LATERAL BIASES IN SCANNING
TWO- DIMENS§0NAL REPRESENTATIONS
0F THREE-DIMENSMNAL SCENES

Bissertatéon for the Degree of Ph. D“
MICHIGAN STATE UNIVERSITY '
FRANK HOLEY
1974

LIBRARY. '

Umty

 

This is to certifmat the
. thesis' chunk?

      
   
 

epfe' ' (dwards fulfillxfient
‘ e retiuirements for

negree in Psychology

. ‘ . i
‘x“’ WW
,7. .

on

»- ' . Major p‘ofessor

 

 

 

 

 

ABSTRACT

LATERAL BIASES IN SCANNING
TWO-DIMENSIONAL REPRESENTATIONS OF
THREE-DIMENSIONAL SCENES

By
Frank Holly

Previous work had indicated that there may be a
preference for looking at the right half of Symmetrical
targets with minimal depth cues. Pilot work supported
this conclusion and extended it to symmetrical targets
representing three-dimensional.scenes.

The present study attempted to relate this lateral
bias in looking behavior to some of the lateral effects
found by Bartley and co-workers using asymmetrical pictures
of three-dimensional scenes. Among other things, these
studies had found that items in the left foreground appear
nearer than items in the right foreground and that this
effect is stronger when a large background item is on the
right side rather than the left.

In attempting to relate these two sets of findings,
observers were presented with symmetrical and asymmetrical

scenes and their eye fixations recorded. For each

Frank Holly

asymmetrical target there was another one which was its
mirrorvimage. It was found that: (1) there was no
difference in fixation time between right and left halves,
(2) the first fixation, however, tends to be to the left
of later fixations, (3) location of the foreground item
does not affect fixation time on this item, and (4) there
are no sex or order effects.

The results were related to other studies which
had indicated that the first fixation is the best indicator
of attensity whereas total fixation time is an indicator
of the difficulty of cognition of a given part of the
target. The lack of sex differences in this and other
eye movement studies using adult subjects was contrasted
to studies of children in which sex differences were found.
There were, however, sizeable individual differences in
the present study and these were discussed in terms of
another study which found certain differences in looking
behavior to be related to I.Q. The Noton and Stark theory
that pictures are remembered and recognized in terms of

eye movement patterns was also discussed.

LATERAL BIASES IN SCANNING
TWO-DIMENSIONAL REPRESENTATIONS OF
THREE-DIMENSIONAL SCENES

BY

Frank Holly

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Psychology

1974

 

 

 

'. n

L‘U.

LIST

inn" A
lO.‘A\L

V': "1: w

Adagiév.

TABLE OF CONTENTS

LIST OF TABLES
LIST OF FIGURES .
INTRODUCTION
METHOD

Subjects

Apparatus .

Procedure .
RESULTS .
DISCUSSION

REFERENCES

ii

Page

. iii

iv

29
29
35
37
54
61

'ia'ale

 

10.

Table

10.

LIST OF TABLES

Latin Square Design Used in Analysis
of Variance . . . . . . . .

Analysis of Variance of Fixation Time
on Left Half of Pictures of Set A

Analysis of Variance of Fixation Time

on the Left Half of Pictures of Set B .

Analysis of Variance of Fixation Time
on the Left Half of Pictures of Set C

Analysis of Variance of Fixation Time
on Foreground Item for Set A

Analysis of Variance of Fixation Time
on Foreground Item for Set B

Trend Analysis of Fixation Time on
Foreground Item in Set B

Trend Analysis of Fixation Time on
Foreground Item in Set B (Men)

Trend Analysis of Fixation Time on
Foreground Item in Set B (WOmen)

Trend Analysis of Foreground to Back-

ground and Background to Foreground
Item.Movements in Set B .

iii

Page

39

4O

42

44

45

46

47

47

48

49

Figure

10.
ll.

12.

13.

14.
15.

Figure

10.
11.

12.

13.

14.
15.

LIST OF FIGURES

PIETER JANSSEN, Reading WOman .

 

Mirror Image of Figure 1

The Glance Curve As Envisioned by Gaffron
and Other Art Commentators .

RAPHAEL, The Change to Peter (cartoon)

 

Mirror Image of Figure 4

 

REMBRANDT, The Return of the Prodigal
Son . . . . .

Mirror Image of Figure 6
RAPHAEL, Death of Ananias (cartoon)

 

RAPHAEL, Death of Ananias (tapestry).

 

Targets Used in M.A. Thesis by Holly

One of the Targets Used by Noton and
Stark with the Scanpath of One Subject
Superimposed . . . . . . . . .

Internal Representation of a Pattern:
(a) pattern, (b) features, (c) scanpath,
(d) feature ring . . . . . . . . .

Position of Median and Quartiles for
First Twenty-two Fixations in Looking
at Each of the Three.Versions, Based
on Fourteen Subjects . . . . .
Mackworth Camera

Pictures in Set A .

iv

Page

10
11
17

19

21

25
30
32

 

Figure
l6.
17.
18.

19.

20.

Figure
l6.
l7.
l8.

19.

20.

Pictures
Pictures

Fixation
Pictures

Fixation
Pictures

Fixation
Pictures

in Set B .
in Set C .

Time on the Left Half of the
of Set A . . . . . . . .

Time on the Left Half of the
of Set B . . . . . . .

Time on the Left Half of the
of Set C . . . . .

Page
33
34

41

43

44

INTRODUCTION

A number of articles have been written concerning
the left-right question in art and how a mirror image of
a painting differs from the original (Schlosser, 1930;
Wolfflin, 1941; Oppé, 1944; Gaffron, 1950; and others).
One of the most complete treatments of the subject is

that given by Gaffrom (1950). Janssen's Reading Woman

 

(Figures 1 and 2) is one of the examples used in her
article to show some of the perceptual changes resulting
from such a reversal. In part, she noted that: (1) In

the original, while seeming to stand further away, we

seem to look directly against the side wall and against the
front of the large chest. In the mirror image, on the
other hand, we seem to look in a different direction,
namely along the side wall. (2) In the mirror image the
slippers seem more important; we perceive them at first
glance and they seem to stand nearer to us. It now seems
easy to look into the inside of the foremost slipper while
we seem to be looking at its outside in the original.

(3) In the mirror image the distance to the back wall seems

greater; the picture space appears deeper.

 

 

 

 

 

 

 

 

. a} .3
74’ ”VI.
:“= Fifi
.V
\ c .
D
O
. z, r
' L
I"‘
x .
C
C
I

"V

3' 4
x.

I , 5 .
“.23; )
V

Figure l.--PIETER JANSSEN, Reading Woman. Munich, Alte
Pinakothek.

 

%g_ 71”,!»
. ...~

 

 

 

 

 

 

 

« J.
-(><I-1 ‘ 7
A.
~' I
. .
P
“ah;
v " 5
‘
a
.4".
. f
.3: r
. V“
— s in
.1???
'7 r- . oi

4:0
9
12,». q ‘
if.

Figure 2.--Mirror Image of Figure 1.

