Eye Tech

Document Sample
Eye Tech Powered By Docstoc
					                                     Eye Tech.
                                  J.A.Renshaw
                           Leeds Metropolitan University
Abstract
This paper describes an experiment in which the eyes of participants were tracked
as they searched a visual display of a matrix of information which would enable them
to complete a given task. The tracking was conducted by means of a non invasive
remote eye tracker using a beam of near infra red light. The purpose of the
experiment was to test the hypothesis that certain design features incorporated into
the displays would assist in the search enabling the task to be completed more
quickly whilst others would not. The data was presented in rows and columns whose
alignment was varied both vertically and horizontally, whilst some designs also had
bands of colour (frames) whose orientation was similarly varied. Interactions
between alignment and the frames were anticipated and it was found that the best
performances were achieved with a combination of vertical alignment and horizontal
frames, but arrangements with horizontal alignment and vertical frames were not
only slower, they were also found to appeal less to participants and gave rise to eye
movements with significantly different properties to those found in other displays.
The experiment underlines the importance of vertical alignment in the design of
visual displays.
Introduction
        The attributes of alignment, proximity and similarity have been found to have
an influence on eye movement in a display of graphs (Renshaw et al, 2004). In a
further eye tracking experiment the impact of design features used in the
presentation of data in a matrix is studied to further test the hypothesis that emergent
properties, in support of a task, will result in predictable eye movements.
        The presentation of data in rows and columns has long been the subject of
 much research (see Tullis, 1997 for a review). In addition to improved task
 completion times (Tullis, 1981) it has also been found that positive subjective rating
 of appeal are associated with structured displays (Tullis, 1983) and that designs that
 optimise user performance do not always optimise user appeal (Tullis, 1988).
        Yamamoto et al (1992) used eye tracking to study matrix design and
 concluded that a columnar format was more efficient than a text based one and that
 a twelve row by four columns layout was the most usable. In further eye tracking
 study, Hornof et al (2003) asked participants to look at a series of designs, in which
 blocks of data were grouped into vertically aligned columns of data of various
 lengths, to find a target in the presence of similar formatted distracters. They
 established that displays with headings were searched more efficiently than those
 without.
        All the above studies incorporated designs with vertical alignment, but this
 study also considers the impact of alignment and misalignment both horizontally and
 vertically, because the possibility of encountering similar design features on the
 World Wide Web. Further more, whereas Hornof et al (2003) considered the
 grouping effect of headings, in this study we consider the effect of coloured frames
 (see Appendix A: Figure 1), as designers have long used frames as a technique to
 assist in the location of data in complex matrixes. Figure 1
        This experiment asks how effective these different devices are in assisting
 users to locate an item of interest and asks the question: do they provide
opportunities for user-exploitation, enabling faster completion of a simple search and
locate task?
       We created designs in which regions of uniform connectedness (Palmer
1994), with either vertical or horizontal orientations, have been superimposed upon
a matrix of data elements grouped into rows and columns. Connectedness was
achieved by superimposing coloured bands or frames arranged either horizontally or
vertically over the matrix and similarity was achieved through the format of individual
elements of data and by means of their horizontal and vertical alignment into rows or
columns.
       We investigate the effect these features have on eye movement using a non
invasive eye tracking technique, whilst recording the time taken to complete a
search and locate task. The users were also asked to give their personal rating of
the appeal of each design.
Display Designs
Twelve displays were designed; see Appendix A, figure 1.The text/data in the
displays describes fictitious individuals in the following sequence:

   Initials (Initials XX), Age (Agegrp NN), Gender (Gender M), Region (Region X)1.

