimming or scanning, easing load on memory in remembering meta information (order of link, order of line etc). The selected text could be accessed when desired.

  • Demarcating sections of a webpage, such as main menu, sider bar menu, main content, so that irrelevant content could be skipped without scanning.

  • Leveraging "item-navigation", providing relevant meta information to provide a table of contents such as xx headings, xx sub headings, xx links.