\i

 

her thec
in the 1
then tux
when vie
standpo:

on the ‘.

 

.—.
h
0'?

Objec-
ruleS
grOu:
Pictt

are I

Gaffron explained all such left-right effects by
her theory that (1) there is a glance curve which begins
in the left foreground, penetrates toward the depth, and
then.turns over toward the right (Figure 3), and (2) people
when viewing a picture tend to locate their imaginary
standpoint at the beginning of this glance curve, i.e.,

on the left side of the picture.

 

 

 

 

Figure 3.--The Glance Curve As Envisioned by Gaffron and
Other Art Commentators. (From Gaffron, nght
and Left in Pictures," Art Quarterly, 1950).

 

The glance curve causes a greater salience for those
objects lying along its path and determines some of the
rules of composition. The left foreground and right back-
ground are good places for the important objects in the
picture while the left background and right foreground

are relatively dull. It also explains the greater depth

 

in the
origin
by the
the pi
the ti
izg th

giving

most 1
viewi:
path.

proces

the IE

03 the
Causes
"7v

We ft

greatq

decid‘
a Pet:

groun<

in the mirror image of the Reading Woman than in the

 

original; in the original the curve is first detained
by the drapery on the chair in the left foreground of
the picture and then meets the back of the woman. In
the mirror image it is virtually uninterrupted, follow-
ing the lines of the floor up to the back wall, thus
giving the impression of greater depth.

Gaffron does not propose a glance curve in the
most literal sense; eye movement recordings of persons
viewing pictures do not show the eye traversing such a
path. Rather, it is a phenomenon based upon the central
processes of vision whereby all objects located within
the range of this path are recognized spontaneously while
we must look separately for those located outside it,
i.e., in the right foreground or upper left background.

The location of the viewer's imaginary standpoint
on the left results in a different set of effects. This
causes the left side of a picture to seem like "our side."
"We feel that items located here are closer to us and have
greater importance to us. A person standing in the left
. foreground with his back turned toward us arouses a
decided feeling of identification with ourselves, whereas
a person looking out of the picture from the left fore-
ground seems directly opposed to us." It also causes

right-to-left movement portrayed in the picture to have

 

the C

the character of approach while movement from left to
right seems to be withdrawing from us. The right side,
on the other hand, seems further away and items on that
side are subjectively less important. One can look for
effects of this nature in two more of her examples
(Figures 4 through 7).

The various commentators on this right-left
question have not necessarily agreed on all points.
WOlfflin (1941), writing earlier than Gaffron, believed
that the right side was generally of greater importance.
He also believed in the physical reality of the glance
curve. Oppé (1944), in turn, disagreed with Wolfflin's
assessment of some of the works of Raphael. Associated
with most tapestries and etchings is an earlier, mirror-
image, cartoon from which the etching or tapestry was
made. A favorite argument among those interested in this
right-left question is whether the cartoon or the mirror—
image tapestry (etching) reflects the composition actually

intended by the artist. In Raphael's Death of Ananias

 

(Figures 8 and 9) Wolfflin finds in the abnormal position
of Ananias an argument in favor of his theory that the
composition was intended to read as in the tapestry, and
from left to right. He supposes that Raphael deliberately
emphasized the catastrophe by throwing the figure of

Ananias against the natural direction of the composition.

 

Figure 4.--RAPHAEL, The Chan e to Peter (cartoon). London,
Victoria and A ert Museum.

 

Figure 5.--Mirror Image of Figure 4.

 

Figure 6.--R.EMBRANDT, The Return of the Prodi a1 Son.
Leningrad, The Hermitage.

 

Figure 7.--Mirror Image of Figure 6.

10

 

 

Figure 8.--RAPHAEL, Death of Ananias (cartoon). London,
Victoria and KIEert Hfiseum.

11

iWF‘Imihum mm... .. '

 

Figure 9.--RAPHAEL, Death of Ananias (tapestry). Rome,
Vatican.

12

Oppé, like most other art historians, considers instead
the cartoon to be a model of composition which flows
logically and dramatically in waves from a center and in
which the action proceeds mainly from left to right. The
crowd on the left have their backs to the spectator and
look upwards to the Apostles, the staircase in the dark
corner and the people on it ascend, while on the right,
under an open space, quiet upright personages frame the
group of the prostrate Ananias and of the man who points
excitedly towards the Apostles in the center from whom
the catastrophe proceeds. In his view, the tapestry
shocks with an immediate effect of confusion and the
figure of Ananias is all but overlooked.

A study by Adair and Bartley (1958) represented
the first attempt to apply psychophysics to this problem

of sideward differences (sideward is used as the perceptual

 

correlate of the physical term lateral). They used five
scenes with varying degrees of lateral asymmetry as
determined by five judges. The judges were used because
of the complexity of the scenes and the consequent com-
plexity of determining asymmetry metrically.

There were two orientations, normal and mirror-
image reversed, of each scene and two sizes, 4 x 4 and
8 x 8 inches, yielding a total of four prints of each

scene. During each trial, there was one small print and

13

one large print visible to the observer. The small prints
were placed at a fixed distance, just to the right of the
track, and the large prints were placed on a carriage on

the track. The two large versions of each scene appeared

in combination with each of the two small versions of the
same scene and the task of the 0's was to adjust the metric
distance of the large print so that the scene in it appeared
to be at the same distance as the corresponding scene in

the small print.

It was found that the pictures which had the
prominent items on the left were placed at a greater
distance than those which had these objects on the right.
In addition, the distance setting of the prints was
influenced by the asymmetry factor in the manner expected;
the greater the asymmetry in a scene, the more accentuated
were the left-right differences.

The next pair of studies relate to Gaffron's state-
ment that "A person standing in the left foreground with
his back turned toward us arouses a decided feeling of
identification with ourselves whereas a person looking
out of the picture from the left foreground seems directly
opposed to us." Bartley and Thompson (1959) used prints
whose asymmetry was established on a more objective basis.
All scenes consisted of a human figure standing on the

center stripe of a roadway extending from the foreground

14

to the horizon. There were also some trees and a few
other objects in the scenes. The human figure was
placed in five different positions: two placements on
the left, two on the right, and one in the center. In
agreement with the previous study, it was found that

the human figure when on the extreme left appeared
nearer than when on the extreme right. The same dif-
ference was found between the less extreme positions but
to a lesser extent. Thompson and Bartley (1959) used
two pictures which contained a human figure standing in
the center. The only difference between the two pictures
was that in one the subject was facing the camera while
in the other he had his back to it. They found that the
man with his back to the camera seemed nearer.

Four other studies in this same tradition were
performed by Swartz and Hewitt (1970), Swartz and Swartz
(1971), Nelson and MacDonald (1971), and Holly (1971).