        For the purpose of analysis each display was divided into three zones: a
question display zone (in the upper left of each display, not shown), an Initials zone (
first column) and a Data zone in which the details attributed to the initials were to be
found.
Eye Movement/ Metrics
        The eye is always on the move but occasionally it is relatively still and these
instances are known as fixations and may be defined as a period of time within
which the locus of an eye’s attention remains within a specified area; periods
between fixations are known as saccades and may be defined as a pre-planned
ballistic movements between fixations. A scan path is the path described by a series
of saccades in sequence and its length may be considered to be the product of the
number of saccades multiplied by an average scan path length the latter being
referred to in this study as the average inter-fixation distance. It is this measure, for
eye movements in the Data Zone, upon which we report.
Participants
        Thirteen volunteer participants (5 male, 8 female) took part in the experiment.
Their ages ranged from 20 to 55 and they were drawn as an opportunity sample from
the mathematics, computing, multimedia, health faculties of two universities and
members of the public. Nine wore no corrective lenses; the remainder wore glasses
or contact lenses. Their dominant eye was identified and all save one were
successfully calibrated.
Apparatus and Materials
       The experiment used an ASL504 pan/tilt eye tracker system capable of
detecting a bright back lit pupil image caused by the retinal reflection of a near
infrared beam of light emanating from the eye tracker, which was positioned
immediately in front of the participant just below the display screen. The resolution of
the eye tracker is better than one degree and it has a sampling rate of 50 Hz. The


1 The letters in parenthesis reflect the terms used in the displays.
data was recorded using GazeTracker software, capable of recording fixation data
and mouse clicks. The displays were shown on a 17 inch PC colour monitor with a
screen area of 1024 x 768 pixels. Software was used to select the images at
random, and to control the timed presentation of various aspects of the displays.
Time stamped fixation and mouse click data was exported to spreadsheets for
subsequent analysis.
Experimental Procedure
        In a repeated measures experiment 12 participants were shown the displays
in a randomised sequence and after a 5 second free view period asked to locate the
answer to a question put to them during the display. The elements making up the
question were randomised so that the probable location of the answer was evenly
distributed over the scene. On clicking on the answer the display disappeared and
they were then asked to rate the display as to its appeal on a scale of 1(low) to 5
(high), again indicating their choice of rating by clicking on a display. Each participant
saw all twelve displays once. Throughout the experiment participants were seated
approximately 60 cm. from the screen in a high backed chair – no other restraining
mechanism was used.
        Prior to the running of the experiment each participant was given a chance to
practice the responses required on a similar display to the experimental displays.
Results
Non Eye Tracking Metrics
Time Taken to Task Completion (Response Times)
        The mean time to complete the task for each display is shown in Figure 2
It shows variations in mean response times for each frame type as a function of
alignment. A 4(alignment) x 3(frame) ANOVA showed the main effect of alignment
[F(3,33) =3.9 p<0.05]. A post hoc t-test did not establish the difference between
horizontal and vertical frames in the presence of horizontal alignment to be
significant but a further post hoc repeated measures t test did show that the
difference between the fastest and the lowest response times was significant (t=-3.7
p<0.05).


                                                                    Mean Response Times (s)
                                                            6.0



                                                            5.5
                  Mean Response Times (s)




                                                            5.0



                                                            4.5                                    F

                                                                                                       1
                                                            4.0
                                                                                                       2

                                                            3.5                                        3
                                                             None        H               V     T


                                                                             Alignment




                                                           Figure 2 Mean Response Times (s)
                                            Legend: Frame: 1 = No frame, 2 = Horizontal frame, 3 = Vertical frame.
                                            Alignment: None = No alignment, H = Horizontal alignment, V = Vertical
                                            alignment, T = alignment both vertically and horizontally.
Ratings (Scores of Appeal)
The results of the users’ evaluations were recorded a plot of the mean scores is
shown in Figure 3. It shows an increasing appeal rating as the degree of alignment
increases from no alignment to total alignment. A 4(alignment) x 3(frame) ANOVA
showed that there was a main effect of alignment [F(3,33) = 15.2 p<0.05], the
highest rating being given to those designs with total alignment.

                                                          Mean Ratings Of Appeal
                                               4.5



                                               4.0



                                               3.5



                                               3.0                                         Frame
                        Mean Rating




                                                                                              1
                                               2.5
                                                                                              2

                                               2.0                                            3
                                                None           H               V       T


                                                                   Alignment




                                                     Figure 3 Mean Ratings of Appeal


                       Legend: Frame: 1 = No frame, 2 = Horizontal frame, 3 = Vertical frame.
                       Alignment: None = No alignment, H = Horizontal alignment, V = Vertical
                       alignment, T = alignment both vertically and horizontally.



