Title: Changes in detectability of direction and motion associated with saccadic eye movements /
Full Citation
Permanent Link: http://ufdc.ufl.edu/UF00099572/00001
 Material Information
Title: Changes in detectability of direction and motion associated with saccadic eye movements /
Physical Description: viii, 124 leaves : ill. ; 28 cm.
Language: English
Creator: Krantz, John H., 1960-
Publisher: s.n.
Publication Date: 1988
Copyright Date: 1988
Subject: Saccadic eye movements   ( lcsh )
Eye -- Movements   ( lcsh )
Psychology thesis Ph. D
Dissertations, Academic -- Psychology -- UF
Genre: bibliography   ( marcgt )
non-fiction   ( marcgt )
Thesis: Thesis (Ph. D.)--University of Florida, 1988.
Bibliography: Bibliography: leaves 115-122.
General Note: Typescript.
General Note: Vita.
Statement of Responsibility: by John H. Krantz.
 Record Information
Bibliographic ID: UF00099572
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: alephbibnum - 001076290
oclc - 19045348
notis - AFG1049


This item has the following downloads:

PDF ( 4 MBs ) ( PDF )

Full Text






Copyright 1988


John H. Krantz


I am deeply indebted to my advisor, Dr. Keith White, both for what he did

in helping me through my graduate career and for what he did not do. At

times when I wanted a crutch, he forced me to stand on my own. For those

times I am chiefly grateful. I am also indebted to the rest of the members of

my doctoral committee: Dr. Keith Berg, Dr. Ira Fischler, Dr. David Green, and

Dr. John Munson. It has truly been an honor to be able to work this group of

people. In addition, without Dr. Robert Moore's assistance in the development

of the EOG trigger circuit, modelled closely after one of his own, this research

would have been more greatly delayed.

There are just too many people in the basement of the Psychology

building that need to be thanked for their support of me. Special mention

must go to Barrie Woods and Dr. Dennis Shuman. Dennis paved the way with

his development of the PSS system and his steadfast friendship. It is also one

of his circuits in the PSS system that I used to develop the amplifier for the EOG

used in these experiments. Barrie was ever constant to let me bounce off ideas

and frustrations. Most of all, he quietly put up with being a subject in all

three experiments. I regret that it looks likely that he will be unable to gain

his revenge.

In my case, without my family, both immediate and extended, this

dissertation likely would not be. My parents have listened to me and supported

me. My inlaws were another source of quiet support. Lastly, I must thank the

majors joys in my life. My wife and children did not, as some complain, hinder

my progress, but enhanced it. My daughter Jenny has provided a final

impetus to complete this document. My son Michael has been such a source of

joy that he often gave me the strength to continue. My wife Margaret put up

with my ratings about the wonders of vision and even acted as an editor for

this manuscript. I am ever thankful that she did.



ACKNOWLEDGEMENTS ................... .......... ............ iii

ABSTRACT .................................................... vii


I INTRODUCTION ............. .. ................ .......... 1

Threshold Elevation during Saccadic Eye Movements. . . . 4
Mechanisms of Saccadic suppression .................... 5
Retinal Im age Smear ............................. 6
Mctacontrast Backward Masking ................... 7
Retinal Shear .................................... 10
Inflow or Feedback Extrarctinal Mechanism ........ 12
Outflow or Feedforward (Corollary Discharge)
Extraretinal M echanism ......................... 13
Matin's Hybrid Extraretinal Mechanism ............ 17
Summary ............................................. 18


Previous Research .................................... 20
The Present Study ..................................... 25
Method ............................................... 28
Subjects ........................................... 28
Apparatus ......................................... 29
Procedure ........................... ............. 31
Data analysis ....................................... 34
Results ................... ............................ 34
D discussion ............................................ 43


Method ............................................... 48
Subjects .................... ......... ............ 48
Apparatus ............................ ............. 48
P procedure ........................... .............. 48
Results ............................................... 50
Discussion ............................................ 54


The "Two Visual Systems" Theory ........................ 60
Saceadic Effects on the Ambient Subsystem:
Previous Studies .................................... 64
M ethod ............................................... 71
Subjects .......................................... 71
Apparatus ......................................... 71
P procedure ......................................... 73
Data Analysis ...................................... 75
Results ............... ............................... 76
FFT Analysis ...................................... 77
Response Average Analysis .......................... 81
Discussion ................... ......................... 84

V. GENERAL DISCUSSION .................................. 88

S um m ary ............................................. 88
Perceptual Stability: Description ........................ 90
Perceptual Stability: Mechanisms ....................... 93
The Central Role of the Corollary Discharge .............. 95
Perceptual Stability: Plasticity .......................... 96
Unresolved Issues and Conclusions ...................... 97



EXPERIMENTS I AND 2 ............................... 102


REFERENCES ................................................. 115

BIOGRAPHICAL SKETCH ...................................... .. 123

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy



John H. Krantz

April, 1988

Chairman: Dr. Keith D. White
Major Department: Psychology

The world is perceived to be stable, even though we move our eyes.

Perceived stability is disrupted, however, if the eye is moved passively by

gently pressing on one with a finger. The present series of studies explores

the nature of perceptual stability during saccadcs, the kind of eye movement

made when looking from one object to another. Saccades are high velocity eye

movements; they are highly stereotypic and usually last less than 100 msec.

They are interesting because they might be expected to disrupt perceptual

stability, yet they do not.

The first study deals with changes in perceived direction associated with

saccades. There are two theories concerning the role of a centrally

originating corollary discharge in readjusting perceived direction. One

theory states that perceived direction is readjusted continuously and slowly,

while the other states that the readjustment is discrete and abrupt. Results

from the first experiment contradict both theories, showing that perceived

direction is adjusted continuously and quickly. It was also found that

perceived direction is readjusted slightly before the saccade begins.

The second study asks why the readjustment observed in the first

experiment is not normally perceived as a motion of the world. Motion

sensitivity was tested using a 2 cycle/0 grating moved such that the patterns

seen before and after the eye movement were identical. Velocity was 7700/sec;

the extent of motion ranged from 0.50 to 80 (1 to 16 cycles). Thresholds for

detecting motion during saccades were found to be about three times greater

than thresholds during fixation. Moreover, the extent of motion at threshold

was larger than the 20 eye movements used in this study.

The third study addresses the issue of stability of the world using

postural stability as the dependent measure. The results are complex. It

appears that some motion information during saccades is still obtained by the

postural control system, although this information may not be enough for a

characteristic response to be made.

Together these studies support the supposition that a corollary

discharge may be an important mechanism for perceptual stability during eye



To introduce the topic, try an experiment. First close one eye. Next,

gently press the other eye through the lid, keeping the eye open and observe

what happens. The world should appear to move in the direction opposite to

the direction the eye rotates. Do not be terribly surprised if the world moves

in the direction you press your eye, for some people the eye rotates against the

direction the eye is pressed (Hill, 1984; Miller, Moore, & Wooten, 1984).

Compare this situation to what happens when you look from one side of the

page to the other. In this situation the world is perceived as stable, and yet,

the displacement of the visual stimulation in the eye is, if anything, more

drastic than what occurred when you pressed your eye. This simple

experiment demonstrates how remarkable it is that we perceive the world as

stable, especially in the face of eye movements. It also indicates that our

perceptual systems must play an active role in this perceived stability.

One question that may arise is why the eye needs to move at all? That

question can best be answered by a brief description of the anatomy and

functioning of the eye. When light from an object enters the eye, the optical

elements of the eye, the cornea and the lens, focus an image of that object onto

the retina. The retina is a thin piece of tissue that lines the back of the eye,

and it is in the retina that light energy is transduced or converted into

nervous system activity. Eye movements are necessitated, in part, because only

a small area of the retina, the fovea, that is capable of detailed vision. When

detailed inspection of an object is desired, the eyes must be directed so that the

image of that object falls on the foveas of both eyes. When a person is looking

at an object, it means that the object is imaged on both of his foveas.

One of the most common eye movements is the saccade. Saccades are

executed when an observer looks from one object of regard to another or from

one part to another part of the same object. They are conjugate eye

movements; that is, the eyes move together. Saccades are also very stereotypic

in that those of a given size and direction are very similar from one time to the

next. They, also, can achieve extremely high velocities (Alpern, 1971;

Carpenter, 1977). The duration of the vast majority of saccadic eye movements

is shorter than the temporal integration period of the visual system, or the

time period across which the visual system sums light energy. From this fact,

it would be expected that saccades would be easily noticed, yet they are not. In

fact, Dodge (1900) spent three years trying to train himself to reliably detect

his own saccadic eye movements. After this extensive training, his ability to

detect his own saccades was still no better than that of an untrained subject.

Two basic questions have been raised about why saccades are not easily

noticed or detected: 1) Why is the image smear not noticed during a saccade?

2) What mechanism(s) is (are) responsible for the perception of a stable world

both before and after a saccade occurs (Helmholtz, 1909/1962). The perceived

stability of the world may seem an intuitively obvious event until two

additional observations are considered. First, recalling the experiment used to

introduce this paper, gently pressing the eye through the lid at the outer

canthus, the world appears to move in the direction opposite to the rotation of

the eye (Helmholtz, 1909/1962; Hershberger, 1984; Miller et al., 1984; Stark &

Bridgeman, 1983). During both the eye press and the saccade, the pattern of

light stimulating the retina is very similar. Thus, it is apparent that the

perception of a stable world is not inherent in the stimulation, per se, but

relates in some way to the act of executing a saccade. The second observation,

from studies of eyes paralyzed via either injury or appropriate medicine,

reinforces the first. If a person with a paralyzed eye attempts to make a

saccade, the world will seem to move or be displaced, this time in the direction

of the attempted eye movement (Stevens et al., 1976). These two observations

together reinforce the idea that the perceived stability of the world before and

after saccadic eye movements results from some active mechanism.

Another way of describing these observations, is that during eye

movements, there is spatiotopic stability in the face of retinotopic change.

Spatiotopic refers to the coordinate system of the outside world. Eye

movements do not interfere with our perceptions of this coordinate system

while eye presses and attempting eye movements with paralyzed eyes do.

Retinotopic refers to the coordinate system of the retina. The fovea is taken at

the center of the retinotopic coordinate system. The relationship of

rctinotopic coordinates to spatiotopic coordinates is altered during an eye

movement. Yet, this change does not disrupt the perceived stability of

spatiotopic coordinates. This spatiotopic stability is the subject of the

experiments discussed below (Chapters II, III, IV).

The rest of Chapter I will review and consider research pertaining to

the question of why saccades are not easily noticed. Considerably more

research has been conducted in an attempt to answer this question than in

trying to understand the stability of perception around the time of saccades.

Nevertheless, the efforts in this area have uncovered some mechanisms that

are useful in understanding the question of perceptual stability.

Threshold Elevation During Saccadic Eve Movements

The first quantitative studies of reduced ability to detect visually

presented stimuli during saccadic eye movements were done by Latour (1962)

and Volkmann (1962). Both researchers measured visual sensitivity by

comparing thresholds during the saccading eye to thresholds for the fixating

eye. In both cases, a decrease in sensitivity or elevated thresholds were

observed when the saccading eye was compared to the fixating eye. Later

research indicates that the reduction in visual sensitivity begins up to 100

msec prior to the onset of the saccade and lasts 160 msec after the saccade ends

(Volkmann, Schick, & Riggs, 1968). Volkmann (1962), measuring the

threshold for detecting a brief flash against a diffuse background, found a

peak threshold elevation of 0.5 log units for the saccading eye relative to the

fixating eye. This time course and magnitude of the reduction in sensitivity

associated with saccadic eye movements, termed "saccadic suppression," is

typical (see Matin, 1974; Volkmann, 1976, 1986 for reviews).

