Metatron and eVRydayVR present VR Speed Reader, an application for rapidly reading text documents in VR. Based on Rapid Serial Visual Presentation (RSVP), VR Speed Reader rapidly presents words from a document embedded in virtual space; this both limits eye movement and discourages bad habits like subvocalization (reading the text in your head) and regression (re-reading portions that were already read) which harm reading speed. Compared to screen-based RSVP apps like Spritz and Spreeder, VR Speed Reader makes more effective use of peripheral vision to provide essential context, and uses head tracking to provide a simple user interface that can be used to configure the reading rate without interrupting reading.
The average adult typically reads printed text at 300 WPM or less, but contestants in the World Championship Speed Reading Competition routinely reach WPM rates in the thousands. What accounts for such variability in reading speeds? Some of the limitations on reading speed include:
- Subvocalization: the act of engaging the muscles associated with speaking without producing audible words, limiting the reading rate to the reader’s speaking rate;
- Regression: re-reading text that was already read earlier;
- Eye movement/saccades: the time needed for the eyes to move between words.
Rapid Serial Visual Presentation (RSVP), used by monitor-based speed reading apps, combats these issues by presenting a stream of single words at a high rate, faster than the reader can subvocalize. There is no need to move the eyes, and no opportunity to regress. However, compared to normal reading, this also removes contextual clues in peripheral vision, and removes the reader’s control over their reading rate.
Virtual reality presents the opportunity to get the best of both worlds by solving these issues within the RSVP paradigm. The center of the user’s field of view still focuses on a single word at a time, but the remainder of the HMD’s wide field of view can be filled with the target text, maintaining the comprehension cues associated with paragraph context and sentence structure even at high WPM rates. Although this text cannot be read, it helps the reader to both recall the structure of the text they have just read and predict the structure of upcoming text. Additionally, VR allows us to solve a common issue with RSVP, in which adjusting options such as reading speed on-the-fly requires pausing the reading task; instead, subtle head tracking controls such as rotating the head slightly left or right can be leveraged for this purpose. The eyes’ natural counter-rotation during head movement keeps them focused on the text in the center of the field of view so that no words are missed.
Controls are based solely on gaze and the touchpad (with gamepad support planned for a future release). The initial prototype features three fixed sample texts of varying textual complexity to demonstrate the core technology, which are selected from a 360-degree main menu using head tracking and the touchpad. During reading, in addition to adjusting speed via head tracking left and right, the touchpad can be tapped to pause reading and display the pause menu, which shows the current reading speed. Speed can be adjusted while paused. Because it is possible for the user to accidentally misconfigure the application to a too-high speed using the head tracking controls, the pause menu includes options to return to an earlier point in the text. Finally, the back button may be used to return to the main menu at any time.
We use a simple language model to define the amount of time spent on each word where word width in pixels is used as a proxy for word complexity/familiarity, and more time is spent on more complex words, allowing higher overall rates to be achieved; future work will develop more sophisticated statistical language models based on corpuses and context.
The current application has a simple streamlined interface with no options other than speed, and is sufficient for most users, allowing them to quickly begin the reading task. To accommodate a wider range of users and explore further improvements, we will in the future permit other experimental options, including adjusting text size, the use of supplemental cues such as color, adjusting curvature of the reading surface, and so on.