Correlation Tests
       In order to determine the correlation between eye movement metrics,
performance measurement and user assessments, it was decided to rank each of
the displays in terms of their results. Table 1 shows the means and standard
deviations of the various measures recorded during the Task phase. On the basis of
these means each display was then awarded a rank indicating performance on a
scale of 12(best) to 1(poorest). For example in the case of fixation duration times
and scan path lengths, the lower the value of

   Display    N                       Appeal             Response          Task Phase         Task Fixation    Task Scan path
  Reference                                              Times (s)          Fixations           Duration           length
                     Mean                  Std.        Mean     Std.      Mean     Std.      Mean     Std.     Mean      Std.
                                          devn.               devn.                devn               devn              devn
          1   12             2.2            1.40        4.80     1.93       8.2     3.19      3.32      1.61    1541       541
          2   12             2.9            1.44        4.86     2.11       8.3     3.84      3.35      1.99    1641       834
          3   12             2.3            1.42        4.87     1.72       9.3     3.44      3.23      1.22    1754       684
          4   12             2.9            1.38        4.99     2.29       8.7     5.07      3.00      1.46    1715       739
          5   12             3.2            1.24        4.82     2.49       9.4     6.82      3.44      2.52    1717      1001
          6   12             3.3            0.98        3.98     1.14       8.0     3.07      3.07      0.83    1480       442
          7   12             4.2            1.19        4.25     1.15       6.9     2.11      2.95      1.26    1520       653
          8   12             3.6            1.70        4.87     2.52       8.8     5.10      3.34      1.36    1978      1066
          9   12             2.4            1.31        4.84     2.04       9.5     5.44      3.45      1.66    1755      1048
         10   12             3.0            1.60        5.89     2.30      10.8     3.83      3.96      1.76    1972       630
         11   12             3.1            0.90        4.11     1.31       8.1     1.88      2.96      1.27    1491       700
         12   12             4.2            0.94        4.75     1.79       8.4     3.70      3.17      1.54    1747       763


                   Table 1 Summary of Descriptive Statistics for the second part of the Experiment
these variables the more effective the display. Table 2 shows the ranking of each of
the displays as a result of this process. It was decided to compute an average
ranking based on the individual eye tracking metric scores to provide a broader basis
for the mean eye tracking score ranking. A subsequent examination of the
correlation between the individual measures and those reported here showed no
significant impact upon the results.
        Correlation tests were conducted in which the correlation between the
participants’ ratings and response time, between the participants’ rankings and the
eye tracking metric rankings and between response times and eye tracking rankings
were determined. The results are shown in Table 3. Response time and the mean
eye tracking rankings show a reasonably strong and positive correlation. There are
no significant correlations between any of the other variables compared.
        Table 4 shows the various displays sorted according to each of the ranking
methods; participants’ preference, response time performance and eye tracking
metrics. The table also shows, against each display reference, the alignment
scheme that display utilises and the frame arrangement if any.

                                                             Ranking
                    Non eye tracking                                   Eye tracking
                     measurements                                     Measurements
 Display          Ratings    Response        Number of       Fixation      Scan path          Mean eye tracking
 Reference          of         times          fixations      Duration         length               score
                  Appeal
      1                  1            8                9                  6              9                     8.0
      2                  4            5                8                  4              8                     6.7
      3                  2            4                4                  7              4                     5.0
      4                  5            2                6                 10              2                     7.7
      5                  7            7                3                  3              6                     4.0
      6                  9           12               11                  9             12                    10.7
      7                 11           10               12                 12             10                    11.3
      8                 10            3                5                  5              1                     3.7
      9                  3            6                2                  2              3                     2.3
      10                 6            1                1                  1              2                     1.3
      11                 8           11               10                 11             11                    10.7
      12                12            9                7                  8              5                     6.7


                               Table 2 Ranking of User Benefit (12 high, 1 low)


                                                                                        Correlation Results
 Variables
                                                                                         Spearman n=12
                                                                                  rho             p
 Participant ratings and response time                                            0.392           >0.05
 Participant Ratings and mean eye tracking score                                  0.256           >0.05
 Response time and mean eye tracking score                                        0.726           <0.05


                                          Table 3: Correlation Results



Eye Movements in the Task Phase – Data Zone
        The average inter-fixation distance measure revealed a significant difference
between designs. Designs with vertical frames were found to have smaller inter-
fixation distances than those designs with horizontal frames regardless of the
alignment see Figure 4. A 4(alignment) x 3(frame) ANOVA reveals a main effect of
frame [F(2,20) = 6.3 p<0.05].