Similar results have been obtained when measuring absolute thresholds

(Volkmann, 1962), incremental and decremental thresholds against a diffuse

background (Riggs, Volkmann, Moore, & Ellicott, 1982), and contrast

sensitivity for gratings (Volkmann, Riggs, Moore, & White, 1978). Moreover,

similar findings have been obtained when studying blinks (Volkmann, Riggs,

& Moore, 1980) and vergence eye movements (Manning, 1986; Manning, &

Riggs 1984). The only difference of note is that thresholds to increments from

a steady background during blinks do not seem to be "suppressed" to as great a

degree as decrements from the same background level (Riggs, White,

Manning, & Kelly, 1984; White, Krantz, Manning, & Moore, 1984). Factors that

increase the amount of suppression that occurs during a saccade are

increasing luminance, increasing the number of contours in the background,

and increasing the size of the saccade (Brooks & Fuchs, 1975; Volkmann, Riggs,

Ellicott, & Moore, 1981).

Mechanisms of Saccadic Suppression

Two classes of mechanisms have been proposed to account for saccadic

suppression. These mechanisms can be broken down into those in which

suppression arises as a result of retinal stimulation associated with the eye

movement and those in which suppression arises extraretinally. The proposed

mechanisms that arise as a result of retinal events are: 1) smearing of the

retinal image as a result of the velocity of the eye movement and the temporal

integration properties of the retina (Dodge, 1900; Matin, 1974) 2) metacontrast

backward masking, where the clear image at the end of the eye movement

blocks perception of the smeared image that resulted from the eye movement

(Matin, 1974; Matin, Clymer, & Matin, 1972), and 3) shearing forces that arise

in the retina as a result of the high velocity of the saccadic eye movement

(Richards, 1968, 1969).

The extraretinal mechanisms can also be broken down into three types:

1) a feedback mechanism where saccadic suppression arises as a result of

proprioceptive feedback from the extraocular muscles. This mechanism is also

referred to as an inflow mechanism. 2) There is also the proposal that the

command to move the eyes sends a corollary signal or discharge to the visual

system which may be involved in saccadic suppression (Helmholtz, 1909/1969).

This mechanism is also referred to as an outflow mechanism. 3) Finally L.

Matin (1976a) has proposed a mechanism that results from a combination of

the two sources of information, referred to as a hybrid mechanism. Each

potential mechanism will be discussed starting with the retinal mechanisms.

Retinal Image Smear

The high velocity of saceades has two effects upon the stimulation of the

retina that fall roughly under the category of image smear. First, a stimulus is

less effective during a saccade since each receptor receives less stimulation as

the image of the stimulus sweeps past (Volkmann, 1986). Given that most

saccades are completed within the temporal integration period of the eye

(Bahill, Clark, & Stark, 1975), the reduction of the stimulus energy at each

retinal location should reduce the effectiveness of the stimulus, that is, the

change in luminance across a contour should be reduced towards threshold.

The second impact on vision that the high velocity of the eye could have

is best described via an example. If a camera is swept across a visual scene

with the shutter held open, the resulting exposure will be a sort of temporal

average of the light distribution across the visual scene. Such an exposure

will not reveal any clear details on the photograph. Again, since saccades are

usually completed within the time period covered by Bloch's Law, which

describes the temporal integration properties of the eye (Bahill, Clark, & Stark,

1975; Cornsweet, 1970), a similar type of temporal average is performed by the

visual system which can reduce the sensitivity of the visual system. That

image smear plays a role in saccadic suppression is supported by observations

that the degree of suppression increases as the complexity of the scene and

luminance is increased, both factors that would increase the amount of image

smear present during a saccade (Brooks & Fuchs, 1975). Moreover, elevation of

visual thresholds has been observed when the visual scene is moved and the

eye is kept still (Brooks & Fuchs, 1975).

The influence of image smear, per se, on saccadic suppression was

investigated directly by Volkmann et al. (1978). In their experiment, they

controlled image smear during both fixations and saccades. As expected,

increasing image smear on the retina increases the contrast threshold for the

detection of a grating, but the effect was much greater for the fixating eye

than for the saccading eye. The lowest threshold found for the detection of the

grating during a saccade, when image smear was near minimum, was still

greater than the greatest threshold found for the fixating eye, with the

maximum image smear used. Volkmann et al. (1978) concluded that image

smear, per se, is not a sufficient factor to completely account for the threshold

elevations observed to accompany saccades.

Metacontrast Backward Masking

Visual masking can be described as the altering of the detectability of

one briefly presented stimulus (the target) by the presentation of another

briefly presented stimulus (the mask) (Breitmeyer, 1984; Breitmeyer & Ganz,

1976). The use of the term "masking" refers to the usual finding that the

ability to detect the target is usually reduced by the presence of the mask. If

the mask precedes the target in time, the masking situation is referred to as

forward masking; conversely, when the mask follows the target in time, the

situation is referred to as backward masking. The temporal offset between the

target and mask is referred to as stimulus onset asynchrony (SOA) where the

time the target is presented is the reference time or a time of 0. Positive SOAs

refer to backward masking and negative SOAs refer to forward masking.

Visual masking can be grouped into two basic types, labelled Types A

and B. Types A and B refer to the existence of two different empirical

functions relating SOA and the amount of masking observed (Brcitmeyer &

Ganz, 1976). The Type A masking function has the greatest magnitude of

masking when the SOA is 0 msec. As the SOA is increased in either the positive

or negative direction, the amount of masking observed reduces quickly and

symmetrically about an SOA of 0 mscc. By the time the SOA is 200 msec positive

or negative, the degree of masking of the target is negligible (Breitmeyer,

1984; Breitmeyer & Ganz, 1976). The stimuli that produce Type A masking have

overlapping contours. Such masking is referred to as either masking by

noise, if the target and mask bear no structural resemblance, or masking by

structure, if elements of the target are used to construct the mask, though

under some circumstances masking by structure will produce Type B masking

(Breitmeyer & Ganz, 1976).

With Type B masking, the degree of masking is not symmetrical around

an SOA of 0 msec. Backward Type B masking produces a much greater degree

of masking than forward Type B masking. The greatest degree of masking

occurs when the mask is presented 100 to 200 msee after the target. For

forward Type B masking, the masking is greatest with an SOA of approximately

-50 msec. The typical stimulus situation that leads to backward masking

employs a stimulus in which the target and mask do not have overlapping

contours (Breitmeyer, 1984; Brcitmeyer & Ganz, 1976). Type B backward

masking is also referred to as metacontrast, which is the term that will be used

throughout the rest of the present paper.

It is the fact that metacontrast does not require overlapping contours

that has made this phenomenon attractive as a potential mechanism of

saccadic suppression to the almost complete exclusion of consideration of other

forms of masking (Breitmeyer & Ganz, 1976; Dodge, 1900; Matin, 1974). The

hypothesis states that the pattern of stimulation on the retina that results from

a saccade resembles the situation used to obtain metacontrast. Specifically, the

clear image present after the eye movement ends has appropriate

characteristics (spatially and temporally) to mask the perception of the

smeared image that occurs during the eye movement.

Matin, Clymer, and Matin (1972) provided an elegant demonstration of

the role that metacontrast may play in saccadic suppression. In their

experiments, subjects viewed a vertical slit displayed during an horizontal

saccade. If the slit was removed from the subject's view prior to the end of the

saccade, the subject would report the perception of a smeared streak, but if the

slit was left on briefly after the end of the saccade the perception of smear

would be reduced or eliminated. It is important to note how this experiment

fits the metacontrast paradigm. The clear image of the slit followed the

smeared image, and, more importantly, the clear image of the slit was not on

the same retinal location as the smeared image.

The role of metacontrast in saccadic suppression is further supported by

the observations that suppression increases as luminance and scene

complexity increases (Brooks & Fuchs, 1975). Elevations of visual thresholds

can be found when the eye is still and the surrounding scene is moved (Brooks

& Fuchs, 1975). These observations taken together have led E. Matin (1974) to

propose that metacontrast masking exists to serve in saccadic suppression.

Yet, note that for metacontrast to operate efficiently, both sufficient

luminance and a contoured visual scene are needed. Thus, metacontrast is

limited in the range of conditions that it can effectively contribute to saccadic

suppression (Volkmann, 1976;1986), though this range of conditions covers

most situations encountered.

Up to now, all other forms of masking have been ignored as potential

contributors to saccadic suppression. While they have not been given much

treatment in the psychophysical literature (Matin, 1974; Volkmann, 1986),

some of the physiological evidence to be discussed below seems to suggest some

role for both forward and backward Type A masking. The clearest example of a

potential role for Type A masking was shown by Judge, Wurtz, and Richmond

(1980). They found cells in the striate cortex of monkeys that showed an

attenuated response to a stimulus that moved across the receptive field at

saccadic velocities when another discrete stimulus either preceded or followed

the target stimulus. The masking was greatest with an SOA of 0 msec and was

nearly absent when the SOA is either a positive or a negative 50 msec. The

masking stimulus was also presented in the receptive field. The findings were

replicated psychophysically with human subjects (Judge, Wurtz, & Richmond,

1980) and in cells in the superior colliculus (Wurtz, Richmond, & Judge, 1980).

Interestingly, metacontrast has yet to be clearly observed in the single unit


Retinal Shear

The final mechanism proposed that depends upon peripheral events is

the retinal shear hypothesis of Richards (1968, 1969). According to this

hypothesis, the great angular acceleration within the eye, coupled with the

different moments of inertia within the eye tissues, leads to shearing forces

within the retina, and it is these shearing forces that cause suppression. This

hypothesis has rarely been investigated directly but several observations

argue against this factor being a major contributor to saccadic suppression.

First, major supplemental hypotheses are required to account for suppression

prior to and following the eye movement. For example, both a neural delay to

account for suppression prior to the eye movement and a prolonged settling

time to account for suppression following the saccade are required. Moreover,

threshold elevations resembling saccadic suppression in terms of magnitude

and time course have been observed to be associated with blinks and vergence

eye movements (Manning, 1986; Manning & Riggs, 1984; Volkmann, 1986;

Volkmann, Riggs, & Moore, 1980). In neither case are the shearing forces

likely to be as large since the accelerations of the eye are small. Finally,

measurements of saccadic suppression with psychophysical techniques led to

the observation that the magnitude of suppression increases as saccade size

increases (Volkmann, Riggs, Ellicott, & Moore, 1981), which is consistent with

the retinal shear hypothesis since suppression increases as acceleration, and

presumably the magnitude of the shearing forces, increase. But Krantz and

White (in preparation) found the opposite trend using body sway as a measure,

a finding counter to the retinal shear hypothesis. If the retinal shear

hypothesis were correct, increasing the acceleration of the eye should, at

least, not lead to a decrease in suppression.

Inflow or Feedback Extraretinal Mechanism

The inflow model for saccadic suppression is attributed originally to

Holt (1903), who proposed that saccadic suppression was caused by feedback

signals from the extraocular muscles. Several sources of information argue

against a pure inflow model as an explanation of saccadic suppression. First,

the time course of saccadic suppression is inconsistent with the feedback idea

because saccadic suppression begins long before the onset of the feedback

signal from the eye muscles. Additionally, the observed motion of the world

during an eye press, when feedback from the eye muscles is presumably

present, also seems to be inconsistent with feedback theories of saccadic

suppression. It is also interesting to note that an afterimage will not appear to

move during an eye press (this observation works best when the eye is closed

to avoid any induced movement of the afterimage moving relative to the

background) (Hershberger, 1987). In this case, the feedback signal should be

present to indicate the eye moved while the stimulus remained still on the

retina. The sum of these signals should produce apparent motion of the

afterimage if feedback were involved in perceptual stability. The observed

displacement of the world during paralyzed eye experiments also argues

against the inflow model. Since the eye does not move and no feedback can

occur, then feedback models should predict that the world should not appear to

move during saccadic eye movements. The fact that the world does appear to

move in this case suggests that some other extraretinal source must be

involved in saccadic suppression.