In the Swartz and Hewitt study (1970) subjects from
grades 1 through university level were presented with
original vs. mirror-image views of each of a series of
twenty pictures. They were asked for their preference
of each pair and the grand mean number of original views
selected was significantly greater than chance expectation.
Preference for the original varied significantly over

pictures with the following properties of lateral

15

organization emerging most distinctly as influential in
the response: (a) pattern of lighting, (b) profile
orientation, (c) handedness characteristics, (d) quadrant
distribution of important objects, and (e) ease of entering
the picture space. Choice was also a function of the
positional arrangement of the two views. With respect

to individual differences, when preference behavior was
averaged over paintings, educational level was a more\
important dimension than either sex or handedness. When
preference was considered for paintings singly, the
influence of sex and handedness was considerable.

Swartz and Swartz (1971) performed a predictive
study in which they selected twelve paintings on the basis
of the five factors of lateral organization found by
Swartz and Hewitt to be of importance. Adult subjects
were shown the original and a mirror-image reversal of
each of these pictures and asked to select the one they
preferred from each pair. The pictures had been selected
so that in six of them thepreponderance of lateral factors
predicted a preference for the original while a preference
for the reversals was expected in the other six. In two
instances there was a preference for the predicted picture
(one original and one reversal) while in the other ten
cases no significant difference was shown. In a replica-

tion with fourth graders, one picture (one of those which

 

16

showed a significant difference in the first part of the
study) showed a significant preference for the predicted
version.

In the Nelson and MacDonald (1971) study, subjects
were shown fifteen displays consisting of a picture and its
mirror-image. The pictures had been selected so that
they contained important items in both the left and right
foregrounds. The experimenter assigned a title to each ‘
display such that in one picture it referred specifically
to the item in the left foreground while in the other it
referred to the item in the right foreground. When sub-
jects were asked to assign the title to the most appropriate
version of each display they showed a significant tendency
to make the title refer to the item in the left foreground
rather than the right foreground.

In the study by Holly (1971), an attempt was made
to force the observers' imagined standpoint to the center
and thereby reduce the saliency of the left side. It was
thought that either of two means would accomplish this
result. In one version of the photographs (Figure 10)
there was a straight highWay which ran from.center fore-
ground to center background and whose center stripe was
in the exact center of the picture (Set III). The result
was a well-defined center point and supposedly a compel-

ling sense of viewing the scene from the center.

 

 

 

 

 

 

 

 

 

 

P—fl-

 

 

 

 

 

 

K/

-.—_..--_—

.—

Set 1

 

 

Set III

 

settv

 

 

 

 

 

 

 

 

 

 

 

h————q

 

 

 

 

 

 

 

 

 

Figure 10.--Targoto Used in M.A. Thooio by Holly.

18

In the other version, the photograph was cropped in such
a manner as to bring the foreground to a point in the
center, again hoping to force the observers' standpoint
to the center (Set IV). The apparent distance of the
critical item (black rectangle) in these pictures was
compared to that in the regular pictures (Sets I and II).
The results revealed no significant change, indicating
that even when the location of the objective vantage point
is well defined the effects do not disappear.
Traditionally, those interested in eye movements
have been either the old—time peripheralists of early
behaviorism or the Hebbians with their emphasis on
redintegrative processes. Recently, however, a new breed
of computer-interfaced psychologists have become intrigued
by saccadic eye movements perhaps because of their analogy
to the sudden, non-continuous shifts in computer process-
ing. Typical of this latter group are Noton and Stark,
whose work has recently gained some prominence. They have
taken note of the regularities in scanning evidenced when
a given person views and re-views a given picture and have
used this to build a serial model in which eye movements
serve as the links between the discrete snapshots etched

in the memory trace.

vi “-v- v-——.—_,.. . .

-._——- ._—-—— "‘" A

 

19

C.
P
0‘. "" ‘re
, -\
'
e 9. (
. V .
. ‘
‘ o \ V"
oi ‘1‘
' C
5 ’ ‘ “
. I o
‘ e
s ' . PS
0 \ "
I ' ‘ .-
‘ - g 3‘ ¢ 4
\ 3‘ ‘.
I
0‘ ’ ~ I‘. ‘ ‘
-. . O " ’ s e '
s o \
I
0‘ I

Figure 11.--(he of the Targets Used by lioton and Stark

with the Scanpath of One Subject Super-
inoeed.

20

In one of their studies (Noton and Stark, 1971)
the targets were pictures of a car, a man's profile, and
a scene with two trees (Figure 11). During the initial
viewing phase, subjects spent 25% of the time following
a fixed scanpath which tended to recur intermittently
with the rest of the time taken up by movements to various
irregular or unpredictable points. In the recognition

phase, subjects tended to begin with the scanpath and,

 

in fact, spent 65% of their time, on the average, fol-
lowing the scanpath during this phase. The authors offer
a model of pattern recognition in which the memory trace
of a pattern is a sort of S-R chain in which the 83 are
xthe various features of the pattern and the Rs are the
eye movements or attentional shifts linking one feature
to another. The addition of attentional shifts to eye
movements as the Rs of the model is in deference to the
fact that pattern recognition obviously can occur with
no eye movements. Graphically, they show this in the

following way:

21

 

D Sensory Memory Trace
0 Motor Memory Trace

Figure 12.--Internal Representation of a Pattern:
(a) pattern, (b) features, (c) scanpath,
(d) feature ring. (From Noton and Stark,
Vision Research, 1971.)

There are reliable differences between subjects
as well as between pictures in the scanpaths. The
authors cite the between subjects differences as evidence
that the scanpath is a bona fide habit and not simply due
to the fact that a given picture always has the same
spatial arrangement and hierarchy of features, thereby
causing the feature detectors to initiate the various
fixations always in the same order. According to them,
if this were the case there would be much more similarity

in the scanpath across subjects since presumably we all

have more or less the same set and hierarchical arrangement

22

of feature detectors. The between pictures differences
in scanning pattern they use as further evidence that the
scanpath must be intimately connected with the way in
which we tell one picture from another. Also, the fact
that the time spent on the scanpath jumps from 25% to 65%
when we go from initial viewing to recognition indicates,
they believe, that the eye movement pattern has something
to do with relating the picture being viewed to one's
memory of it.

Doubt, however, about the significance of eye
movements to pattern recognition was cast by some work
which Chris Gilbert and I did on the Mackworth Eye
Camera and which was suggested by his MLA. thesis. For
his thesis he made composite pictures of faces by split-
ting them down the middle and combining each half with
its mirror image reversal, thus yielding two fairly
normal looking pictures, One composed of two right halves
and one composed of two left halves. He found that when
he presented the original pictures to subjects and asked
them to tell him which of the composites most resembled
the person, Ss showed a significant tendency to choose
the composites of the right halves of the original face
(right halves of the original faces = left visual field
when looking at the center of the original faces). It

made no difference in performance whether the original

23

face or its mirror image reversal was used as the
inspection target; in both cases subjects chose the
composite corresponding to the left visual field of
the inspection target indicating that the effect was
not caused by some imbalance in the targets themselves.