                                                                                 Mean Inter Fixation Distances
                                                                                            Data Zone
                                                                       400
              Mean Inter Fixation Distances




                                                                       300


                                                                                                                       F

                                                                       200                                                 1

                                                                                                                           2

                                                                       100                                                 3
                                                                         None           H                    V     T


                                                                                             Alignment




                                              Figure 4: Mean Inter-fixation Distances in the Data Zone During the Task Phase.

                                                            Legend: Frame: 1 = No frame, 2 = Horizontal frame, 3 = Vertical frame.
                                                            Alignment: None = No alignment, H = Horizontal alignment, V = Vertical
                                                            alignment, T = alignment both vertically and horizontally.




Measure                                            Appeal                              Response time (s)                       Eye tracking score
Ranking    Display                                 Alignment    Frame            Display    Alignment    Frame                  Display     Alignment   Frame
   by     reference                                                             reference                                      reference
measure
   1st.      12                                        T         V                  6                    V        H               7            T        None
  2nd.        7                                        T        None               11                    V        V               6            V         H
  3rd.        8                                        T         H                  7                    T       None             11           V         V
   4th.       6                                        V         H                 12                    T        V               1            N        None
   5th.      11                                        V         V                  1                    N       None             4            H         H
   6th.       5                                        V        None                5                    V       None             12           T         V
   7th.      10                                        H         V                  9                    N        V               2            N         H
   8th.       2                                        N         H                  2                    N        H               3            H        None
   9th.       4                                        H         H                  3                    H       None             5            V        None
  10th.       9                                        N         V                  8                    T        H               8            T         H
  11th.       3                                        H        None                4                    H        H               9            N         V
  12th.       1                                        N        None               10                    H        V               10           H         V


                                                                   Table 4: Summary of ranking results.
                                                                 Key 1st. letter = alignment, 2nd. Letter = frame


Discussion of the Results
Completion Times and Inter-fixation Distances
The results show that, in a horizontal search and locate task, vertical frames induce
shorter inter-fixation distances, or put another way, horizontal frames permit longer
saccades which in turn are generally considered to be indicators of more efficient
scanning. Either way this result is supportive of the proposed hypothesis; it suggests
that the emergent feature (the horizontal frame) assists in the completion of the task,
whereas the vertical frames may not.
An examination of the mean average fixation duration showed a main effect of
alignment as did a count of the number of fixations but neither presented an easily
discernable pattern as to likely cause (neither is reported in this paper). However,
response times for designs with vertical frames and horizontal alignment were
significantly longer than for those with vertical frames and vertical alignment
suggesting some form of inhibiting effect possibly caused by an interaction between
vertical frames and horizontal alignment (or an acceleration effect of vertical frames
and vertical alignment). The effect is not seen with vertical frames and total
alignment because it is suggested that the task can still be effected by the vertical
frames and vertical alignment route, an option not available in the vertical frame,
horizontal alignment design.

