Association for Business Communication
78th Annual International Convention
New Orleans, 2013
Tharon Howard,
Clemson University UsabilityTesting Facility Director
tharon@clemson.edu
Part One: An Introduction to Eye Tracking Foundational
Concepts and Two Approaches to Website Navigation
(Tharon)
Part Two: Webby Awards and an Eye Tracking Study
(Heather)
Part Three: Limitations and Strengths of Eye Tracking
(Brian)
Fovea
Saccades
Fixations
Parafovea
How FovealVisionWorks
Example of FovealVision
Source: http://themetapicture.com/it-was-worth-it/
FovealVision
Instead of:
Credit: Dan Liddle, Clemson UTF
FovealVision
We can only see:
Credit: Dan Liddle, Clemson UTF
FovealVision
We can only see:
Credit: Dan Liddle, Clemson UTF
FovealVision
We can only see:
Credit: Dan Liddle, Clemson UTF
FovealVision
We can only see:
Credit: Dan Liddle, Clemson UTF
Saccades and Fixations
Can be as short as 20 milliseconds or
as long as 200 milliseconds
The fastest movement produced by
the human body
Substantial image blur during a
saccade due to this speed
fixation fixation
saccade
Fixation: Eye-movement pauses to acquire
Content
Saccades: The period of eye-movement from
one fixation to the next
AlfredYarbus –TheVisitor (1967)
Saccades are not
arbitrary; they’re
influenced by
parafovea vision
Users’ tasks and
goals play a
significant role in
the gaze plot
“Depth of Processing” Gaze Plot
• “Mechanical” or Search Coil
• “Glint” or Dual Purkinje
• Limbus or PupilTracking
• Dual-PointVideo
Popular EyeTracking
Technologies to Consider
Search Coil & Magnetic Field
or “Mechanical”
Uses contact lenses
with search coils and
a magnetic field
Usually combined with reflected light
Extremely accurate
Very costly
Uncomfortable & not portable
Source: http://www.iovs.org/content/44/1/275.long
Dual Purkinje Image
or “glint”
Uses infrared light to create reflections from
different parts of eye
Video tracks 2 reflections and software
calculates angles
Good accuracy
Comfortable for users
Allows more head movement
Still costly
Source: http://www.fourward.com/dconcept.htm
Source: http://medical-dictionary.thefreedictionary.com/Purkinje-
Sanson+mirror+images
Limbus or PupilTracking
or “Single-PointVideo-Based”
Image processing of video is used to locate pupil
and calculate POG
A Starburst algorithm projects rays from an initial
starting point detecting pixels with high image
gradients
Source: Howard & Sulak
Example of rays projected from initial point
Source: Ryan, Duchowski, & Burchfield
“Dual-PointVideo-Based”
Image processing tracks pupil and corneal reflection
of IR light is also tracked
Less expensive
Comfortable
Head movement is limited
Only accurate to 1-2 degrees (1o is about 0.5”
at 27” distance)
Example of rays projected from initial point
Source: Ryan, Duchowski, & Burchfield
Source: Howard & Sulak
Some OtherTerms to Know
Accuracy—Measured in degrees and typically ranges from 0.5
to 2 degrees. Keep in mind that 1o is about half an inch of
resolution at the normal 27” viewing distance for computer
monitors.
Precision—Smallest change in eye position that can be
measured. Usually this is also measured in fractions of
degrees.
Sampling Rate—Measured in Hertz, this factor is important
because eye movements are the fastest humans can produce.
Over 1000 Hz is needed to capture many eye movements, but
most systems typically run at 50 to 60 Hz.
The “F Pattern” and Golden
Triangle
Source: http://blog.mediative.com/en/2011/08/31/eye-
tracking-google-through-the-years/
Source: http://www.nngroup.com/articles/f-shaped-pattern-reading-web-content/
“Eyetracking visualizations
show that users often read
Web pages in an F-shaped
pattern: two horizontal
stripes followed by a vertical
stripe.” Jakob Nielsen
“We recorded how 232 users
looked at thousands of Web
pages.We found that users'
main reading behavior was
fairly consistent across many
different sites and tasks.”