If feedback were a good source of extraretinal information for saccadic

suppression, then feedback would be expected to be a major contributor to eye

position sense. Eye position sense refers to the sensory impression of the

direction of the eye relative to the head. In the case of human eyes, afferent

(feedback) information apparently provides only a coarse input to the eye

position sense. Human extraocular muscles are amply supplied with sensory

endings (Cooper, Daniel, & Whitteridge, 1955; Matthews, 1972; Sherrington,

1897; Wolter, 1955), including a similar quantity of muscle spindle organs

sensitive to stretch in other skeletal muscles (Cooper, Daniel, & Whitteridge,

1955; Matthews, 1972). Yet, what role these sensory organs play in an eye

position sense is presently unclear. Their contribution to an eye position

sense would seem to be minimal as noted from the observations in the

preceding paragraph and from the finding that the eye can be displaced

passively to a large extent without the subject being aware of such a

displacement (Brindley & Merton, 1960; Merton, 1964; Skavenski, 1972). The

area that the eye can be passively moved without the subject being aware of

the movement extends to at least 10 degrees to either side of fixation, and this is

only for practiced subjects (Skavenski, 1972). With unpracticed subjects,

displacements of up to 40 degrees to either side can go unnoticed (Brindley &

Merton, 1960).

Outflow of Feedforward (Corollary Discharge) Extraretinal Mechanism

The most popular model for saccadic suppression that arises from

extraretinal sources is the corollary discharge. The initial model of the

corollary discharge (Sperry, 1950) was based upon Helmholtz's (1909/1962)

idea of "effort of will." Basically, the model states that the motor system sends a

copy or efferenz copie of the motor command to the visual system to "inform"

it of the perceptual effect of the intended motor act (Held, 1961; Hoist, 1954).

This model seems corroborated by studies of perceptual rearrangement as long

as the interaction of the corollary discharge with visual signals is assumed to

be to some degree plastic (Held, 1961). The issue of plasticity will be discussed

in greater detail below.

No matter what form the corollary discharge takes, psychophysical

evidence seems supportive of some role for a centrally originating corollary

discharge being involved in saccadic and other forms of suppression. In

many studies, stimuli are presented briefly on diffuse backgrounds to avoid

potential effects of image smear and masking (Volkmann, 1962; Volkmann,

Schick, & Riggs, 1966). To rule out potential masking from the slightly visible

contours of the face (e.g., the nose), Riggs and Manning (1982) placed cut out

ping-pong balls, effecting a ganzfeld, close to the eyes for more complete

removal of contours and found the typical amount of suppression. Perhaps the

most impressive support for a corollary discharge being involved in saccadic

suppression comes from Riggs, Merton, and Morton (1974) who used visual

phosphenes in complete darkness and still found saccadic suppression of 0.4

log units.

Another source of information about any role the corollary discharge

might have in saccadic suppression is from physiological studies of the

activity of the visual system during saccades. Discovering evidence at the

physiological level consistent with a corollary discharge has proven to be a

difficult task. Much of the difficulty may have arisen because the operation of

the corollary discharge may be more subtle and complex than at first had been


A study by Richmond and Wurtz (1980) does find support for such a

function of the corollary discharge in the superior colliculus of the rhesus

monkey. They found cells in the superficial layers of the superior colliculus,

the "visual" layers (Schiller, 1972), that showed a reduced sensitivity to visual

stimulation around the time of saccades. They determined that the reduction

in sensitivity was due to a centrally originating corollary discharge, because

the magnitude of the reduction of sensitivity was the same in light and dark

conditions and after retrobulbar block. Since the reduction in sensitivity

occurred in darkness, masking and smear were eliminated as potential factors.

The retrobulbar block eliminated feedback from extraocular muscles, so inflow

information was not available (Richmond & Wurtz, 1980).

Judge et al. (1980) did not observe such a suppression of response of

cells in the striate cortex of rhesus monkeys. Yet, other investigators have

found some evidence of suppression of visual responses during saccadic eye

movements along the primary visual pathways. Crcutzfeldt, Noda, and

Freeman (1972) have found that some cells in the lateral geniculate nucleus

(LGN) are suppressed by saccades. The observed suppression did not depend

upon the presence of visual stimulation but was only related to the presence of

a saccade. Bartlett, Doty, Lee, and Sakakura (1976) observed suppression of

visual evoked potentials recorded in the optic radiations and striate cortex.

This suppression continued in the dark and was replaced by facilitation 100

msec after the end of the saccade. Yet, these findings do not represent the

entire pattern of physiological results around the time of saccades. In the

same report, Creutzfeldt et al. (1972) found some cells in the striate cortex

being facilitated, while many cells do not respond at all during an eye


There is a sizable negative potential observed in the LGN during

saccades that seems to have an extraretinal origin. The negative potential

during saccades seems to be related to the pontine-geniculo-striate (PGO)

potential first observed during rapid eye movements in sleeping subjects

(Singer, 1976). This negative-going potential does not seem to be inhibitory

but seems instead to function to release inhibition within the lateral

geniculate nucleus (LGN) of the thalamus, the structure where the neurons

leaving the eye synapse with neurons that will carry the information to the

cortex. During normal viewing of a complex visual scene, substantial

inhibition within several levels of the visual system develops (Coren, Porac, &

Ward, 1984). This inhibition is usually thought to be essential to normal

viewing. However, to be useful, the pattern of inhibition must be related to

the pattern of contours in the environment (lateral inhibition, Bekesy, 1967;

Cornsweet, 1970). This pattern of inhibition at the time of a saccade would not

be related to the new visual scene present after an eye movement and could

disrupt the clear viewing of the new scene. It has been proposed (Singer,

1976) that the function of the PGO wave is to release the visual system from

this build up of inhibition so that the new scene at the end of the saccade can

be efficiently processed. Such a function seems essential and may well operate

along side the suppression of visual responsiveness during the saccade. Thus,

physiological studies suggest that a corollary discharge may both suppress

visual sensitivity during a saccade and prepare the central pathways for

processing the next clear image. It should be noted that both facilitation and

inhibition observed at one level of the nervous system may ultimately have

the opposite effect at another level of the nervous system, so such conclusions

should not be held to be firm.

6) Matin's hybrid extraretinal mechanism

In 1976, L. Matin (1976a; see also L. Matin, 1982) published an article in

which he tried to resurrect an inflow theory of saccadic suppression by

combining it with outflow information. Basically, the theory proposes that

feedback from the extraocular muscles is the primary extraretinal signal, but

that this feedback is altered by centrally originating signals. The proposal

states that the corollary discharge exists, but acts only as a gate for inflow


This proposal allows him to account for the observations from eye press

and paralyzed eye experiments in a manner similar to how corollary discharge

explanations account for these observations. In fact, he admits that data from

these two sources do not discriminate between the two hypotheses. Thus, the

hybrid model seems to be counter to the usual scientific tenet of parsimony.

In addition, an observation by Stevens et al. (1976) seems contrary to

Matin's hypothesis. Within the context of the hybrid theory, feedback

information must be able to leave the eye for the illusions observed during eye

presses and paralyzed eyes to occur. In the last experiment in Stevens et al.

(1976), a retrobulbar block was used which prevented both motor commands

from reaching the eye and feedback information from leaving the eye. Still,

they observed that the world appeared displaced after an attempted eye

movement, as measured by the presence of the subject pointing beyond objects

in the direction of the attempted saccade and subject reports. In addition, the

poor feedback mechanisms in the eye make even gated inflow information

about eye position of a dubious quality (Brindley & Merton, 1960; Steinbach,



This first chapter has heavily emphasized the literature on saccadic

suppression from both psychophysical and physiological research and thus

has emphasized the first of the two questions related to vision around the time

of saccadic eye movements: why the smear that results from executing a

saccade is not noticed. This research has primarily focused on elevations of

some type of threshold during the saccading eye as compared to the fixating

eye. Suppression is incomplete, typically measured at about 0.5 log units (E.

Matin, 1974; Volkmann, 1976; 1986). The actual degree of suppression depends

upon numerous factors including the size of the eye movement, luminance,

and complexity of the visual scene (Brooks & Fuchs, 1975; Volkmann et al.,


This observed suppression is best accounted for by three factors

working in conjunction. One important factor seems to be visual masking (E.

Matin, 1974). Usually metacontrast masking is invoked where the clear image

at the end of the saccade masks perception of the smeared image during the

saccade. Yet, some physiological evidence suggest that both forward and

backward Type A masking may be involved. One limiting factor about masking

is that it depends upon sufficient luminance and the presence of contours to

be effective. Retinal smear is another mechanism that could act to cause

saccadic suppression (Volkmann et al., 1978). The major extraretinal source of

saccadic suppression appears to be a centrally originating corollary discharge

which can operate under any visual condition (Vo!kmann, 1976, 1986). Other

potential extraretinal sources of saccadic suppression, inflow and L. Matin's


(1976a) hybrid mechanism, do not seem to adequately account for all of the eye

press and paralyzed eye data. While Richard's (1968, 1969) retinal shear

hypothesis has not been greatly examined, it has difficulty with suppression

observed during blinks and vergence eye movements and the decrease in

"suppression" for larger saccades when body sway is measured (Krantz &

White, in preparation).

The next chapter will discuss previous research on the second question

of how the world maintains a perceptual stability around the time a saccade

occurs. After this discussion, the first experiment will try to resolve some

conflicts in the data gathered to date.


Previous Research

The importance of studying what happens to the sense of

direction during a saccade is related to the second question mentioned

in the introduction. Namely, why the world is perceived to remain

stable during a saccadic eye movement. Principally, the eye press

experiments and paralyzed eye experiments seem to suggest that some

sort of centrally originating corollary discharge seems to be involved.

This idea was originally espoused by Helmholtz (1909/1960) and

expanded by Hoist (1954) and Held (1961). In general, the hypothesis

states that any intended movement leads to a corollary signal that

counters the sensory signal which results from the movement and

which might indicate motion of the world. In effect, the sensory signal

indicating that the world is moving is cancelled, leaving the perception

of stability. Note that in this situation, a corollary discharge does not as

much inhibit visual functioning as it selectively cancels incoming

sensory information that is expected to be the result of the intended


L. Matin and his colleagues (Matin, 1972; Matin, Matin, & Pearce,

1969, 1970; Matin & Pearce, 1965; Matin, Pearce, Matin, & Kibler, 1966)

were the first to investigate these ideas directly as they related to

saccadic eye movements. In their paradigm, they presented briefly

flashed stimuli either just before, during, or after the saccade and their

subjects were asked to judge whether the flashed stimulus was to the left

or to the right of the initial fixation target which had been removed at

least 500 msec earlier. Their results indicated a slow readjustment of

perceived direction, consonant with the change of eye position,

beginning well before the saccade and lasting about 300 msec after the

saccade. From their data, Matin et al. (1969, 1970) argue that the

corollary discharge portion of perceptual readjustment takes up to 0.5

seconds to complete. Another feature of these data to note is the wide

intersubject variability in their judgements before, during, and after

the saccades. Such intersubject variability is unusual in a large

proportion of psychophysical studies.

Matin observed another pattern of results when judgements of

direction were made to a briefly presented stimulus presented earlier in

the same saccade (Matin, 1976b). Matin (1976b) refers to this study as

the two-flash paradigm. In this case, perceived direction seemed to be

locked to the same retinal location of the first flash, regardless of its

spatial direction, until after the end of the eye movement when a more

rapid readjustment of perceived visual direction occurs. Matin (1976b,

1982) argues that these results do not reflect the operation of the

corollary discharge but the effects of visual persistence of the first

flash overriding the corollary discharge.

By discounting the results for the two-flash paradigm, L. Matin

uses the results from the first experiments to argue that the corollary

discharge is a slow signal. Specifically, he argues that the corollary

discharge takes a relatively long time to readjust perceived direction to

match the new spatial coordinates resulting from the eye movement.