Participants’ Rating of Appeal
        Higher scores were awarded by participants for displays that were more
aligned than others, that is to say those displays, in which both horizontal and
vertical alignment were present simultaneously, scored more highly than other
displays, confirming the findings of other researchers.
Summary
        The relative rankings of the displays (Table 4) indicates that alignment
(vertical or orthogonal) is an important factor in securing better response times and
in the appeal ratings of users. The importance of frames is less clear; save for the
fact that the best task completion time relates to displays with vertical alignment and
horizontal frames and the worst to horizontal alignment and vertical frames. This
would indicate that designers should use such vertically orientated grouping devices
with care.
Future Work
        Whilst there is evidence of a positive correlation between performance and
eye tracking measures there is no correlation between appeal and performance and
as appeal may be considered an element of satisfaction this may adversely impact
upon usability assessments incorporating users views. Further research is planned
to investigate this further to determine whether eye tracking can contribute toward an
understanding of this apparent discrepancy.
Acknowledgements
I would like to thank the University of Huddersfield for the use of their eye laboratory
and their technical staff and my supervisors: Professor J.Finlay, (Leeds Metropolitan
University), Dr. D.Tyfa (University of Huddersfield) and Dr. R.D. Ward (University of
Huddersfield).
References
Hornof, A. J. a. H., T (2003). "Cognitive Strategies and Eye Movements for Searching Hierarchical
         Computer Displays." ACM CHI 2003: Conference on Human Factors in Computing Systems:
         249-256.
Palmer, S. E., & Rock, I. (1994). "Rethinking perceptual organisation: The role of uniform
         connectedness." Psychonomic Bulletin & Review 1((1)): 29-55.
Renshaw, J. A., Finlay,J.E.,Ward,R.D.,and Tyfa,D (2004). "Understanding visual influence in graph
         design through temporal and spatial eye movement characteristics." Interacting with
         Computers Vol 16(3): p557-578.
Tullis, T. S. (1981). "An Evaluation of Alphanumeric, Graphic and Colour in Information Displays."
         Human Factors 23: 541-550.
Tullis, T. S. (1983). Predicting the Usability of Alphanumeric Displays PhD. Thesis, Rice University:
         172 p.
Tullis, T. S. (1988). Screen Design. Handbook of Human-Computer Interaction. M.Holander(Ed.),
         Elsevier Science Publishers BV (North Holland): 377-412.
Tullis, T. S. (1997). Screen design,In Handbook of Human-Computer Interaction., Elsevier,
         Amsterdam.
Yamamoto, S., Kuto,Y. (1992). "A Method of Evaluating VDT Screen Layout by Eye Movement
         Analysis." Ergonomics 35(5/6): 591-606.
                                                                                                              Appendix A Figure 1: Representations of Displays Used.

                             No Alignment                                                         Horizontal Alignment                                                      Vertical Alignment                                    Total Alignment

                                                                               Initials PX              Agegrp 88              Gender M     Region N                   Agegrp 88          Gender M      Region N    Initials BN   Agegrp 88            Gender M    Region N
  Initials ER        Agegrp 88      Gender M          Region N                                                                                          Initials JE


Initials LD           Agegrp 15 Gender F           Region E                    Initials HH              Agegrp 15   Gender F                Region E                                                    Region E    Initials NM   Agegrp 15            Gender F    Region E
                                                                                                                                                                       Agegrp 15
                                                                                                                                                        Initials OL                       Gender F
    Initials PL      Agegrp 21          Gender M                Region S         Initials GH            Agegrp 21   Gender M                Region S                   Agegrp 21          Gender M                  Initials PA                        Gender M    Region S
                                                                                                                                                        Initials ST                                                               Agegrp 21
                                                                                                                                                                                                        Region S
  Initials AZ          Agegrp 21                            Region S
                                       Gender M                                                                                                                                                                     Initials SO   Agegrp 21            Gender M    Region S
                                                                               Initials KD      Agegrp 21                Gender M           Region S    Initials GY Agegrp 21                           Region S
                                                                                                                                                                                          Gender M

      Initials ZS                         Gender F                Region N                                                                              Initials LC    Agegrp 35          Gender F      Region N    Initials MQ   Agegrp 35            Gender F    Region N
                        Agegrp 35                                                 Initials YS           Agegrp 35              Gender F   Region N

                                                                                                                                                        Initials LL    Agegrp 25          Gender M                  Initials QW   Agegrp 25            Gender M    Region E
 Initials SX                            Gender M       Region E                Initials DR      Agegrp 25                      Gender M     Region E                                                    Region E
                      Agegrp 25
                                     Gender F                                                                                                           Initials PR    Agegrp 35                        Region W    Initials WE   Agegrp 35            Gender F    Region W
    Initials XD         Agegrp 35                  Region W                      Initials KA    Agegrp 35               Gender F            Region W                                      Gender F


                                          1                                                                         3                                                                          5                                               7