As a follow-up, the next step seemed to be to
determine whether this greater importance of the right
side was associated with a greater time spent looking at
that side. We would not have been surprised to find
either no difference or a greater time spent looking at
the right side of the face, but the result was the one
that no one would have predicted: Ss spent more time
looking at the 12;; side (p < .001) of the face (right
visual field). Both the originals and mirror-image
reversals of the originals were used as targets but the
results were the same in either case; more time was spent
looking at the right visual field. Also, the test which
he had administered in his thesis was given again and
the results came out the same; Ss chose the right come
posites (left visual field of the originals). There was
no correlation between the amount of time which a subject
spent viewing one side of the face over the other and his
tendency to choose one composite over the other.

There was an extensive 1935 study by Buswell in

which, using methodology similar to our own, he showed a

24

large number of pictures to a large number of subjects and
obtained records of the subjects' fixations superimposed
on the paintings. Careful examination of his records
yielded some interesting findings. For one, artists
generally have little success in controlling or prede-
termining the fixation pattern of the observer in spite
of art school talk about controlling the sweep of the
viewers' gaze. The only picture which elicited more or
less the same fixation sequence from every subject was a
picture of a large wave in the ocean in which fixations
tended to start at the bottom of the wave and move up and
around to the tip.

One of his targets might be described as a section
of typical wallpaper border. It was simply a wide, thin
strip containing a row of identical dog forms (Figure 13).
As can be seen from the fixation records, there is a strong
tendency for $8 to immediately shift their gaze to the left
and proceed to the right in step-wise fashion and then
back towards the left, after which follows a series of
smaller oscillations. In other versions of the same
target, the dogs were all running to the left or right
but the fixation pattern remained the same, again demon-
strating little correlation between the sweep of the

picture's motion and the pattern of eye movements.

25

is fiéfiflfifiefifl

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Q.

Figure l3.--Position of Hedisn and Quartiles for First
Twenty-two Fixations in Looking at Each of the
Three Versions, Based on Fourteen Subjects. (From
Buswell, How People Look at Pictures, 1935.)

26

Buswell notes in passing that his two nearly
symmetrical pictures showed a greater number of fixa-
tions in the right visual field than in the left. The
only other reference in the literature to a fixation
bias for one hemifield over the other which I was able
to find was in an article by Thomas (1963) in which he
mentions that subjects look more to the right side
(visualfield) of symmetrical Rorschach cards than to the
left.

In the present study, observers will be shown
asymmetrical pictures of the type used by Bartley and
co-workers as well as some symmetrical versions with
strong depth cues. Of course, the very asymmetry of the
Ascenes which we will be using would be expected to cause
an imbalance in looking of one sort or another. This can
be controlled by showing subjects both the original
scenes and their mirror images. In this way, first one
half of the picture and then the other half will appear on
the left side and the same for the right side. The time
spent looking to the right side totalled over the two
different halves of the picture can then be compared to
the time spent looking to the left totalled over the two
halves. It is predicted that there is a bias for looking
at the right side even with asymmetrical pictures which

will emerge under these conditions. Symmetrical versions

27

of these scenes will also be shown, and it is expected
that there is an even stronger bias with symmetrical than
with asymmetrical scenes. Because of the redundancy of
the two halves of a symmetrical scene there is no need to
look at the left half.

Bartley has commented on the immediacy of the
lateral effects in his scenes and the possible significance
therefore of the first fixation. It is predicted that
because of the greater prominance of the left foreground
more of these first fixations will fall on the left half
than on the right.

In a further analysis, it will be determined
whether there is any hint of a glance curve going from
the lower left corner to the upper right corner as postu-
lated by Wolfflin. 'Although it has already been shown
that there is no distinct glance curve of this sort, it
is possible that a weak effect of this sort exists. In
order to determine this, the number of eye movements
from the critical item in the foreground to the large
background item in the original will be compared to that
for the mirror-image reversal. If any sort of weak glance
curve exists, one would expect more movements of this sort
when the critical item is in the left foreground and the
large background object is on the right side than vice-

versa .

28

Also, the scanpaths a la Noton and Stark of both
the original scenes and their mirror images will be
examined for differences and similarities. Specifically,
an attempt will be made to determine: (1) if the scan-
path is as strong in the mirror-image reversal as in the
original and (2) if the direction of movement in the
mirror image tends to be in a direction opposite to that
of the original.

In another analysis, the perception of the
critical foreground item as affected by its physical
relationship to the large background item will be examined
more closely. Here, the subjects will be shown four dif—
ferent placements, right, right center, left center, and
left of the critical foreground item with the rest of the
scene remaining the same. It will be determined whether
the number of fixations on the critical item changes in
a systematic way as the critical item is moved from.right
to left. Additionally, it will be determined whether the
number of eye movements from the critical item to the
large background item changes in'a systematic way as the
critical item is moved from right to left. With respect
to all of these questions sex and order effects will also

be determined.

METHOD

Subjects
Sixty-four subjects (thirty-two male and

thirty-two female) were chosen from.the undergraduate

population at Michigan State University.

Apparatus

 

Eye movements were recorded by means of a Mack-
worth camera (Figure l4). This camera utilizes the
corneal reflex technique in which a light source is
positioned so that its image reflects from the cornea
of the left eye into a lens system which aims it towards
the front (silvered) surface of a half-silvered type beam
splitter. From there, it reflects into the lens system of
a Beauliou movie camera. This image reflects off of the
beam splitter at approximately a 90° angle but the exact
angle depends, in a quite linear fashion, upon the angle
of reflection from.the cornea and hence eye position.
Meanwhile, an image of the target which the subject views
is directed by a system of mirrors towards the rear of the
beam splitter. It passes straight through the beam split-
ter and into the lens of the camera. The result is a

moving picture record of the target material with a white

29

3O

 

Figure 14.--Hackworth Camera.

31

dot (the reflected light source) indicating each fixation
superimposed on it. The system is quite linear up to
about 20°. A Hunter timer was hooked in series with the
movie camera and its 6-volt power supply to provide exact
timing for the recordings.

The target material consisted of three sets of
four 4%" X 7%" pictures of outdoor scenes. In Set A
(Figure 15) there was: (1) a picture of the type which
has been found to maximally produce the lateral effects,
i.e., a scene in which there is a small foreground item
on the left and a large background item on the right,
(2) the mirror—image of (l), (3) a scene in which both the
small foreground item and large background item are on the
right, and (4) the mirror image of (3). The four scenes
of Set B (Figure 16) were identical except for the differ-
ent placements, right, right center, left center, and left,
of the foreground item. The first two pictures of Set C
(Figure 17) were the same as pictures (1) and (2) of Set A.
Pictures (3) and (4) were symmetrical scenes composed from
prints identical to those of (l) and (2). Picture (3) was
the result of joining the right half of scene (1) to left
half of scene (2) and picture (4) resulted from joining
the left half of scene (1) to the right half of scene (2).
Care was taken to insure that the symmetry in Sets A and C

was exact .

)1.

.>H

.< mom c“ monouu«m-u.