Thus, the corollary discharge is not the primary factor in perceived

stability around the time of eye movements. In fact, he argues that

visual context, even as primitive as a single dot fixed in space, can

provide sufficient information to readjust the perception of direction

(Matin, 1972, 1976b).

The conclusions that Matin draws from his data could be

incorrect for two basic reasons. First, Matin (1976b; Matin et al., 1969)

does not provide any evidence to support the supposition that the

persisting image of an object controls the perception of the direction of

that object. His argument is based on the observation simply that

persistence exists (Bowen, Pola, & Matin, 1974). The possibility remains

open that persistence may not explain the results from the two-flash

paradigm. If this is the case, then there is no a priori reason to believe

that the data obtained when judgements were made to the initial fixation

target are any more representative of the action of the corollary

discharge than data from the two-flash paradigm data.

In fact, it is possible that the temporal separation between the

initial fixation target and the flashed stimulus whose direction is to be

judged in the original experiments (Matin et al., 1969, 1970) could have

been responsible for the pattern of results observed. Since, the

judgements of direction are made to a location in space that is only

remembered, errors in memory could have contaminated the results in

these experiments, perhaps explaining the wide intersubject variability

(Hershberger, 1987; Skavcnski, 1976; Skavenski, & Steinman, 1970).

Such considerations lead to difficulty in interpretation of Matin's data.

Hershbergcr (1987) advanced a faulty-memory explanation to

account for the data Matin ct al. (1969, 1970) observed, but only for the

data following the saccade. The linear trend in the changes of judged

direction following the saccade fits Skavenski's expectancy for the

pattern in the data if memory were to account for the results.

Hershberger (1987) does not use a memory explanation to account for

the results Matin ct al. (1969, 1970) obtained for changes in perceived

direction prior to the saccade for two reasons. First, the flashes are

closer in time to when the initial fixation target was removed, and the

eye has not moved, so memory for the location of the initial fixation

target should not be as much a problem. Secondly, the curvilinear

trend in the data does not fit with Skavcnski's (1976) expectation of the

linear trend the data should take if memory is the explanation.

Hershberger (1987), instead, finds the pattern of results for

flashes presented prior to the saccade consistent with his own model of

the readjustment of the perception of direction around the time of

saccades. His model includes von Hoist's (1954) contention that the

efference copy or corollary discharge is present and gives highly

accurate information about the position of the eye at any moment, but

this signal is not used when making psychophysical judgements.

Instead, he proposes that there is also an afference copy (Hershberger,

1976, 1987) which he conceives of as a reference signal indicating

where the eye will be after the end of the eye movement. The sense of

"copy" used in the term "afference copy" is something to be duplicated

or matched, whereas "copy" in the term "efference copy" has the sense

of something that is itself a duplicate. The afference copy serves as a

reference signal that the eye movement mechanism attempts to match

such that the eye movement sequence is complete when the efference

copy is the same as the affcrence copy.

Hershberger's (1976, 1987) model of the perception of direction

around the time of saccadic eye movements relies on the validity of

Robinson's (1975, 1987) model of saccadic eye movements. Unlike

typical models of the control of saccadic eye movements (Carpenter,

1977; Clark & Stark, 1974), Robinson proposes that saccadcs are not

ballistic, nor is eye position sampled around the time of saccades, as

opposed to continuously during smooth pursuit movements. Robinson

(1975, 1987) proposes that eye position is continuously monitored even

during saccadcs and that saccades are even mutable during their course.

He argues that since the latency to execute saccades is long relative to

the time it takes a normal saccade to be completed (Alpern, 1971), it only

appears that saccades are ballistic. In fact, Robinson believes that the

control signal used to generate saccadcs, and all eye movements, is

derived from a signal that is proportional to the final position of the eye

(an affercnce copy?). Given this model of saccadic eye movements, it

seems more reasonable to propose some sort of reference signal against

which current eye position is compared. If saccades are truly ballistic

then eye position would not need to be monitored during a saccade.

Using Robinson's (1975, 1987) model of saccadic eye movement,

Hershberger (1987) proposes that the affcrence copy controls the

perception of direction as measured by psychophysical techniques.

Since the reference signal, the afference copy, is present prior to the

execution of the saccade, perceived visual direction is proposed to shift

in the direction of the saccade by the size of the intended eye movement

prior to the eye movement. Hershberger (1987) makes a supplementary

addition to his hypothesis to account for the curvilinear nature of the

data of Matin et al. (1969, 1970) for judgements of direction prior to the

saccade. To explain these data in a manner consistent with his theory,

Hershberger proposes that the time that the afference copy takes effect

and controls judgements of direction is variable with respect to the

onset of the saccade. Thus, as the onset time for the saccade nears, the

probability that the afference copy has been set up and is now

controlling "conscious" judgements of direction increases. In this

manner, the curvilinear nature of Matin et al.'s (1969, 1970) data can be

accounted for.

The Present Study

The present study was designed with two purposes in mind. First,

it seems important to test directly Matin's (1976a) contention that the

pattern of results obtained from the two-flash paradigm was due to

visible persistence (Coltheart, 1980). One way to test this hypothesis is

to minimize the time interval between presentations of the stimuli used

to judge direction while also reducing the possible effects of visible

persistence. The present study will therefore minimize visible

persistence by making all the stimuli decrements from a background

luminance, which leads to far less persistence. To further reduce the

amount of visible persistence from the stimuli against which direction

is to be judged, these stimuli will be left on for at least 2.5 seconds.

Increasing stimulus duration tends to reduce the persistence of the

stimulus following its offset (Coltheart, 1980).

Still, a simple replication of Matin et al.'s (1969, 1970) paradigm

will not be able to address Hershberger's (1987) hypothesis. If

Hershberger's proposal is correct, then judgements of visual direction

for stimuli presented prior to saccade onset should break down into two

distributions. One distribution, representing judgements made prior to

the afference copy being present, should center on the actual position

of the reference target (the initial fixation target in Matin, Matin, and

Pearce's experiments). The other distribution, representing judgements

made after the afference copy is present should center around the

position the reference target would occupy on the retina after the

saccade was executed. In other words, these judgements should be

shifted in space with the same direction and size as that of the upcoming


In Matin, Matin, and Pearce's original experiments, the above

proposal could not be disproved. Yet, a minor adjustment to the stimulus

situation would allow for judgements contrary to or consistent with

Hershberger's model to be observed. To accomplish this aim, a stimulus

will be placed halfway between the initial fixation target and the

saccade target stimulus, and judgements of the position of the test flash

will be made relative to the middle stimulus (see Figure 1).

Four, instead of two, responses are now possible. A judgment of

"far right" indicates that the test flash appeared to the right of all three

initial stimuli. A judgment of "right" indicates that the stimulus

appeared between the rightmost target and the middle target.

Judgements of "left" and "far left" have comparable meanings as

presented in Figure 1. Critical target flashes to test Hershberger's



"far left"



"right" I "far right"



Figure 1. Stimulus arrangement used for Experiment 1. The eye movement is from
the left outer target to the right outer target after the targets are removed.
Judgements of the direction of stimuli flashed at locations 1 and 2 are critical for
Hershberger's hypothesis. See text for an explanation.

hypothesis will be presented in locations 1 and 2 as shown in Figure 1.

To understand the following discussion, it is important to note that all

target flashes will be presented after the three initial stimuli are

removed, and that all saccades are to the right. This second point is

important as perceived direction should shift to the right, if perceived

direction anticipates the saccadic eye movement. Prior to the saccade, a

test flash presented in location 2 should not ever be judged to be "right"

because the hypothesized afferenee copy leads to a discreet

readjustment of judged position having the extent of the intended

saccade. Either the target presented at location 2 should be judged to be

"left" (before the perceived direction shifts) or "far right" (after the

perceived direction shifts). Judgements of "right" to test flashes

presented in location 2 indicate a shift in perceived direction, but this

shift is less than the extent of the intended saccade. Thus, a judgement

of "right" under these circumstances is inconsistent with Hershberger's

hypothesis. Using the identical line of reasoning, under Hershberger's

hypothesis, a stimulus presented in location 1 prior to the saccade

should be judged either "far left" or "right," but not "left."



Three male subjects, volunteers, were run in this experiment.

Their ages were 24 to 30 years old, and all had normal acuity.


The experiment was run on a Commodore 128 microcomputer.

The computer presented stimuli, controlled and measured timing,

collected subjects judgements and analyzed the data. Data were stored on

a disk and analyses were performed off-line. The stimuli were

presented on a Magnavox RGB Monitor 80 (Model #CM8562), using the

green phosphor only. The decay of the phosphor was measured to be 4

msec to 10% of the original luminance. The light color reflective case

of the monitor was masked off to flat black to reduce stray light on the


The monitor screen was viewed through a high powered cylinder

lens placed 24 inches (60 cm) from the face of the screen. Viewing was

accomplished by placing the right eye of the subject just behind the

lens (Figure 2). The effect of this optical arrangement was to cause a

vertical integration of the light across the screen. For example, a black

square presented against a white background would be seen as a

vertical gray bar the width of the black square but extending vertically

the entire field of view. With this arrangement, the vertical position

along the screen raster could be used to time stimulus presentations for

shorter periods of time than the typical screen frame (16.67 msec).

Saccadic eye movements were detected using an amplified

electrooculogram (EOG). Appendix A contains a description of the

amplifier and trigger device used. The amplified EOG signal was fed into

a level detector that upon the detection of a rapid voltage change

associated with saccades, sent a pulse to the Commodore 128

microcomputer to generate an interrupt so the computer would record

the time of the saccade onset. The interrupt could be detected with an



Figure 2. Optical placement of the monitor, cylinder lens and subjects right eye in
Experiment 1.

accuracy of 65 microseconds. The amplified EOG and trigger pulse were

also fed into and displayed on a Techtronix Oscilloscope for online



All of the stimuli used were 25' of visual angle wide. Each subject

participated in at least 1000 trials, across several sessions lasting about 2

hours each. The EOG electrodes were placed bitemporally with a ground

electrode placed on the left mastoid bone, just behind the ear.

Subjects began every trial by pressing a button on a joystick.

This button press initialized a variable delay period of 2 to 4 seconds.

Then the three initial stimuli were presented (Figure 3). Subjects were

instructed to fixate on the stimulus to the left. The stimulus to the far

right was the saccade goal, indicating a saccade of 6.8. The middle

stimulus is placed exactly halfway between the other two targets. Then,

2 seconds after the three initial stimuli were presented, a tone was

sounded for a duration ranging from 0.5 to 1.5 seconds. The subject was

instructed to execute the saccade to the far right stimulus when the tone

was turned off. The three initial stimuli were removed 150 msec after

the end of the tone.

The flashed probe stimulus was presented with a variable delay of

0 to 300 msec following the offset of the three initial stimuli. This

variable timing allowed the flashed probe stimulus to be presented

before, during and after the saccade had occurred. The time when the

probe stimulus occurred relative to the saccade onset was determined by

the computer which kept track of when the probe stimulus was

presented and when the saccade onset was detected.


Eye Movement








3 initial stimuli Probe

0.5- <
2-4 sec. 2 sec 1.5 ;

150 mseco 300 msec


Figure 3. (a) Stimulus arrangement presented to the subject in Experiment 1. The "X"'s
indicate the locations used to flash the test probe stimulus. (b) The timing of events in
a single trial in Experiment 1.

The position of the flashed probe stimulus was also variable,

being presented in 3.40 steps starting 5.10 to the left of the fixation

stimulus (the leftmost of the three initial stimuli) and ending 1.60 to the

right of the saccade goal (the rightmost initial stimulus). Thus probe

stimuli were presented to the left of the fixation stimulus, halfway

between the fixation stimulus and middle stimulus, and halfway between

the middle and saccade goal stimulus as well. Probe stimulus positions

are indicated by X's on Figure 3.