Found consistent pattern
from 2005 - 2011
Rosenblatt and Web Page Genres
Efferent(to carry away) Aesthetic (emotional pleasure)
“In readings that fall somewhere in the efferent half of
the continuum, the reader selects out predominantly
more public or cognitive elements than private.The
aesthetic stance, in contrast, accords selective
attention to predominantly more of the penumbra of
private feelings, attitudes, sensations, and ideas than
to the public aspects.” Louise Rosenblatt
Source: “TheTransactionalTheory:Against Dualisms.” College English,Vol. 55, No. 4. (Apr., 1993), p. 383.
Rosenblatt and Gaze Plots
Efferent(to carry away) Aesthetic (emotional pleasure)
Do the GoldenTriangle and F-Pattern
obtain for “aesthetic” sites as well as
for “efferent” sites?
Tharon Howard, Clemson UsabilityTesting Facility Director
tharon@clemson.edu
Part One
Honoring the Best of theWeb
Source: http://www.grumpycats.com/grumpy-cat-wins-meme-
of-the-year-at-the-17th-annual-webby-awards/#.UmiDsyToVWZ
Source: http://www.webbyawards.com/press/
Source: http://www.morethanamap.com/
Google More than a Map – Efferent
(Transactional)
HuffPost Live - Aesthetic
Source: http://live.huffingtonpost.com
EyeGuide®Tracking System
EyeGuide® Hardware:
• Camera
• LED
• 2.4 GHz transmitter
Source: http://www.otshow.com
Source: Howard,Christiansen & Gaines
Calibrating the EyeGuide®
Source: Howard,Christiansen & Gaines
Source:T. Howard
Research Question:
• Do users' eye movement patterns differ between websites which are
designed to provide users with information (efferent/ transactional) versus
websites which are designed to provide users with aesthetic experiences?
Preparing for the Study
Participants:
• 21+ in age with a bachelors degree or higher
• Recruited a convenience sample
• Goal of 10 participants
• Approximately 30 – 45 minutes/ participant
Fit participant with the head-mounted eye tracking device.
Study Procedures
Calibrate the eye tracker.
After a satisfactory calibration, begin tasks.
Participant is shown a series of ten screen captures from website interfaces.
Each screen capture is assigned a specific task.
While completing tasks, the EyeGuide® software captures eye movement data.
Stimulated RetrospectiveThink Aloud: after completing the tasks, the
participant is shown gaze plots of his or her eye movements and asked a series
of questions.
Gaze Plots
Source: Howard,Christiansen & Gaines
Heat Maps
Source: Howard,Christiansen & Gaines
Stimulated RetrospectiveThink Aloud
Source: Howard,Christiansen & Gaines
Task-based eye tracking - the Golden
Triangle and F Pattern are still present,
but current web conventions tend to
dominate the gaze.
What DidWe Learn?
Source: Illustration adapted fromYarbus
1967, figure 109, for Land andTatler 2009
“Commenting Policy”Task
All Participants
Source: Howard,
Christiansen & Gaines
Source: Howard,
Christiansen & Gaines
Heather Christiansen, PhD Student
Rhetorics,Communication and Information Design
Clemson University
hdunn@clemson.edu
Part Two
Brian Gaines,
Graduate Research Assistant,
Clemson University
blgaine@g.clemson.edu
EyeTrackers can be expensive.This
usedTobii T120 system is $27,500.
(Image courtesy of eBay, inc.)
SensoMotoric Instruments (SMI) Eye
Tracking Glasses 2.0 are priced on a
case by case basis.This could
potentially be quite costly. (Image
courtesy of SensoMotoric Instruments,
GmbH)
Price
Lab costs, as well as the costs for participants can be costly. For example,
Eye Tracking, Inc. charges between $5,000-$250,000 per study.These costs
are determined by the type of study, length of study, number of participants,
etc.