.HHH

m

d ou:m«m

 

33

.>H

.m uom c“ monouoamuu.oa ouowam

.HHH

 

 

34

.>H

.0 new

ad «eunuuwman

.HHH

2 3%:

 

35

Procedure

 

Before each subject entered the room, all the
targets were placed in the target holder so that whenever
a target was withdrawn from the holder the next one was
automatically revealed. When a subject was brought in,
he was seated in front of the apparatus and the function
of the biteboard explained to him. He was exhorted to
attempt to keep his head still.

Dental impression compound was then placed on the
biteboard and S bit down to form an impression. When it
had hardened, he bit it again and fixated the center dot
. of a test pattern containing five fixation dots, one in
the center and one in each corner. E then adjusted the
optics until the marker light appeared superimposed upon
the center dot. To insure the linearity of the system 0
then, on command from E, repeatedly fixated all the dots
‘including the center one and E made further adjustments,
if necessary, until the marker light was superimposed
upon which ever dot 0 was fixating. At the end of align-
ment, 0 was instructed: "Your job in this experiment is
simply to look at the pictures which I will soon show you."

After the instructions, all lights including those
illuminating the target were turned off and the subject
was told to get ready for the first picture. The test

pattern was then removed from the target holder and the

36

camera was turned on, followed a split-second later by
the lights. The camera stayed on for exactly 6 seconds
for each recording and ran at a speed of 16 frames/second.
The first 8 men and 8 women were shown the pictures of
Set A. The first male saw them in the order 1, 2, 3, 4;
the second in the order 2, 3, 4, l; the third, 3, 4, l, 2;
the fourth, 4, l, 2, 3; the fifth, l, 2, 3, 4; etc. The
same procedure was followed with the women so that two men
and two women saw the pictures in each of the four orders.
The next 16 subjects (each group of 16 subjects contained
8 men and 8 women) were shown the pictures of Set B in the
same manner; The third group was shown the pictures of
Set C by the same procedure. In the final group of 16
subjects each person saw only one picture: 4 subjects
(each group of 4 was composed of 2 men and 2 women) saw
picture B2, 4 saw picture B3, 4 saw picture C3, and 4 saw
picture C4. The final group was used to collect more
observations on these four pictures since the first three
groups yielded fewer observations on these pictures than
on the others. That is, pictures Al and A2 also appeared
as Cl and C2 and pictures B1 and B4 also appeared as Al
and A3.

RESULTS

The data was scored by means of a Bell and Howell
film analyzer. This projected the film onto a screen and
allowed E to advance the film one frame at a time and
record on a copy of the target the position of each
fixation. Fixation time, the parameter used in the
statistical analyses, is thus in units of one frame.
Since the film speed used in recording was 16 frames/
second, each frame represents .0625 seconds. Of course,
the shutter was not open during the entire .0625 second
cycle so that where there was a movement between one
frame and the next it may represent more or less than
.0625 seconds.

First, a t—test was performed to learn whether
there is an overall tendency to look more to the right
side than to the left. Set A and Set C were used for
this analysis since all the pictures in these two sets
either were symmetrical (pictures C3 and C4) or had
mirror images (pictures A1, A2, A3, A4, C1, and C2) so
that the sideward differences in content were balanced
out. Data from Set B could not be included since this

set did not contain its own mirror images. In this test,

37

38

data from Sets A and C were totalled across subjects and
the expected time on the left half (50% of the total time
spent on both halves divided by 32) was subtracted from
the mean actual time spent on the left half and the dif—
ference divided by the standard error of the mean as
determined from the left-half sample. No significant
difference was found.

In order to gain information about sex and order
variables as well as to gain more detailed information
about the picture variable, separate analyses were per-
formed on Sets A and C according to the following Latin

square design:

39

Table l.--Latin Square Design Used in
Analyses of Variance.

 

 

 

 

 

 

 

b; be b1 b;
by b1 b2 b3

r

 

a a ' a a
G b1 b2 b3 bu
G b2 b3 b. b.
c G b3 bu b1 b2
G by b1 ' b2 b3
G b1 b2 b3 be
G b. b. 1... b.
c - .. A I
G
G

 

 

 

 

 

where time on the left side is the

dependent measure and where:

a = Order

b = Pictures

c = Sex

G = Groups (pairs of the same sex

seeing the same order of
presentation--not to be
confused with the groups
of subjects defined
earlier)
This analysis of Set A (Table 2) shows no signifi-
cant Sex or Order effects but does show a significant
effect for Pictures (p < .01). This significant F reflects,

in part, the large differences in time spent on the left

40

Table 2.--Analysis of Variance of Fixation Time (No. of
Frames) on Left Half of Pictures of Set A.

 

 

 

 

 

Source SS df MS F
Between subjects 9,437 15 629 .87
Sex 564 1 564 .78
Groups within Sex 3,118 6 520 .72
Subjects within Groups 5,755 48 719
Within Subjects 39,203 48 817 1.51
Order 1,313 3 438 .81
Pictures 14,178 3 4,726 8.74**
Order X Sex 1,015 3 338 .62
Pictures X Sex 1,630 3 543 1.00
Residual ‘8,093 12 674 1.25
Error 12,973 24 541

 

 

 

 

 

*Denotes significance at .05 level.

**Denotes significance at .01 level.

side between pictures A1 and A2 (a total of 830 frames vs.
a total of 522 frames, respectively) and between pictures
A3 and A4 (a total of 709 frames vs. a total of 496 frames,
respectively). It is not surprising that there should be
a large difference between pictures A3 and A4 since in one
(A3) of them nearly all the items are on the right side,
whereas on the other (A4) nearly all of them are on the
left. However, the direction of this difference might
seem rather surprising, i.e., the majority of fixation

time in both cases is on the relatively empty side. Also,

41

it might be considered surprising that pictures A1 and

A2, which are.more balanced, should reveal an even larger

effect of this sort. In fact, a Newman-Keuls test shows

the difference between A1 and A2 to be significant while

that between A3 and A4 is not. Figure 18 shows these

differences graphically.

’a?

Q)

5

H

£3

1.41000

93

p 800

m

,3 600

8 400

0)

E 200
0

Figure 18

 

 

I l I. l

 

A1 A2 A3 A4
Pictures

.--Total Fixation Time on the Left Half of the
Pictures of Set A. Dashed lines indicate the
expectations at .50 probability of fixating
on the left half. Expectations were obtained
by dividing the total fixation time on each
picture by 2.

42

The analysis of variance of Set B is shown in

Table 3.

Again, the only significant effect was that for

Pictures (p < .05) although the Order X Sex interaction

approached significance (F = 2.59).

Table 3.--Ana1ysis of Variance of Fixation Time (No. of
Frames) on the Left Half of Pictures of Set B.

 

 

 

 

 

Source SS df MS F
Between Subjects 30,875 15 2,058 .691
Sex 133 1 133 .045
Groups within Sex 6,921 6 1,153 .387
Subjects within Groups 23,821 8 2,978
Within Subjects, 48,298 48 1,006 1.189
Order 556 3 185 .219
Pictures 11,277 3 3,759 4.443*
Order X Sex 6,579 3 2,193 2.592
Pictures X Sex 3,840 3 1,280 1.513
Residual 5,734 12 478 .565
Error 20,312 24 846

 

i

 

 

 

 

*Denotes significance at .05 level.