After the end of every trial, the subject was instructed to make a

judgement of the position of the flashed probe stimulus relative to the

middle initial stimulus. Judgements of "far left" were to indicate that

the subject perceived the probe stimulus to be to the left of where all

three initial stimuli were located, "left" indicated the subject perceived

the probe stimulus to be between the left fixation and middle initial

stimuli, "right" indicated that the probe stimulus appeared between the

middle and saccade target initial stimuli, and "far right" indicated that

the probe stimulus appeared to the right of all three initial stimuli. An

experimenter entered the judgement in the computer for storage on

disk. The experimenter also monitored the EOG and trigger signals on

the oscilloscope. If there was either a false trigger or if the probe

stimulus was not perceived (subjects were encouraged to guess), the

experimenter rejected the trial which was signaled to the subject.

The first two experimental sessions began with 10 practice trials

which were identical to experimental trials except that the data were

not stored on the disk. Sessions were not run to collect a predetermined

amount of data but instead for a period of approximately 2 hours. The

session could be terminated sooner is the subject requested. Breaks

were at the subjects discretion with at least one break after an hour.

Data Analysis

The times when the stimulus probe occurred relative to the onset

of a saccade were grouped into 25 msec bins relative to the time of the

onset of the saccade to give more stability to the data. The onset of the

saccade was the center time for the 0 msec time bin. Positive times

indicate that the probe stimulus occurred after the onset of the saccade.

Within each time bin, the data were grouped by the position of the

probe stimulus. Within the positions the data were also grouped by the

judged position of the probe stimulus. Thus, in each time bin there were

20 groups of judgements (5 probe stimulus positions x 4 judgements).

This data arrangement was used for the analyses described below.


In order to compare the present results with Matin's earlier

reports (Matin, 1972, 1976a; Matin et al., 1969, 1970), the point of

subjective equality (PSE) was determined for the position of the probe

stimulus relative to the middle initial stimulus at each of the 25 msec

intervals. To make the data more similar, judgements of "left" and "far

left" were grouped together for the present analysis as "left", and

judgements of "right" and "far right" were grouped together as "right".

The PSE was calculated as the probe position where subjects responded

"left" on 50% of the trials. If no probe position was at exactly the 50%

position, a linear interpolation was made between the two flash

positions that bounded the 50% position.

Figure 4 shows the calculated PSE in terms of spatiotopic

coordinates for probe position as a function of the time relative to

saccade onset. On Figure 4, the 0 location refers to the spatiotopic

location of the middle initial stimulus. The three panels report data

separately for each of the three subjects. Four features of the data are

worth noting. Prior to saccade onset, all three subjects are reasonably

accurate in their judgements of direction, as indicated by PSE near 00.

Secondly, all three subjects make errors in judged location in the

direction towards the left (PSE < 0) near the time that the saccade begins.

This would be consistent with perceiving that the eye has already

moved to some extent to the right, the direction of the saccades in this

experiment. The third feature is the tendency for judgements of

direction to be more accurate during and shortly after the onset of the

saccade. Typical 6.80 saccades take approximately 35 msec (Bahill, Clark,

& Stark, 1975), so the 25 and 50 msec bins are important for this

observation. The final feature is the breakdown of consistency of

judged direction across subjects after the saccade has ended (bins 75

msec and greater). Two subjects judge the stimulus to be too far to the

left (negative PSE's of up to 3.50), while one subject (CBW) remains

highly accurate in his judgements (PSE's near 00).

There are several important differences between the present

pattern of results and the pattern observed by Matin ct al. (1969, 1970).

These differences are more easily observed by representing the same

data as in Figure 5. Figure 5 presents the same data as in Figure 4 but

recalculated to reflect the position that the middle initial stimulus would




o -2

*" -3


0 0

0 -0




Figure 4. Point of subjective equality (PSE) between probe stimulus and middle
initial stimulus. A PSE of 0 deg represents the actual spatial position of the middle
initial stimulus. Left of the central initial stimulus is < 0.



-100 -50 0 50 100 150 200
Time Relative to Saccade Onset

have on the retina before, during and after the saccade (a retinal locus

of 0 is the position of the fovea and negative positions are to the left).

Thus, these data are in terms of retinotopic coordinates. The dashed line

in Figure 5 represents the position on the retina that the middle initial

stimulus would occupy if it had been present during the saccade. The

present figure is analogous to Figure 10 in Matin (1972). Matin's (1972)

data is different from the present data in two important respects: 1) He

observed the changes in perceived direction beginning about 200 msec

before saccade onset and lasting up to 300 msec thereafter. In the

present data, the change in perceived direction is closely tied to the

onset of the saccade, not being appreciably present until the 0 msec bin

(which does contain some judgements up to 12.5 msec before the saccade

begins). Moreover, the present data suggest perceived direction

stabilizes by only 100 msec after saccade onset, not the 300 msec

reported by Matin. 2) The data on the three subject's of Matin et al.'s

(1969, 1970) are widely different in the accuracy, relative to the actual

position of the initial fixation target in his experiments. No two of the

subjects are at all similar in their judgements before or after the

saccade. In the present experiment, a much greater degree of

intersubject agreement was obtained. All three subjects were similar in

their judgments before, during, and immediately after the saccade

(Figure 4 and 5). After the saccade the three subjects diverged with two

subjects overestimating the extent of the saccade while one accurately

estimates the extent of the saccade (represented by judgements more

negative that the position of the dashed line in Figure 5 after the

saccade was finished).

0 4
".- -6

S -+4
S- S = JHK


-0 -2

4 -

S-a 6 -

C., I I I I I I I I I I I I
.o +4 -
..A-.. S = RLD
0 +2

0, 0



-6 -

-100 -50 0 50 100 150 200
Time Relative to Saccade Onset
Figure 5. Corrected Point of subjective equality (PSE). Same data as in Figure 4 but
replotted to indicate the position of the PSE on the retina. The fovea is set as a PSE of 0.
The dashed line indicates the position on the retina where the middle initial stimulus
would be imaged during a typical saccade, if present.

Still, the data are consistent with Matin's proposal that the data

from the two-flash paradigm are affected by visible persistence.

Despite the much closer proximity in time between the offset of the

initial stimuli and the probe stimulus in the present experiment, there

is no indication that judged direction maintains the same retinal locus

until after the end of the saccade as in the above experiment (Matin,

1976a). Instead, the data follow closely the spatial position of the initial

stimuli and, if anything, anticipate the eye movement to some degree.

The analyses to present have not allowed any assessment of

Hershberger's hypothesis that perceived direction changes discretely

and not in the continuous manner suggested by the above data

presentation, especially in Figure 5. To test Hcrshberger's, proposal it is

only necessary to look at the judgements made prior to or just after the

saccade begins (-50 to 0 msec bins). Additionally, examining only the

judgements from the two critical probe stimulus locations are important

(1 and 2 in Figure 1). These prove stimulus locations are the only

relevant locations because these are the only two probe stimulus

locations for which judgements representing half the distance of the

eye movement (3.40) and the full distance of the eye movement (6.80)

are possible. Combining the data from these two stimulus locations and

across subjects leads to the data presentation in Figure 6. If the stimulus

at location 1 listed in Figure 1 is judged "far left", then that judgement

falls into the 00 judged category. A judgement of "left" falls into the 3.40

category since the center of the "band" for judgements of "left" is 3.40 to

the right of that probe stimulus location. Note that this band roughly

represents a mislocalization of half the size of the eye movement. A

judgement of "right" falls into the 6.80 bin for similar reasons. These









0) U) 0) CD
C0 N- U) C

uig owi.L ui sluawaF~pnr io a6~juoadi

O U)


E ,

E _~


O, 0

0) >,
2 a)

) a,

a, a,
0 0
E ,

(D 0
0) it a
u~ C's

C) Co )

C) E 25



:3 0
il 0c

judgements of "right" represent mislocalizations of roughly the size of

the intended saccade. For probe stimuli positioned at location 2, the

judgements of "left", "right", and "far right" fall into the 0, 3.4, and 6.80

bins, respectively. The bins were then converted to percentages of the

total number of judgements that fall into each time bin since the

numbers were not equal.

If Hershberger's hypothesis regarding the afference copy were

correct then Figure 6 should show progressively more judgements in

the 6.8 position as the time bin is closer to saccade onset. Moreover,

there should not be an increase in the judgements made in the 3.40 bin

beyond that observed in the -100 msec bin which is shown as a

comparison time period where judgements of direction were accurate

(Figures 4 and 5). As can be seen, the predicted pattern does not hold at

all, but instead judgements gradually become more concentrated in the

3.40 position until during the 0 msec time bin half of all judgements of

direction fall in the 3.40 position bin, while only 9% of the judgements

in this time bin fall the full 6.80 to the left. A chi-square test was

performed on the data in the 3.4 position to test whether the increased

judgements found in the 3.40 time bin represented a reliable increase.

The proportion of judgements in the -100 msec time bin was used to

determine the expected number of judgements for each of the other

time bins as indicated in Table 1.

From these data, a chi-square was calculated for the three time

bins (-50, -25 and 0 msec) on the possibility that the observed frequency

at 3.4 differed from the expected frequency. The difference was

significant (X2 = 190, df = 2, p < .025). Given the fact that in the 0 msec

bin contains judgements from trials where the eye has begun to move,

it could be argued that judgements could occur in the 3.4. The basis of

this argument lies in the fact that the actual eye position catches up

with the perceived position of the eye under Hcrshberger's proposal.

Thus, a second chi-square test was performed using only the -50 and -25

msec time bins. This chi-square was also significant (X2 = 50.3, df = 1, p <

.025). Both tests indicate a large increase in the proportion of

judgements of direction to be in the in-between position (3.40), with few

judgements ever reaching the position representing the full extent of

the eye movement (6.80). This test also indicates that perceived

directions does begin to move in the direction of the eye movement

prior to the onset of the saccade.

Table 1
Actual and Expected Frequency of Judgements of Varying
Direction Prior to Saccade Onset

Frequency of Judged Position Relative to Actual Position (deg)
Time Relative to 0 3.4 6.8
Saccade Onset

Actual 89 8 0
(Percentage) (91) (8) (0)

Actual 84 12 0
(Expected) (88) (8) (0)

Actual 82 30 1
(Expected) (104) (9) (0)

Actual 29 35 9
(Expected) (64) (6) (0)


Basically, the results indicate that judged direction remains

relatively closely aligned to actual direction with the deviation not

becoming significant until about the time the eye movement begins. At

that time, for all subjects the judged position becomes more accurate

immediately after the end of the eye movement, while two subjects (JHK

and RLD) show errors indicating that the eye was perceived to move

farther than the actual extent of the saccade. Moreover, the results are

not at all consistent with Hershberger's hypothesis that the afference

copy controls psychophysical judgments of direction. The pattern of

perceived mislocalizations prior to the beginning of the saccade are

much more indicative of a gradual readjustment of perceived direction

that does at least begin but does not finish prior to the onset of the

saccade. This interpretation is supported by introspective reports made

by all three subjects. All three reported on some trials perceiving the

flash probe to move to the right. The motion was perceived to be

continuous. Such a perception might be the result of the probe stimulus

being perceptible for a period long enough prior to saccade that its

perceived position was displaced while visible. Of course, the

observations are only informal.

While the results are not strongly consistent with either Matin et

al.'s (1969, 1970) data or Hershberger's hypothesis, the results do appear

to replicate Bischof and Kramer (1968). They observed perceived

mislocalizations prior to the onset of the saccade in the direction of the

intended eye movement. Their data indicated a rapid readjustment of

the perceived coordinate system just prior to the saccade in a manner

similar to what is observed here. The implications of the present

experiment, along with that of Bischof and Kramer's (1968), is that the

corollary discharge may be more able to be responsible for the shifts in

perceived direction necessary for perceptual stability than indicated by

Matin's (1972, 1976a) proposal of a slow corollary discharge. Yet, the

corollary discharge seems to operate in a continuous manner more

similar to an efference copy than Hershberger's proposed afference


Still, how does one account for the slow change in perceived

direction observed in Matin's (1972) display of the data of Matin et al.