                                                Gender M                       Initials NT              Agegrp 88            Gender M      Region N      Initials PA          Agegrp 88                             Initials EC    Agegrp 88           Gender M     Region N
   Initials OP       Agegrp 88                                                                                                                                                               Gender M    Region N
                                                                   Region N
   Initials FJ                                                     Region E                                                                              Initials AS                                                Initials CR    Agegrp 15           Gender F     Region E
                          Agegrp 15              Gender F                         Initials SX       Agegrp 15                Gender F      Region E                           Agegrp 15      Gender F    Region E

   Initials AB                                                                   Initials MN    Agegrp 21           Gender M               Region S                           Agegrp 21      Gender M               Initials HJ    Agegrp 21           Gender M     Region S
                     Agegrp 21                                     Region S                                                                              Initials RT
                                          Gender M                                                                                                                                                       Region S

       Initials BC                              Gender M                                                                                                                      Agegrp 21                  Region S   Initials HK    Agegrp 21           Gender M     Region S
                                                                                                                                                                                             Gender M
                             Agegrp 21                             Region S      Initials KG           Agegrp 21             Gender M      Region S      Initials JQ

                                                                                                                                                         Initials SD                         Gender F
                                 Agegrp 35 Gender F                Region N                                                                                                                                         Initials RV    Agegrp 35           Gender F     Region N
   Initials CD                                                                 Initials LB      Agegrp 35                    Gender F      Region N                           Agegrp 35                  Region N

                                                     Gender M                                                                                                                 Agegrp 25      Gender M
                             Agegrp 25                                           Initials KA    Agegrp 25                    Gender M      Region E                                                      Region E   Initials VT    Agegrp 25           Gender M    Region E
                                                                   Region E                                                                              Initials DF
   Initials EF
                                                                                                                                                         Initials FG          Agegrp 35                  Region W   Initials TB    Agegrp 35           Gender F    Region W
   Initials BV                            Gender F                             Initials YY              Agegrp 35       Gender F           Region W                                          Gender F
                         Agegrp 35                                Region W


                                          2                                                                         4                                                                          6                                                   8




                        Agegrp 88         Gender M                                Initials NT           Agegrp 88              Gender M   Region N                                        Gender M      Region N    Initials EC    Agegrp 88            Gender M    Region N
                                                                   Region N                                                                              Initials PA Agegrp 88
 Initials OP
                      Agegrp 15                                                Initials SX        Agegrp 15                    Gender F     Region E                   Agegrp 15          Gender F                  Initials CR    Agegrp 15            Gender F    Region E
 Initials FJ                              Gender F                                                                                                       Initials AS
                                                                Region E                                                                                                                                Region E

                                                                                 Initials MN      Agegrp 21             Gender M             Region S                  Agegrp 21          Gender M                  Initials HJ    Agegrp 21            Gender M     Region S
 Initials AB               Agegrp 21                             Region S                                                                                Initials RT                                    Region S
                                                Gender M
                                                                                                                                                                       Agegrp 21          Gender M
                                                                    Region S   Initials KG              Agegrp 21        Gender M         Region S       Initials JQ
                                                                                                                                                                                                                    Initials HK    Agegrp 21            Gender M     Region S
                     Agegrp 21            Gender M                                                                                                                                                      Region S
 Initials BC
                                                                                                                                                         Initials SD                                    Region N
                                                                                                                                                                       Agegrp 35          Gender F                  Initials RV    Agegrp 35            Gender F    Region N
    Initials CD                                                                   Initials LB       Agegrp 35                  Gender F   Region N
                           Agegrp 35            Gender F           Region N
                                                                                                                                                         Initials DF Agegrp 25            Gender M                  Initials VT    Agegrp 25            Gender M    Region E
                     Agegrp 25                                     Region E      Initials KA       Agegrp 25             Gender M           Region E                                                    Region E
 Initials EF                                    Gender M
                                                                                                                                                         Initials FG Agegrp 35
                        Agegrp 35               Gender F           Region W                                                                                                               Gender F      Region W    Initials TB    Agegrp 35            Gender F    Region W
 Initials BV                                                                   Initials YY         Agegrp 35                   Gender F   Region W


                                          9                                                                             10                                                                     11                                                  12

				
DOCUMENT INFO
Shared By:
Categories:
Tags: Tech
Stats:
views:101
posted:2/24/2010
language:English
pages:9
Description: Eye Tech