The difference across pictures is shown in Figure 19.

Time on the left side is maximal with B2 and falls off

on either side.

43

 

 

13

3

:‘3 1000...

EL

44 800 ~ N

H ----.. _____ ..----

m dd-

== 600 -

U

“a: ..

,4 400

8 200 -

Q)

OE O 1 1 l 1

9" 131 32 33 134
Pictures

Figure l9.--Total Fixation Time on the Left Half of the
Pictures of Set B. Dashed lines indicate the
expectations at .50 probability of fixating
on the left half. Expectations were obtained
by dividing the total fixation time on each
picture by 2.

In Set C (Figure 20 and Table 4) none of the F3
reached significance. Pictures Cl and C2 show the same
order of difference (a total of 754 frames on the left
for C1 vs. a total of 558 frames on the left for C2) as
they showed in Set A (A1 and A2) but this effect does not
reach significance because of the similarity of pictures
C3 and C4 with respect to fixation time on the left side

(610 frames vs. 543 frames, respectively).

44

Table 4.——Analysis of Variance of Fixation Time (No. of
Frames) on the Left Half of Pictures of Set C.

 

 

 

 

 

Source SS df MS F
Between subjects 17,957 15 1,197 1.211
C 963 l 963 .975
Groups within C 9,093 6 1,515 1.533
Subjects within Groups 7,902 8 988
Within Subjects 31,911 48 665 .901
A 602 3 201 .272
B 5,242 3 1,747 2.367
AC 146 3 49 .066
BC 736 3 245 .332
Residual 7,482 12 624 .846
Error 17,702 24 738

 

 

 

 

 

*Denotes significance at .05 level.

 

 

 

 

 

7:?

a 1000 "

H

E: 800..

'8. ' _-.\_-- -----

:3 600 1" r? A \n:--
ct} 400' ..

3

c: 200*-

o

G) O 1 s 1 l
.3 C1 C2 c3 04
H Pictures

Figure 20.--Total Fixation Time on the Left Half of the
Pictures of Set C. Dashed lines indicate the
expectations at .50 probability of fixating
on the left half. Expectations were obtained
by dividing the total fixation time on each
picture by 2.

 

45

In order to learn how the sex, order, and pictures
variables influenced fixation time on the critical fore-
ground item, another set of analyses was performed on
Sets A and B using the same Latin square design. The only
difference was that in this case the dependent variable
was fixation time on the foreground item rather than
fixation time on the left side. Neither Sets A nor B
(Tables 5 and 6, respectively) showed any significant
differences. Set C was not analyzed by this method since

pictures C3 and C4 contained two foreground items.

Table 5.--Ana1ysis of Variance of Fixation Time (No. of
Frames) on Foreground Item for Set A.

 

 

 

 

 

Source SS df MS F‘
Between Subjects 3,532 15 1M235 1.11
Sex 248 l 248 1.18
Groups within Sex 1,594 6 266 1.26
Subjects within Groups 1,690 8 211
Within Subjects
Order 296 3 97 .87
Pictures 855 3 285 2.57
Order X Sex 276 3 92 .83
Pictures X Sex 692 3 231 2.08
Residual 1,144 12 95 .86
Error 2,666 24 111

 

 

 

 

 

*Denotes significance at .05 level.

46

Table 6.--Ana1ysis of Variance of Fixation Time (No. of
Frames) on Foreground Item for Set B.

 

 

 

 

 

Source SS df MS F
Between Subjects 10,000 15 667 1.38
Sex 447 l 447 .93
Groups within Sex 5,692 6 949 1.97
Subjects within Groups 3,861 8 482
Within Subjects 22,799 48 475 .97
Order 3,073 3 1,024 2.08
Pictures 397 3 132 .27
Order X Sex 1,193 3 398 .81
Pictures X Sex 1,944 3 648 1.32
Residual 4,378 12 365 .74
Error 11,815 24 492

 

 

 

 

 

*Denotes significance at .05 level.

To further examine the question
tion time on the foreground item varied
way with its position, a trend analysis
the pictures of Set B. The pictures of
remembered, varied only with respect to

the critical foreground item; all other

of whether fixa-
in a systematic
was performed on
Set B, it will be
the position of

elements were held

constant. Fixation time on the foreground item was the

dependent variable and linear quadratic

were tested. As Table 7 shows, no such

and cubic trends

trends were found.

47

Table 7.—-Trend Analysis of Fixation Time (No. of Frames)
on Foreground Item in Set B.

 

Source SS df MS F
Between 424 3 ‘
Linear 367 1 367 .68
Quadratic 9 l 9 .02
Cubic 47 l 47 .09
Error 34,443 64 538

Totals 34,867 67

 

 

 

 

 

*Denotes significance at .05 level.

Nor were any effects of this sort found when the data from
men and women were analyzed separately as shown in

Tables 8 and 9.

Table 8.--Trend Analysis of Fixation Time (No. of Frames)
on Foreground Item in Set B (Men).

 

 

 

Source SS df .1 MS . F
Between 1,893 3
Linear 410 1 410 .58
Quadratic 3 l 3 .00
Cubic 102 l 102 .14
Error 22,806 32 713
Totals 24,698 35

 

 

 

 

 

*Denotes significance at .05 level.

48

Table 9.--Trend Analysis of Fixation Time (No. of Frames)
on Foreground Item in Set B (Women).

 

Source SS df MS F
Between 304 3 154 .45
Linear 154 1 154 .45
Quadratic 3 1 3 .01
Cubic 137 l 137 .40
Error 9,637 28 344

Totals 9,961 31

 

 

 

 

 

*Denotes significance at .05 level.

Again using the pictures in Set B, a trend
analysis was performed on the number of foreground item
to background item (tree) and background item to fore—
ground item movements. This was done to determine whether
the numberrof such movements varied systematically as the
foreground items moved from left to right. Table 10 shows°
that no such trends were found.

The next set of tests was directed at the glance
curve question which meant looking for certain types of
regularities across subjects with respect to the sequence
of their fixations across subjects with respect to the
sequence of their fixations. Specifically, this meant
looking for any hint of the lower-left-to-upper right

sequence postulated by Wolfflin, Gaffron, and others.

49

Table 10.--Trend Analysis of Foreground to Background and
Background to Foreground Item Movements in

 

 

 

 

Set B.
Source
Between .093 3 .031 .323
Linear .024 l .024 250
Quadratic .040 1 .040 417
Cubic .021 l .021 219
Error v 6.157 64 .096
Totals 6.25
{

 

 

 

*Denotes significance at .05 level.

In the first, a t-test was performed on the number
of foreground item to background item and background item
to foreground item movements in pictures Al and Cl vs.
those in A2 and CZ. Picture Al-Cl would, according to
Wolfflin and Gaffron, have good lateral composition since
the major items in this picture fall along the glance
curve whereas picture A2-CZ is an example of bad lateral
composition since here the major items do not fall along
this curve. It was expected that there would be more such
movements in picture Al-Cl where the item placement and
postulated glance curve are in harmony than in picture
A2-CZ where they conflict. However, no such difference

was found.