(1969, 1970). Hershberger (1976, 1987) placed a great deal of weight on

the fact that the data were curvilinear in arguing that the data obtained

by Matin et al. (1969, 1970) prior to the saccade was not due to memory

factors. Hershberger, like Skavenski (1976), expected that memory

factors should show a linear trend. Yet, if their expectation that the

trend should be linear were in error, then in fact memory factors could

explain the entire slow trend observed in the data of Matin et al. (1969,

1970) data.

The results of the present experiment raise some interesting

questions. To best present these questions it is necessary to explore in

more detail the observations made by Stevens et al. (1976) in a paralyzed

eye experiment. Stevens et al. had been concerned that some of the

perceptions of motion of the world made during attempted saccades

when the eye was paralyzed actually resulted from small residual

movements of the eye. To test this hypothesis, one of the subjects used

by Stevens et al. was induced into extremely deep paralysis which did

not allow any residual movements. In another procedure, the subject

was given a retrobulbar block which prevented eye movements while

allowing movement of the remainder of the body. In both cases, the

attempt to make a saccade was accompanied by a displacement of the

perceived location of the world. Motion of the world was not perceived.

These results were one of the observations that lead Hershberger (1987)

to propose that the afference copy controlled readjustment of perceived

location in a discrete manner. Given the results of the present

experiment that perceived location is readjusted in a continuous

manner, why is this motion not perceived, particularly in the

experiment of Stevens et al. (1976)? One possible reason could be that

there is sufficient suppression of the ability to perceive motion that

during a saccade the smooth displacement of the world would not be

perceived. The second experiment will explore this possibility.


There have only been a few attempts to measure the degree to which a

subject's sensitivity to motion detection is reduced during saccadic eye

movements. The general conclusion is that motion sensitivity is reduced

somewhat, but not nearly to the extent necessary to obscure motion of the size

created by the extent of the eye movement (Bridgeman, Hendry, & Stark, 1975;

Heywood, 1981; Mack, 1970; Stark, Kong, Schwartz, Hendry, Bridgeman, 1976).

The two most systematic studies find detection for motions larger than about

1/5th to 1/3rd the size of the saccadic eye movement (Bridgeman et al., 1975;

Mack, 1970). Such a meager amount of suppression can not account for the

apparent degree of motion suppression observed by Stevens et al. (1976) in

their paralyzed eye experiment.

Careful analysis of the experimental situations used in the various

experiments yield some possible explanations for the apparent discrepancies

between the two sets of observations. In the studies reporting on motion

detection, it is possible that the studies actually only measure sensitivity to

displacement of a stimulus to a new location after the eye movement had

ended. By displacement, it is meant that the subject perceives the stimulus is

in a new location but does not perceive the actual translation of the stimulus,

that is, the subject does not perceive the actual motion of the stimulus. Such a

confound is possible because these experiments employ stimuli that are 1)

asymmetrical, 2) nonrepeating, and 3) present before and after the end of the

saccade. These three elements allow the subject to perceive that a

displacement had occurred in the stimulus without perceiving the stimulus

traversing the positions in between the end points of the motion.

For example, Mack (1970) used a single dot on an oscilloscope that was

moved during the eye movement in response to a signal derived from the EOG.

Bridgeman et al. (1975) used a pattern of 13 fixation lights. The subjects were

instructed to make saccades at will along these lights and it was these patterns

of lights that were moved during the saccades, i.e., the target fixation light

would move during the saccade. Moreover, the mechanical system used by

Bridgeman et al. (1975) to make stimulus motions could be heard by the

subjects. Bridgeman et al. admitted that the estimate of reduced motion

sensitivity during eye movements was conservative.

Basically, these previous experiments suggest that motion sensitivity is

reduced during eye movements. Despite the biases against finding such

suppression, the results in all cases do display some reduction of sensitivity to

motions over the condition where the eye is still. Yet, these experiments do

not provide an adequate measure of the degree of motion suppression. To

obtain a better estimate of the magnitude of motion suppression, at least two

alternatives are possible. First, the entire stimulus configuration that is in

motion could be only presented during the saccadic eye movement. With this

procedure, no displacement would be present, but unless the stimulus is

chosen very carefully, motion detection in this situation could become

confounded with detection of luminance changes. The second alternative

would be to use a stimulus that moved in a specific way so that the stimulus

configuration was identical both before and after the end of the eye

movement. Again, no displacement information would be present outside of

detecting the motion. This alternative could be realized by presenting a

grating to the subject and moving that grating during the saccade across

distances that are whole multiples of the spatial period (the inverse of the

spatial frequency). If this motion is accompanied by removal of cycles at the

edge towards which the motion occurs and addition of cycles at the other end

of the stimulus, then the static stimulus will not change at all relative to the

subject either before or after the saccade has occurred. Using this

methodology, the degree that sensitivity to motion detection was reduced

during saccadic eye movements was measured in the second experiment.



Four subjects (3 male, I female) were used, ages 20 to 30 years old. All

subjects had normal acuity in the eye used in the study. One subject (JBM) was

discovered to have a mild ablyopia in her left eye after she ran in the



The apparatus was the same as in Experiment I.


In this experiment the viewing distance was increased to 32". The

electrode placement to record the EOG was identical to that used for the first


The procedure for the present experiment employed a two-alternative-

forced-choice paradigm, with two types of trials, saccade and nonsaccade.

During saccade trials, an initial fixation bar will appear superimposed upon a 2

cycle/0 horizontal square wave grating with a Michaclson contrast of 16%.

The duty cycle was 40% lighted. The subject was instructed to fixate on this

target. After 1 second, this initial fixation target was extinguished and

simultaneously a second fixation bar was presented 20 to the right. The second

target was on for 180 msec, shorter than the normal latency for a saccadic eye

movement. This sequence of events was then repeated with both periods being

subject-initiated. Randomly, in either the first or the second time period, the

grating moved 1, 2, 4, 8, 12, or 16 complete cycles (or 0.5, 1, 2, 4, 6, 80). During

the motions, the contrast of the grating reduced to 12%, still above the contrast

threshold for still gratings during 20 saccadic eye movements (Volkmann,

Riggs, White, and Moore, 1978). The motion was triggered off the horizontal

EOG so that the motion occurred during the early portion of the eye movement

when suppression for motion, as well as detection, appears to be maximal

(Bridgeman et al., 1975; Volkmann, et al., 1968). The subject's task was to report

whether the motion of the grating occurred during the first or second period.

Both the EOG and the trigger derived from the EOG were monitored during both

periods of the trial. The trial was rejected and repeated later if a false trigger


The nonsaccade trials were identical except that the second target never

occurred. The motion of the grating occurred briefly after the end of the first

target either during the first of second period. As in the saccade trials, the

subject was to report during which period the motion occurred.

To better compare the present results with the results obtained by

Bridgeman ct al. (1975), the velocity was kept constant across all motion sizes.

The velocity was 770/second. Thus, the different motions sizes confound

different motion extents with different durations.

Subjects were run in 4 sessions of 180 trials each (not including trials

rejected). Within each session there were 6 blocks of trials (3 saccade and 3

nonsaccade). There were 120 saccade trials, 40/block, and 60 nonsaccade trials,

20/block. During a session, each motion size was presented 20 times during

saccade trials and 10 times during nonsaccade trials. Half of the motions for

each size and each trial type were to the left and half were to the right. Block

type was alternated within each session with the type of block presented first

during the first session randomized across subjects. After the first session, the

block type presented first was alternated, that is, if a block of saccade trials

was presented first during the first session, then a nonsaccade block of trials

would be presented first during the second session, et cetera.

The responses were input into the Commodore microcomputer by the

experimenter for offline analysis. Threshold of motion detection was

calculated at the motion size detected on 75% of the trials, corresponding to a d'

of 1 in signal detection methodology (Green & Swcts, 1966). If no motion size

was detected exactly 75% of the trials then a linear interpolation between the

two motion sizes on either side of the 75% detection level was performed to

estimate the motion size that would be detected 75% of the trials.


Thresholds did not differ for motion to the right or to the left during

nonsaccade trials. Consequently, the data were combined and will be reported

together. Motion to the right during saccade trials will be referred to as

"motion-with" since the stimulus motion is in the same direction as the eye

movement. Motion to the left will be referred to as "motion-against" as the

stimulus motion is in the opposite direction of the eye movement.

Figure 7 plots the psychometric function for each subject relating the

percentage of correct detections to the size of the stimulus motion. For all four

subjects the psychometric functions for the nonsaccade functions have

steeper (that is, larger) slope than those obtained from saccade trials. Without

exception, detections were perfect for nonsaccade trial motions 40 and larger.

Perfect detections were never observed for any motion size during saccade

trials. Subject CBW approaches completely perfect detection during saccade

trails, especially for motion-with trials where detection of greater than 90%

correct were observed. The shallower slopes indicate that using a stricter

criterion for calculating the threshold, for example 90%, would indicate a

larger magnitude of suppression during saccades than reported in Table 2

below. Thus, the thresholds reported in Table 2 represent a conservative

estimate of motion suppression. Two more features of note: 1) for the three

subjects that reached threshold during saccade trials (GER, JHK, CBW) the slope

of the motion-against trials up to at least threshold is lower than for motion-

with trials, and 2) for all subjects, the percentage of correct detections for 60

motions during motion-with saccade trials are less than either 40 or 8

motions. In fact, for subjects JHK and GER, the percent correct detections

during the 6" motions during motion-with saccade trials fall below the 75%

threshold levels. The thresholds to be reported in Table 2 were calculated from

the first motion size that reached the threshold criterion and, in essence,

represent lower bounds for the estimates of the thresholds for subjects GER

and JHK. A more stringent threshold would lead to an even larger estimated

magnitude of suppression.


100 S=JHK ...0--------0- --------0 Nonsaccade
S8 0 Saccade:
S70 motion
E 70
D .with
"o V"""""V
0 50
S4 0 against
| _J -------- ^ ----- I ----- \--- A7--A
100 S=CBW 0---------------
_ I

--------- ......

7 0 -

100- S=JBM ..0 ------ ---------

50 --------- ---
4 O0

0.51 2 4 6 8

Size of Motion (0)
Figure 7. Percentage of correct detections of stimulus motion for each

Table 2
Motion Detection Threshold During Saccade and
Nonsaccade Trials ()

Subject Nonsaccade Trials Saccade Trials:
Motion-Witha Motion-Againsta
GER 1.4 2.7(1.9) 4.0(2.9)

JHK 1.5 3.6(2.4) 8.0(5.2)

CBW 0.9 3.1(3.4) 2.9(3.1)

JBM 1.9 --b _.b

a Numbers in parentheses represent the ratio of the threshold during the saccade trials to
observed for nonsaccade trials for the same subject.
b This subject never reached to 75% correct detection level for even the largest extent of
stimulus motion (8).

As a summary, Table 2 shows the thresholds for motion detection

calculated as described above. One subject, JBM, never reached the 75%

correct detection level for motion in either direction during saccade trials,

which is why there is not a figure reported for threshold. Also reported in

Table 2 are the ratios of the thresholds obtained for saccade trials relative to

the thresholds obtained during nonsaccade trials. This ratio represents one

way of measuring the magnitude of suppression of motion detection.