Price
The Grinbath Eye Guide is roughly $2,495, with discounts
applied to multiple unit purchases. (Image courtesy of Grinbath,
LLC)
Affordable options are available.
Eye Tribe’s USB 3.0
Hardware is only $99, but
comes with a Software
Development Kit (SDK),
which means code must be
written in order for this
device to be used. While the
base price is attractive, the
time and cost of hiring a
software engineer is
something to consider.
But, how affordable is affordable?
Equipment Sensitivity
Because of the sensitivity of the equipment, the
participants were uncomfortable. (Image Courtesy of Brian Gaines)
Equipment Sensitivity
Calibrating the eye tracker is in itself an art and a science.
(Image courtesy of Tharon W. Howard, Ph.D.)
Equipment Sensitivity
Differences in Participants
Retrospective think aloud is a technique used in usability,
and eye tracking in particular, to gather qualitative
information on the user intents and reasoning during a
test. It's a form of think aloud protocol performed after
the user testing session activities, instead of during them.
Fairly often the retrospective protocol is stimulated by
using a visual reminder such as a video replay.
(Guan, Lee, Cuddihy, and Ramey, 2006)
Stimulated RetrospectiveThink Aloud
More than 80% of subjects’ verbalizations of what they
were attending to corresponded with the eye movement
data. We reject the notion of subjects’ fabrication since
only less than 3% of their verbalization failed to match up
with objects identified by their eye movement.
(Guan, Lee, Cuddihy, and Ramey, 2006)
Stimulated RetrospectiveThink Aloud
Immediately following the eye tracking session, the data is available.
Instantaneous Results
Instantaneous Results
Cons
• Certain eye tracking technologies are
expensive
• Lab/participants costs
• Equipment sensitivity
• Accommodating for differences in
participants
Pros
• As technology advances, prices
will drop
• Stimulated Retrospective Think
Aloud is a valid data collection
method
• Results are intantaneous
Conclusions

How eyetrackerswork

  • 1.
    Association for BusinessCommunication 78th Annual International Convention New Orleans, 2013 Tharon Howard, Clemson University UsabilityTesting Facility Director tharon@clemson.edu
  • 2.
    Part One: AnIntroduction to Eye Tracking Foundational Concepts and Two Approaches to Website Navigation (Tharon) Part Two: Webby Awards and an Eye Tracking Study (Heather) Part Three: Limitations and Strengths of Eye Tracking (Brian)
  • 3.
  • 4.
    Example of FovealVision Source:http://themetapicture.com/it-was-worth-it/
  • 5.
  • 6.
    FovealVision We can onlysee: Credit: Dan Liddle, Clemson UTF
  • 7.
    FovealVision We can onlysee: Credit: Dan Liddle, Clemson UTF
  • 8.
    FovealVision We can onlysee: Credit: Dan Liddle, Clemson UTF
  • 9.
    FovealVision We can onlysee: Credit: Dan Liddle, Clemson UTF
  • 10.
    Saccades and Fixations Canbe as short as 20 milliseconds or as long as 200 milliseconds The fastest movement produced by the human body Substantial image blur during a saccade due to this speed fixation fixation saccade Fixation: Eye-movement pauses to acquire Content Saccades: The period of eye-movement from one fixation to the next
  • 11.
    AlfredYarbus –TheVisitor (1967) Saccadesare not arbitrary; they’re influenced by parafovea vision Users’ tasks and goals play a significant role in the gaze plot
  • 12.
  • 13.
    • “Mechanical” orSearch Coil • “Glint” or Dual Purkinje • Limbus or PupilTracking • Dual-PointVideo Popular EyeTracking Technologies to Consider
  • 14.
    Search Coil &Magnetic Field or “Mechanical” Uses contact lenses with search coils and a magnetic field Usually combined with reflected light Extremely accurate Very costly Uncomfortable & not portable Source: http://www.iovs.org/content/44/1/275.long
  • 15.