50

Although, as shown above, there is no hint of a
lower-left-to—upper right movement tendency, further
tests revealed a fairly strong tendency for the eyes to
move initially from left to right when viewing a picture.
Looking at the first fixation alone, we find that more of
them fall on the left half of the pictures than on the
right half. In this absolute sense of left half vs.
right half the effect does not quite reach significance
(t==1.62). However, when the position of the first fixa-
tion relative to the second was noted, there was found a
significant tendency for the first fixation to fall to the
left of the second.

To analyze these sequential effects, binomial
tests were used and the relative position of the first
fixation to that of the second, i.e., to the right or
left, was noted. Since only one score could be assigned
to each subject in these tests, four identical tests (one
for each of the four pictures viewed by most subjects)
were run. One of these four tests (I) used data from.the
first picture viewed by each subject, another (II) used
the data from.the second picture seen by each subject,
etc. Data from.those subjects who viewed only one picture
were included in (I). In all cases, there was a signifi-
cant tendency for the first fixation to lie to the left
of the second (p <.01, p:<.05, p:<.01, and p:<.05,

respectively).

51

Also, using only the first picture seen by each
subject, a binomial test was performed comparing the
absolute side of the first fixation to the side at which
he looked most. Instances in which both of these fell
on the same side were discarded, leaving 36 usable obser-
vations. There was a significant tendency for the initial
fixation to be on the left half relative to the half which
was looked at most (right half).

As in the Noton and Stark (1971) study, there was
a strong tendency for subjects to repeat previous fixa-
tions rather than fixate on a new point every time. Each
6-second record was examined to determine the longest
sequence of new fixations, i.e., the maximum number of
movements made before returning to a previous fixation
point. The average of these longest sequences over all
subjects and all pictures was 2.08 new fixations before
returning to a previous fixation point. The mode number
of such new fixations was two, with one being the second
most common and then three, four, five, and six in that
order. The maximum number of these new fixations found
in any record was six (there were two of these). In most
cases, the previous fixation point returned to was the
one immediately preceding the new fixation or fixations.

Generally, about four types of looking behavior
could be seen in the records: (1) a relatively long

period of time during which the subject simply fixated

52

on one point, (2) a series of short, redundant movements
covering perhaps two or three points of which the initial
fixation point was one, (3) a series of new, exploratory
fixations but again returning to the initial fixation
point, and (4) a relatively long sequence of fixations
covering anywhere from three to seven points repeatedly
traced. A sequence of this latter type (4) was often
momentarily interrupted for a backward movement to a
previous fixation point. A variation on (3) was for the
subject to move back to the initial fixation point after
every single exploratory foray to a new fixation point.

The different looking behaviors generally tended
to occur in the order listed, but all behaviors were by
no means included in every record. In fact, it was more
common for one or more to be omitted on any given record.
Looking behaviors showed a great deal of consistency
within subjects across pictures. It might also be noted
that the one pair of brothers used in the study showed
no particular similarities in looking behavior.

The final t-test was designed to learn whether
observers tend to form a stronger scanpath (whatever the
shape of that path) when viewing pictures with "good"
lateral composition than when viewing pictures with "bad"
lateral composition. The number of times a scanpath was

traced in picture Al-Cl was compared to the number of

53

such tracings in picture A2-CZ. No significant difference
was found.

In short, the results showed that: (1) there was
no significant difference between time spent on the left
and right halves, (2) the only evidence for any sort of
a glance curve was the tendency to look first to the left
and then, in a series of one or more fixations, move to
the right, (3) there was no correlation between fixation
time on the critical foreground item and its position in
the picture and, (4) none of the tests of sex or order

effects turned out to be significant.

DISCUSSION

The fact that the first fixation tends to fall to
the left of later fixations indicates that first fixation

may measure something different than does length of fix-

 

EEiQE: A study by Hackman and Guilford (1936) is of
interest in this connection. In a study of eye movements
and attnetion, they chose targets involving four of the
well-accepted factors of visual attention: position,
isolation, size, and novelty. The position slides con-
tained nine letters (ordinary typewritten capital letters)
arranged symmetrically in three rows and three columns with
the center letter in the exact middle of the slide. The
isolation slides contained one letter on one side of the
slide and a group of letters (ranging from seven to twelve
in number) on the other side. The size slides had ordinary
capital letters on one side and large typewritten capital
letters in a larger group on the other side. The groups

of large letters and small letters were arranged each with
five letters, two in the top row and bottom.row and one in
the middle, the whole slide being arranged like a domino
with five spots on each side. The novelty slides had five

ordinary typewritten capital letters (arranged in the same

54

55

grouping as above) on one side and a group of five novel
items, arranged in the same manner, on the other side.
The novel items were colored or black geometric forms
and typewritten digits. Both short (100 msec.) and long
(10 sec.) exposure times were used and subjects were
instructed to "observe the exposure field and report
which part of the field is most prominent or most compel-
ling, most intriguing or the part you feel most inclined
to look at.” As a double check, subject reports of
attensity value were taken after each presentation and
the portion of the slide intended by E to be the most
attention-getting was nearly always the one reported.

They found that locus of the first fixation cor-
related more highly with reports of attensity than did
length of fixation. Further, the average correlation
between the first fixation and length of fixation was only
.337. They, in fact, found some tendency for greater fixa-
tion time on the side of the field not containing the
factor of attention, possibly due to the greater difficulty
of cognition involved in the small letters, the group of

letters opposite the isolated letter, or the familiar

 

items, which were letters, as opposed to geometric figures
and digits . They concluded that the length of fixation was the
best indicator of what they called cognition. This was

supported in a second part of the study in which subjects

56

were instructed to ”observe the exposure field and report
what you have seen. Describe or name as many objects as
you can observe." In this case, length of fixation cor—
related best with the objects reported. In a study to

be described more fully later, Lewis, Kagan, and Kalafat
(1966) also found the first fixation to be a better
measure of differential attention than was the longest
fixation.

If these results can be directly related to the
present study it would seem that the left side of a pic-
ture is more attention-getting than the right side. This
pattern of the first fixation being to the left of subse-
quent fixations was also quite pronounced in the face
study described earlier. Beyond this, it remains, to
be seen just how general this pattern is, i.e., how many
different types of targets evoke this response.

No significant sex differences were found with the
adult 83 used in the present study. Other eye movement
studies have generally found sex differences when children
were used as the subjects but not when adults were used.
Hoats, Miller, and Spitz (1963), for example, found that
normal eight-year-old boys had more perceptual curiosity
than did girls of that age when such 33 could request either

a simple or complex pattern during a viewing spell. Their

57

evidence was that the boys chose the complex designs
twice as often (50% of the occasions) as did the girls
(25%). No such difference, however, was found in subjects
aged 17 years.