Inspection of Table 2 shows that motion needed to have been approximately 3

times the extent (or three times the duration, see below) in order to be detected

during a saccade than when the eye is fixating. This ratio is very similar to

the 0.5 log unit elevation typically observed for saccadie suppression studies

measuring detection of luminance changes. Recall that this estimate of

suppression may be conservative, especially for JHK and GER. It is also

noteworthy that all of the thresholds in the present experiment indicate that a

motion larger than the 20 eye movement is necessary to be perceived. Also,

for GER and JHK the saccade trial thresholds for motion-with (stimulus motion

to the right) seem to be much smaller than thresholds for motion-against.


The results of the present experiment clearly demonstrate that the

ability to detect motion is reduced or suppressed during saccadic eye

movements. The amount of suppression, if measured using the 75% threshold

detection criterion, is about the same 0.5 log unit suppression typically found

in saccadic suppression studies measuring thresholds for detection of changes

in luminance. Importantly, at least under the stimulus conditions employed in

the present experiment, a motion the size of the eye movement in either

direction was not reliably detectable. Based on this result, it appears that

suppression of motion detection could be sufficient to prevent the subject in

Stevens' et al. (1976) study from perceiving movement during attempts to

execute saccades.

One of the more unexpected results was the difference in the

psychometric functions for motion-with versus motion-against saccade trials.

The reasons for differences in the psychometric functions for motion-with

versus motion-against trials are not immediately obvious. One possible reason

is that the stimulation of the retina is not identical for motion-with versus

motion-against. Each direction of motion differentially adds on top of the

motion caused by the motion of the eye. As such, motion-against, being in the

same direction as the motion caused by the eye, is additive, while motion-with

is subtractive being motion in the opposite direction. Thus, motion-with

reduces or even changes the direction of motion on the retina while it occurs.

To explore what possible effects these vector sums might have on the present

data, the vector sums were calculated with the simplifying assumption that the

eye's velocity is constant during a saccade. A typical 20 saccade takes 30 msec

(Bahill et al., 1975), giving a velocity of 670/sec. Since these are vector sums,

the signs are important. For these calculations motion to the right will be

given a positive sign. Thus, the eye movement creates a motion vector of

-670/sec on the retina. Recalling that the stimulus velocity is 770/sec, the

resulting retinal velocities are:

Motion-with (stimulus moves to the right)

1) 770/sec 670/sec = 7030/sec,

Motion-against (stimulus moves to the left)

2) -7700/sec 67/sec = 837/sec.

The change in retinal velocity is less than 10% of the presented stimulus

velocity of 7700/sec. Still, these changes in velocity might play a role in the

differences in the detectability of motion-with versus motion-against stimulus


Several issues not explored by the present experiment should lead to

caution in interpreting the present results. The first issue is the confound

between motion extent and duration that exists in this and other studies of

motions suppression (Bridgeman et al., 1975; Stark, 1976). In all of these cases,

the velocity of the motion was kept the same for all sizes of motion, but this

fact meant that as the motion size increased, so did the duration of the motion.

Thus, the thresholds obtained might not indicate the size of the motion

necessary for detection, but the duration of the motion necessary for

detection. While in either case the results indicate suppression of the ability

to detect motion, the interpretation of the results could be much different. If

the duration of motion is the only relevant factor in the results from the

present experiment, then motion much shorter than the duration of the eye

movement is detectable (-10 mscc for an 80 motion as compared to -30 msec for

a 20 saccade, Bridgeman ct al., 1975). Mack (1970) chose to confound extent and

velocity in his study of motion suppression. In all cases, the duration of the

motion was the same as the duration of the saccade, only the extent, and thus

the velocity, was changed. Still, his results are very comparable to those of

Bridgeman et al. (1975) and Stark et al. (1976). From this lone comparison

study, it appears that it is motion extent and not duration or perhaps even

velocity that is the more critical variable for motion suppression.

Still, the velocity of the motion may be a variable that can affect motion

suppression. To obtain sufficiently large motions in the present experiment

extremely high velocities (7700/scc) were used. Bridgeman et al. and Stark et

al. used an even higher velocity (9000/sec). Even in Mack's study, the

velocities used were much higher than typical velocities perceived during

normal experience. Yet, whether only high velocity information is

suppressed or all motion is suppressed, it is high velocity stimulation that

occurs as a result of a large saccadic eye movement. Suppressing only high

velocity motion would still effectively suppress the perception of motion of the

world caused by executing a saccade.

Another unresolved factor is the time course of the observed motion

suppression. The timing of motions during trials was selected to occur near

the beginning of the saccade. This linme period was chosen because the

maximum suppression of detection thresholds tenJcd to be observed near the

onset in other studies of visual suppression (Volkimann, 1986). If the time

course of what might be termed motions suppit.sr.ion is similar to the time

course of the suppression of detection of luminance changes, then this

observation might explain why on some trials in Experiment 1, motion was

observed to occur for the probe stimulus. If the probe was presented either

early enough or late enough, the suppression might not have been great

enough to prevent it from being observed. The observed motion tended to be

in the direction of the eye movement. Thus, the likelihood is that the stimulus

had been presented prior the saccade since the adjustment of direction

appears to begin before the eye movement (see Chapter II).

One other factor of potential importance that could affect the ability to

detect motion which has not been investigated in the present study is the

contrast of the grating. Specifically, it seems reasonable that higher contrast

gratings would be easier to detect. While this does not affect the comparison

between saccade and nonsaccade trials, the absolute thresholds should be

taken as specific to the contrasts used and not general. Still, as mentioned in

the opening chapter of the present paper, the high velocity of the eye motion

leads to a smearing of the retinal image which reduces the effective contrast.

So, perhaps, the low contrast gratings used here might effectively mimic real

world saccadic motion stimulation.

Unfortunately, one stimulus condition could not implemented on the

present equipment and, as a result, a determination of the source of

suppression is made ambiguous. Ideally, a second nonsaccade condition would

have been employed to some degree clarify the interpretation of the

mechanisms responsible for the observed suppression. In this condition,

during both parts of the trial, the grating would move in a manner consistent

with the motion on the retina generated by the saccade, simulating saccade

trials retinal effects. During one of these motions, one of the above used

motions would be superimposed and the subject would report during which

part of the trial the superimposed motion occurred. Thus, the retinal motions

would be identical for this trial type and for the saccade trials. The reason that

this condition could not be implemented on the present equipment was that a

video monitor was used to display the stimuli. Because typical 20 saccades last

longer than video frames (30 msec versus 16.67 msec) the blank interval

between frames would disrupt the presentation of the simulated saccade

motion, even with the cylinder lens.

Still, a discussion of the logic behind this trial type will help in

understanding the possible mechanisms of motion suppression during

saccades, and the small uncertainty which remains after the current study.

With this trial type, if thresholds between the nonsaccade trials and the

simulated saccade trials were identical, then any observed suppression could

be said to have arisen from an extraretinal source, since the critical factor

leading to the elevated threshold was executing the eye movement and not the

retinal stimulation. If, however, the saccade trial thresholds and the simulated

saccade trial thresholds were the same, then the suppression would be likely to

arise from a retinal source as changing the retinal stimulation, and not

making the saccade, lead to the elevated thresholds. As a result, the

suppression observed for motion detection in the present experiment (or any

other experiments to date) cannot be ascribed to either retinal or extraretinal

sources. Some tentative conclusions about the mechanisms of motion

suppression are offered based on other sources of information. The

suppression observed for changes in luminance suggests that it is likely both

retinal and cxtraretinal sources are involved. In addition, if the present

suppression is the mechanism that accounts for the subject in Stevens ct al.

(1976) only observing displacement and not motion of the world when

attempting to execute a saccade with paralyzed eyes, then these results argue

for at least some role for an extraretinal mechanism. The fact that

displacement and not motion were observed with the retrobulbar block, which

prevents feedback from the extraocular muscles as well, argues that a

corollary discharge must be involved to some extent.

In summary, the magnitude of motion suppression in the present

experiment seems adequate to explain why motion of the world is not observed

during either normal saccades or during carefully controlled paralyzed eye

experiments. The present results suggest that there may be some type of

asymmetry in the characteristics of suppression associated with the direction

of the motion during the eye movement. Also, the mechanisms involved in the

suppression have not been clarified, although some role for a corollary

discharge seems likely.

In the first two experiments, psychophysical methods have been

employed. These methods require verbal responses. The third experiment will

examine the issue of stability around the time of saccadic eye movements from

the perspective of a different response system. The final experiment will use a

motor response, specifically body sway.


The vast majority of research concerning saccadic effects on vision,

including the two previous studies reported here, have used psychophysical

techniques and required verbal responses. Verbal responses seem a natural

choice to ascertain visual functioning because verbal reports appear to be so

closely tied to our conscious perceptions. More importantly, these

techniques have proven valuable in uncovering clinically useful

information (e.g., as measures of acuity, Riggs, 1965), and in uncovering

functional aspects of sensory physiology (e.g., color mixture data and the

different classes of cones and color opponent cells, Boynton, 1979). Still, a

growing body of literature, that falls generally under the heading of "the

two visual systems" theory, suggests that other types of responses use visual

information differently (Leibowitz & Post, 1980).

The "Two-Visual-Systems" Theory

The two-visual-systems theory proposes that there exists a functional,

if not anatomical, division of "the visual system" into two different modes of

processing (Held, 1968, Leibowitz & Post, 1980; Schneider, 1967, 1969;

Trevarthen, 1968). In essence, the theory proposes that the visual system is

divided into two subsystems. One subsystem, referred to as the focal

(Trevarthen, 1968) or cognitive (Bridgeman, Kirch, & Sperling, 1981)

subsystem is proposed to be the system primarily tapped by psychophysical

techniques. The focal subsystem is proposed to be responsible for pattern

recognition and other tasks requiring fine discrimination. Information

stimulating the retina on or near the fovea contributes most heavily to focal

subsystem processing (Johnson, Leibowitz, Milladot, & Lamont, 1976;

Leibowitz, Rocdemer, & Dichgans, 1979; Trevarthen, 1968). Foveal stimulation

is of primary importance in focal vision because of the high acuity of the

fovea allowing fine discrimination and the poor acuity of the periphery not

allowing fine discrimination.

In contrast, the other subsystem, the ambient (Trevarthen, 1968),

localization (Leibowitz & Post, 1980; Leibowitz et al., 1979), or motor-oriented

subsystem, does not require the high acuity of the fovea for its processing.

Thus, visual information from all portions of the retina contributes to

ambient visual processing about equally (Johnson et al., 1976; Leibowitz et al.,

1979; Trevarthen, 1968). The ambient subsystem processes information

concerning the location of objects in space and the orientation of the

organism with respect to the environment (Held, 1968; Schneider, 1967; 1969;

Trevarthen 1968).

Anatomically, the retinal projections to the lateral geniculate nucleus

(LGN) of the thalamus, and from the LGN to the striate cortex (that is, the

central visual system) subserves the focal subsystem. The projections from

the retina to the superior colliculus of the midbrain are proposed to subserve

the ambient subsystem (Schneider, 1967, 1969). Placing a lesion in the visual

cortex of the golden hamster causes the animal to be unable to learn a

pattern discrimination task, while its ability to perform orientation tasks

appear unimpaired. Lesioning the superior colliculus causes the hamster to

be unable to learn orientation tasks. If an orientation response is not

involved, then the hamster can learn pattern discrimination (Schneider,

1967; 1969). A similar dissociation has been observed in monkeys with the

forebrain commissures (anterior telencephalic commissure and corpus

callosum) cut (Trevarthen, 1968).

Evidence supporting the two-visual-systems theory has also been

gathered from human patients with brain damage to various areas of the

visual system. Some of this evidence has fallen under the term "blind sight."