    Dual Purkinje Image or“glint” Uses infrared light to create reflections from different parts of eye Video tracks 2 reflections and software calculates angles Good accuracy Comfortable for users Allows more head movement Still costly Source: http://www.fourward.com/dconcept.htm Source: http://medical-dictionary.thefreedictionary.com/Purkinje- Sanson+mirror+images
  • 16.
    Limbus or PupilTracking or“Single-PointVideo-Based” Image processing of video is used to locate pupil and calculate POG A Starburst algorithm projects rays from an initial starting point detecting pixels with high image gradients Source: Howard & Sulak Example of rays projected from initial point Source: Ryan, Duchowski, & Burchfield
  • 17.
    “Dual-PointVideo-Based” Image processing trackspupil and corneal reflection of IR light is also tracked Less expensive Comfortable Head movement is limited Only accurate to 1-2 degrees (1o is about 0.5” at 27” distance) Example of rays projected from initial point Source: Ryan, Duchowski, & Burchfield Source: Howard & Sulak
  • 18.
    Some OtherTerms toKnow Accuracy—Measured in degrees and typically ranges from 0.5 to 2 degrees. Keep in mind that 1o is about half an inch of resolution at the normal 27” viewing distance for computer monitors. Precision—Smallest change in eye position that can be measured. Usually this is also measured in fractions of degrees. Sampling Rate—Measured in Hertz, this factor is important because eye movements are the fastest humans can produce. Over 1000 Hz is needed to capture many eye movements, but most systems typically run at 50 to 60 Hz.
  • 19.
    The “F Pattern”and Golden Triangle Source: http://blog.mediative.com/en/2011/08/31/eye- tracking-google-through-the-years/ Source: http://www.nngroup.com/articles/f-shaped-pattern-reading-web-content/ “Eyetracking visualizations show that users often read Web pages in an F-shaped pattern: two horizontal stripes followed by a vertical stripe.” Jakob Nielsen “We recorded how 232 users looked at thousands of Web pages.We found that users' main reading behavior was fairly consistent across many different sites and tasks.” Found consistent pattern from 2005 - 2011
  • 20.
    Rosenblatt and WebPage Genres Efferent(to carry away) Aesthetic (emotional pleasure) “In readings that fall somewhere in the efferent half of the continuum, the reader selects out predominantly more public or cognitive elements than private.The aesthetic stance, in contrast, accords selective attention to predominantly more of the penumbra of private feelings, attitudes, sensations, and ideas than to the public aspects.” Louise Rosenblatt Source: “TheTransactionalTheory:Against Dualisms.” College English,Vol. 55, No. 4. (Apr., 1993), p. 383.
  • 21.
    Rosenblatt and GazePlots Efferent(to carry away) Aesthetic (emotional pleasure) Do the GoldenTriangle and F-Pattern obtain for “aesthetic” sites as well as for “efferent” sites?
  • 22.
    Tharon Howard, ClemsonUsabilityTesting Facility Director tharon@clemson.edu Part One
  • 23.
    Honoring the Bestof theWeb Source: http://www.grumpycats.com/grumpy-cat-wins-meme- of-the-year-at-the-17th-annual-webby-awards/#.UmiDsyToVWZ Source: http://www.webbyawards.com/press/
  • 24.
    Source: http://www.morethanamap.com/ Google Morethan a Map – Efferent (Transactional)
  • 25.
    HuffPost Live -Aesthetic Source: http://live.huffingtonpost.com
  • 26.
    EyeGuide®Tracking System EyeGuide® Hardware: •Camera • LED • 2.4 GHz transmitter Source: http://www.otshow.com Source: Howard,Christiansen & Gaines
  • 27.
    Calibrating the EyeGuide® Source:Howard,Christiansen & Gaines Source:T. Howard
  • 28.