In a habituation study by Mackworth and Otto
(1970) subjects were presented with repeated viewings of
a display of 16 circles, one of which changed from red to
white on each trial. Boys spent 23% of the time inspecting
this circle whereas the girls spent only 8% of their time
on this circle. At the beginning of this test there was
an even greater difference since the boys averaged 42%
and the girls 12%. V

A study by Lewis, Kagan, and Kalafat (1966) found
that six-month-old female infants spent a significantly
longer time looking at faces (one male, one female, and
one schematic) than at non-faces (checkerboard, bulls-eye,
bottle). The six-month-old boys, however, showed no dif-
ferences. In a finding reminiscent of that of Hackman
and Guilford (1936), they found the first fixation to be
a more sensitive index of differential attention than was
total fixation time, since the girls showed an even
greater difference between the faces and non-faces when
first fixation was used as the measure. Again, the boys

showed no difference.

58

Whether any of the rather large inter-subject
differences in scanning behavior found in the present
study are related to intelligence or any other personal
variables is not known. A study by Guba, Wolf, deGroat,
Knemeyer, van Atta, and Light (1964) which dealt with much
smaller eye movements found such a relationship. They.
recorded the eye movements of subjects viewing a TV screen
and counted the instances of what they called NOMs (no
observable eye movement from.one frame to the next) and
MINs (a movement of between 15 minutes and one degree of
visual angle). They found that the intelligence groups
differed sharply for NOMs and MINs; high intelligence
subjects displayed more NOMs than MINs while low intelli-
gence subjects displayed more MINs than NOMs.

The scanpaths noted by Noton and Starkxwere very
much in evidence in the present study. Still, a great
deal of time was spent in not tracing any predictable
scanpath. This shift back and forth from the predictable
to the unpredictable is perhaps best handled by the model
proposed by Furst (1971). He used pictures of real-life
objects and focussed in on the learning and habituation
aspects of the scanning pattern, a process which he called
automatizing of visual attention. He noted that as tflme
went on the locii of eye fixations became more and more

predictable and the rate of fixations dropped. This

59

decrease in rate he called habituation. He also found a
tendency for fixations of short duration to be followed
by eye movements of short distance and for fixations of
long duration to be followed by movements of a longer
distance. He proposed an interesting model to account
for his results:

The significance of the finding that long fixations
tend to be followed by long eye movements may be
seen in terms of a speculative hypothesis about
momentary shifts of attention. It is suggested that
the difference in average transition distance re-
flects a difference in central attentional state
which is signified by long or short fixations.
Assume that there are two central states, correspond-
ing to "attention" or "inattention" to the visual
channel. Commands for a new fixation can occur from
either of these states but the parameters of control
of eye movements will be different for each state.
Probably the commands from an "inattentive" state will
be coarser and result in relatively longer movements.
It has previously been seen that average fixa-
tion time increases during habituation. If one can
further assume that long fixations reflect periods of
central inattention to the visual channel, the dif-
ference in average transition length between brief
and long fixations is explained.

For purposes of the present study we can think of two
central states, one giving rise to predictable movements
and the other giving rise to unpredictable movements.
If it is the case that long fixation durations and long
movement distances are associated with predictable move-
ments and short fixation durations and short movement
distances are associated with unpredictable movements,
then the results of the present study fit directly into

this model.

60

The question of the relationship of eye movements
to pattern recognition and memory remains open. For as
Noton and Stark themselves point out, the regularities in
scanning behavior do not prove that these movements are
an integral part of the process by which we recognize
pictures as their serial model states. Rather, they
assumed such a model and set out to examine the implica-
tions of this assumption. Their stimulus conditions, in
fact, were designed to promote serial processing; the
stimulus pictures were dimly illuminated relative to the
general illumination so that subjects could barely see any
parts of the pictures not viewed foveally. It is note-
worthy, however, that the same regularities in scanning
behavior were noted with the more normal illumination con-
ditions of the present study, thus indicating that the
scanpaths they found were not just artifacts of their

design.

REFERENCES

REFERENCES

Adair, H. J., and Bartley, S. H. Nearness as a function
of lateral orientation in pictures. Perceptual
and Motor Skills, 1958, 8, 135-141.

 

 

Bartley, S. H., and Thompson, R. A. A further study of
horizontal asymmetry in the perception of pictures.
Perceptual and Motor Skills, 1959, 9, 135-138.

 

Buswell, G. T. How People Look at Pictures. Chicago:
University of Chicago—Press, 1935.

 

Furst, C. J. Automatizing of visual attention. Percep-
tion anvasychophysics, 1971, 19, 65-69.

 

Gaffron, M. Right and left in pictures. Art Quarterly,
1950, 13, 312-331.

Guba, E., WOlf, W., deGroat, S., Knemeyer, M., vanAtta, R.,
and Light, L. Eye movements and TV viewing in
children. Audio-Visual Communications Review,
1964, 12, 386-401.

Hackman, R., and Guilford, J. A study of the visual fixa-
tion method of measuring attention value. Journal
of Applied Psychology, 1936, 20, 44-59.

 

Hoats, D., Miller, M., and Spitz, H. Experiments on per-
ceptual curiosity in mental retardates and normals.
American Journal of Mental Deficiency, 1964, 68,
386-395.

Holly, F. Effects of certain target and observer variables
upon the phenomenal distance of left side and right
side foreground items. Unpublished Masters Thesis,
Michigan State University, 1971.

Lewis, M., Kagan, J., and Kalafat, J. Patterns of fixation
in the young infant. Child Development, 1966, 31
331-339.

61

62

Mackworth, N., and Otto, D. Habituation of the visual
orienting response in young children. Perception
and Psychophysics, 1970, 1, 173-178.

 

 

Nelson, T., and McDonald, G. Lateral organization, per-
ceived depth, and title preference in pictures.
Perceptual and Motor Skills, 1971, 33, 983-986.

 

Noton, D., and Stark, L. Scanpaths in saccadic eye move-
ments while viewing and recognizing patterns.
Vision Research, 1971, 33, 929-942.

Oppé, A. Right and left in Raphael's cartoons. Journal
of the Warburg and Courtauld Institute, 1944, VII,
82.

Schlosser, J. W. Intorno alla leture dei quadri. Criticg,
1932, 33, 72.

Swartz, P., and Hewitt, D. Lateral organization in pictures
and aesthetic preference. Perceptual and Motor
Skills, 1970, 33, 991-1007.

Swartz, P., and Swartz, S. Lateral organization in pictures
and aesthetic preference: II. a predictive study.
Perceptual and Motor Skills, 1971, 33, 319-324.

Thomas, E. L. Eye movements and fixations during initial
viewing of Rorschach cards. Journal of Pro ective
Techniques and Personality Assessment, 31, 45-
349.

Thompson, R., and Bartley, S. H. Apparent distance of
material in pictures associated with higher order
meanings. Journal of Psychology, 1959, 43, 353-

 

358}
Wolfflin, H. Uber das rechts und links in bilde.
Gedanken zur kunstgeshichte. Basel, Switzerland:

 

Sdhwabe, 194I.

   

"'711'!lelele111}jiujljitjfljfljj‘rmfi'ES