I’ve been thinking about eyetracking as a useful indicator of reading, and how other indicators may be used to track user activity on the web. Many eyetracking studies presume to know when reading occurs via movement patterns. Even accepting the eye-mind hypothesis, Jakob Nielsen reminds us, “By itself, eyetracking can do one thing and one thing only: discover what users look at. (2010, 10)”. And yet our understanding of the automatic and unconscious eye movements in reading is fairly sophisticated, and can even provide prediction of fixation location and saccade lengths based on an analysis of the text itself (this is one way that computational models of reading are tested). So though we have not seen a comprehensive theory of eyetracking reading, there appear to be strong relationships that allow for reliable prediction.
The complexity and automaticity of eye movements when reading is such that it seems unlikely that a human could trick eyetracking into thinking it was reading when it was not. However, I personally have experienced a kind of reading without comprehension, a movement through the text, recognizing the textbase without comprehending, because my mind was on other matters. Friends have described similar phenomenon, when tired, when distracted, when disconnected from the text. I have not examined literature on the question of “reading” without comprehending, and how that might look compared to reading with comprehension; however, I suspect a robust enough algorithm could detect anomalies in oculomotor activity in disconnected reading.
But this particular question is a bit far afield from my original pondering of the logic of the assumptions we make about eye tracking. Thinking through this logic may allow me to then extend these to other possible indicators of reading, such as computer user interactions.
As with all theories, we can not logically verify that reading is occurring by observing user activity; we can only hope to falsify that reading is occurring. However, I am optimistic that through consistent lack of falsification we can come to a reasonable, inductive assumption about when reading is occurring, and substantiate that assumption with assessment of reading comprehension.
Eye tracking falsification syllogism
If one is reading*, one’s eyes will move in “reading patterns”.
Your eyes are not moving in “reading patterns”.
You are not reading*.
*Reading here means the current mode of reading as rauding, where the eye moves through the text fairly consistently and completely while the text remains relatively static. RSVP reading is another matter.
Presuming one had no ability to track oculomotor activity, one would have to rely on other possible indicators, which also can only rely on falsifications. For instance, looking at home PC reading on the web:
If one is reading on the web, …
…the text must remain visible.
…cursor activity will not suggest interest in navigating the site.
…time spent on the page will be no shorter than (formula based on words, difficulty, reading speed).
…windows other than the window containing the text will not gain focus and receive interaction.
…lack of activity on the the page will not reach a “time out” threshold(?).
…scrolling occurs when text extends below “the fold”.
Still not good enough! Even if a method of indicating reading without eyetracking (or other more invasive, lab-based monitoring techniques) may be feasible from a falsification point of view, such a method is still subject to a number of confounding variables. Yet such a method may be substantiated or validated by an assessment of reading comprehension; it may additionally be substantiated by cursor movement if a correlation between eye movement and cursor movement patterns when reading is shown to exist.


Challenges for online reading researchers
Some thoughts re. future challenges and opportunities for researchers interested in online reading that were burbling as I drove to Layton today:
* reading detection without eyetracking
* eyetracking in mobile computer device
* ecologically valid studies of reading for learning; more authentic environments and tasks
** quasi-experimental research of online users in their natural environments engaged in real tasks
* relationship of Csikszentmihalyi’s flow to reading (offline or online)
* usability/ui design for electronic reading (spatial orientation, readability)
* distraction/attention in online reading environments (hyperlinks, media, accessibility of external stimuli, user’s mental models and preconceptions of “online” v. “offline” activity)
* annotation/note-taking interaction (review of current tools)
* permanence of user document metadata (e.g. annotations, note-taking) (permanence, stickiness, context, import/export)
* descriptive research or phenomenology of online reading behavior for learning (students and academics)