    Research Question: • Dousers' eye movement patterns differ between websites which are designed to provide users with information (efferent/ transactional) versus websites which are designed to provide users with aesthetic experiences? Preparing for the Study Participants: • 21+ in age with a bachelors degree or higher • Recruited a convenience sample • Goal of 10 participants • Approximately 30 – 45 minutes/ participant
  • 29.
    Fit participant withthe head-mounted eye tracking device. Study Procedures Calibrate the eye tracker. After a satisfactory calibration, begin tasks. Participant is shown a series of ten screen captures from website interfaces. Each screen capture is assigned a specific task. While completing tasks, the EyeGuide® software captures eye movement data. Stimulated RetrospectiveThink Aloud: after completing the tasks, the participant is shown gaze plots of his or her eye movements and asked a series of questions.
  • 30.
  • 31.
  • 32.
    Stimulated RetrospectiveThink Aloud Source:Howard,Christiansen & Gaines
  • 33.
    Task-based eye tracking- the Golden Triangle and F Pattern are still present, but current web conventions tend to dominate the gaze. What DidWe Learn? Source: Illustration adapted fromYarbus 1967, figure 109, for Land andTatler 2009
  • 34.
  • 35.
  • 36.
  • 37.
    Heather Christiansen, PhDStudent Rhetorics,Communication and Information Design Clemson University hdunn@clemson.edu Part Two
  • 38.
    Brian Gaines, Graduate ResearchAssistant, Clemson University blgaine@g.clemson.edu
  • 39.
    EyeTrackers can beexpensive.This usedTobii T120 system is $27,500. (Image courtesy of eBay, inc.) SensoMotoric Instruments (SMI) Eye Tracking Glasses 2.0 are priced on a case by case basis.This could potentially be quite costly. (Image courtesy of SensoMotoric Instruments, GmbH) Price
  • 40.
    Lab costs, aswell as the costs for participants can be costly. For example, Eye Tracking, Inc. charges between $5,000-$250,000 per study.These costs are determined by the type of study, length of study, number of participants, etc. Price
  • 41.
    The Grinbath EyeGuide is roughly $2,495, with discounts applied to multiple unit purchases. (Image courtesy of Grinbath, LLC) Affordable options are available.
  • 42.
    Eye Tribe’s USB3.0 Hardware is only $99, but comes with a Software Development Kit (SDK), which means code must be written in order for this device to be used. While the base price is attractive, the time and cost of hiring a software engineer is something to consider. But, how affordable is affordable?
  • 43.
    Equipment Sensitivity Because ofthe sensitivity of the equipment, the participants were uncomfortable. (Image Courtesy of Brian Gaines)
  • 44.
    Equipment Sensitivity Calibrating theeye tracker is in itself an art and a science. (Image courtesy of Tharon W. Howard, Ph.D.)
  • 45.
  • 46.
  • 47.
    Retrospective think aloudis a technique used in usability, and eye tracking in particular, to gather qualitative information on the user intents and reasoning during a test. It's a form of think aloud protocol performed after the user testing session activities, instead of during them. Fairly often the retrospective protocol is stimulated by using a visual reminder such as a video replay. (Guan, Lee, Cuddihy, and Ramey, 2006) Stimulated RetrospectiveThink Aloud
  • 48.
    More than 80%of subjects’ verbalizations of what they were attending to corresponded with the eye movement data. We reject the notion of subjects’ fabrication since only less than 3% of their verbalization failed to match up with objects identified by their eye movement. (Guan, Lee, Cuddihy, and Ramey, 2006) Stimulated RetrospectiveThink Aloud
  • 49.
    Immediately following theeye tracking session, the data is available. Instantaneous Results
  • 50.
  • 51.
    Cons • Certain eyetracking technologies are expensive • Lab/participants costs • Equipment sensitivity • Accommodating for differences in participants Pros • As technology advances, prices will drop • Stimulated Retrospective Think Aloud is a valid data collection method • Results are intantaneous Conclusions