In this case, if the lesion affects the central visual pathways then the patient

reports being unable to see ("blind") in the affected portion of the visual

field. This report is confirmed by normal perimetry, i.e., the affected area of

the visual field is a scotoma. Examinations of the patient using stimulus

motion or a motor response have revealed that, if the superior colliculus is

spared, visual function is not completely absent from the scotoma. Visual

functions still present include 1) significant tendencies to execute saccades

in the direction of objects imaged in the scotoma (Poppcl, Held, & Frost, 1973;

Weiskrantz et al., 1974), and 2) the ability to accurately point to an object

imaged in the scotoma given that the object is sufficiently large (Perenin &

Jeannerod, 1975; Wciskrantz et al., 1974). These spared visual functions are

consistent with the functional and anatomical distinctions proposed by the

two-visual-systems theory.

Functional separations of the operation of the visual system consistent

with the two-visual-systems theory have also been observed in normal adult

humans. The visual acuity of the periphery does not seem to be affected by

changes in the refractive error of the image over a large range (Johnson et

al., 1976). This fact is in dramatic contrast with the effects of refractive

error on foveal vision where errors of less than 1 diopter (inverse of the

focal distance in meters) are considered clinically relevant and requiring

treatment. Apparently, the normal functioning of the periphery does not

require a clear image, as does the fovea (Johnson et al., 1976). A similar

dissociation between stimulus parameters affecting foveal vision and an

orientation response has been observed with circular vection (CV)

(Leibowitz et al., 1979). CV is the perception of circular self-motion by a

person when the person is actually still. CV is induced by a stimulus pattern

moving in a circular direction around the person. The direction of self-

motion is opposite to the direction of the stimulus pattern, thus the direction

of CV is consistent with the actual direction a person would move to create

the motion of the stimulus pattern if it were still. CV is only minimally

affected by the presence of refractive errors of over 16 diopters or by

reductions of luminance to near threshold levels. Both the large refractive

errors and the low luminances used Leibowitz et al. (1979) would greatly

disturb the perceptions of fine detailed patterns. Given that the perception

of fine details requires the fovea, and CV can be thought of as an illusion of

the subject's orientation relative to the environment, these results are

consistent with the distinction between focal and ambient visual processing

in the two-visual-systems theory.

Thus, several lines of evidence, which have been briefly reviewed,

support the two-visual-systems theory. One implication of this theory is that

psychophysical studies of vision primarily tap focal vision and not ambient

vision (Bridgeman et al., 1981). By implication then, psychophysical studies

of saccadic effects on vision may not completely describe how visual

functions change during saccadic eye movements. The next section will

review previous studies supporting this proposition.

Saccadic Effect on the Ambient Subsystem:
Previous Studies

Only two studies have examined what might be termed saccadic

suppression of ambient vision. In both cases, the studies have used postural

stability as the response (Krantz & White, in preparation; White, Leibowitz, &

Post, 1980). The rationale for using postural stability as a response for the

ambient subsystem rests on the observations that 1) vision contributes to

postural stability (e.g., Dichgans & Brandt, 1978), 2) the primary information

necessary for postural stability is information about the organisms

orientation relative to the environment (White, et al., 1980), and 3) the

retinal stimulation associated with a saccade would indicate that the person

was losing balance if it occurred when a saccade was not occurring (Krantz &

White, in preparation; White et al., 1980). In other words, informal

observation indicates that, during saccades, some sort of suppression of an

ambient visual function may be occurring as well as suppression of visual

functions required in psychophysical tasks. Both experiments support this

informal observation.

Both experiments mentioned above used similar designs to measure

saccadic suppression of body sway. The primary comparison was between

trials where subject made saccades across a still stimulus surround and trials

where the stimulus surround moved in a saccade like manner while the

subject fixated. The retinal stimulation was very similar in both conditions.

The only difference being whether the motion was made by a saccade or

motion in the environment. Both studies found that the conditions in which

the saccade-like image motions were produced by the stimulus surround

caused more sway than trials with actual saccadcs (Krantz & White, in

preparation; White et al., 1980). In fact, Krantz and White observed that

saccades may have even led to less sway than the baseline condition during

which both the stimulus surround and the subject's eyes were still. White et

al. did not observe this trend.

These studies also suggest ways in which saccadic suppression of body

sway is different from saccadic suppression observed in typical

psychophysical studies. Recall that Brooks and Fuchs (1975) found that

motion of the visual scene can elevate thresholds to about the same degree as

saccadic eye movements when the eyes are kept still. This observation is

important to metacontrast explanations of saccadic suppression (Matin, 1974).

The condition used by Brooks and Fuchs (1974) when high velocity stimulus

motion was presented to the fixating eye parallels the simulation of the

image motion caused by saccades used by both Krantz and White (in

preparation) and White et al. (1980). In fact, by simulating the image motion

of a saccade, the stimulus parameter for eliciting metacontrast effects should

be very similar in the saccade and in simulated saccade conditions. Yet,

suppression of body sway is only observed in the actual saccade conditions.

Actually, all peripheral conditions are similar, except for any shearing force

on the retina, between the actual saccade conditions and the simulated

saccade conditions in the postural stability studies, suggesting that only an

extraretinal mechanism may be involved in saccadic suppression of body

sway (Krantz & White, in preparation; White et al., 1980). This conclusion is

distinct from that reached with psychophysical studies of saccadic

suppression, which suggests that both retinal and extraretinal mechanisms

are involved.

The distinction between saccadic suppression observed with

psychophysical studies, perhaps tapping the focal subsystem, and the studies

of saccadic suppression using postural stability was further strengthened by

the relationship between the magnitude of saccadic suppression and the size

of the eye movement. As mentioned in the Introduction (Chapter I),

psychophysical measures of saccadic suppression indicate that the

magnitude of suppression increases as the saccade size increases. Krantz and

White (in preparation), however, observed that the magnitude of saccadic

suppression of sway is actually greater for small saccades (<40) than for

larger saccades (40 and larger), which is opposite the trend found in

psychophysical studies. This trend further supports the notion that some

sort of extraretinal mechanism must be responsible for saccadic suppression

of body sway. Above, retinal shear was the only peripheral mechanism that

was not ruled out by the simulated saccades conditions to account for the

suppression of body sway. Yet, since accelerations increase with increasing

saccade size, the magnitude of any shearing force should increase with

larger saccades (Alpern, 1971; Bahill, Clark, & Stark, 1975; Carpenter, 1977).

It would be expected from this fact that increasing saccade size should lead to

more, not less, suppression as observed. In fact, this observation tends to

rule out retinal shear as a factor in any type of saccadic suppression since it

operates at the level of the retina before any divergence between focal and

ambient subsystems would be possible.

Even more directly relevant to the topic of the present paper is the

observation by White et al. (1980) that motion of the surround stimulus

during the time of a saccade does not seem to be as effective as the same

motion when the eye is still. Several weaknesses in this particular

experiment by White et al. (1980) suggest that this conclusion is tentative. In

their study, White et al., did not measure eye movements at all, but simply

made the stimulus motion occur approximately 200 msec after the subject was

cued to make a saccade by the appearance of the target stimulus. While the

200 msec delay represents a typical delay time for the execution of a saccade

(Alpern, 1971; Carpenter, 1977), there is no objective measurement to verify

that the stimulus motion occurred during the eye movement. And if Krantz

and White's (in preparation) observation that simply executing saccades

reduces body sway relative to baseline, then this factor could account for the

differences observed by White et al., as there were not any saccades in the

comparison condition. One of the motivations for the present experiment is

to replicate White et al.'s observation when triggering stimulus motion off

the saccadic eye movement.

While both Krantz and White (in preparation) and White et al. (1980)

suggest that visual information to the ambient subsystem is suppressed

during saccades, other studies indicate that information in the ambient

visual subsystem is preserved. Studies involved in the debate over whether

saccades are retinotopically or spatiotopically programmed provide one

source of evidence that at least some aspects of visual information are

preserved in the ambient subsystem. This debate centers on whether

saccades are executed towards the location of the image of the object on the

retina relative to the fovea (retinotopic programming), or if the brain tries

to execute the saccade in order to image the spatial location of the object on

the fovea (spatiotopic programming). In other words, are saccades executed

using spatiotopic or retinotopic coordinates as discussed in Chapter I. While

in most situations both types of programming will produce identical results,

it is possible to identify situations where this may not be the case. Hallett and

Lightstone (1976a,b) attempted to address this controversy by presenting

saccadic targets during a saccade. Their hypothesis was that if saccades are

retinotopically programmed then the second saccade should not be accurate

but in error when referenced to the spatial location of the stimulus, as

indicated by psychophysical studies finding errors in perception of direction

(the first experiment can be included in this list). Yet, they observed that,

although a few saccade targets were "ignored" the saccades were accurate

and of "approximately normal latency" (Hallett & Lightstone, 1976b, p. 107).

But they did note that latencies to targets presented during saccades did have

slightly longer than normal latencies, such that the latency of the second

saccade as measured from the end of the first saccade was actually of a more

normal duration (Hallett & Lightstone, 1976a,b). Still, these findings indicate

a preservation of localization information in contrast to the findings of both

psychophysical studies and White et al. (1980). Again referring to Chapter 1,

these studies indicate a preservation of spatiotopic information despite

changes in retinotopic information.

Hansen and Skavenski (1985) found a similar preservation of

localization information with yet a different motor response. In their case,

they measured accuracy of a hammer strike at the location of a flashed target

presented before, during, or after a saccade. They found that the accuracy of

the hammer blows varied little around the time of a saccade occurrence. This

result appears to agree well with the results of Hallett and Lightstone (1976a,

b). Actually, the small variations in the position of the hammer blow

accuracy near the saccade onset may not be terribly inconsistent with the

results of Experiment 1 which found only small localization errors until after

the saccade. If the conjecture that failure of memory of the position of the

three initial targets is correct in explaining the errors of localization after

the end of the saccade in Experiment 1, then the results may be even more

similar to Hansen and Skavenski's (1985) which does not seem to have the

same memory demand.

Thus, some results suggest a suppression of localization responses

during saccades (Krantz & White, in preparation; White et al., 1980) and

others suggest a preservation of possibly similar responses (Hallett &

Lightstone 1976a,b; Hansen & Skavenski, 1985). But these studies may address

different aspects of visual functioning of saccadic effect on ambient vision,

in a manner parallel to the division between the types of visual information

in Experiments 1 and 2 of the present paper. While information about the

spatial location of a stimulus pattern is essential to perceive motion, White et

al.'s study of the effects of motion on body sway during saccades is formally

similar to the present Experiment 2. On the other hand, the studies of Hallett

and Lightstone (1976a,b) and Hansen and Skavenski (1985) are more formally

similar to the present Experiment 1 which does not incorporate any motion

in the stimulus. Moreover, in the latter studies, the temporal information,

when present, indicates that possibly the response uses information about

location obtained after the end of the eye movement. Hallett and Lightstone

directly observed an increase in saccade latency, while Hansen and

Skavenski have not measured latency at all leaving open the possibility that

the movement is programmed from information after the eye movement is


The present study is an attempt to better clarify the above

discrepancy. The study will measure body sway as the response as in Krantz

and White (in preparation) and White et al. (1980). The stimulus conditions

will be similar to Experiment 2. Some trials will present motions of a grating

stimulus during a saccade, and during other trials the same stimulus motions

will be presented while subjects fixate. In addition, the sway responses that

follow either a saccade or a stimulus motion will be averaged as in Krantz and

White. Thus any time locked component, like those observed by Krantz and

White can be detected and measured. This analysis provides some ability to

look at issues of localization of the stimulus. Krantz and White found

response average movements of subjects to depend upon the direction of

stimulus motion for lateral sway. The present study will be able to test

whether sway responses to stimulus motions depend on the direction of the

motion during a saccade.

The basic design of the experiment includes a baseline trial in which

the subjects maintain fixation and no stimulus motion occurs. This trial type

serves as a way of reducing intersubject variability by making every subject

his/her own control. Other conditions are trials with subjects executing

saccades without any stimulus motions, stimulus motions during fixations,

and trials of subjects executing saccades which trigger the motion of the

stimulus. In this experiment, the eye movement was chosen to be 30 since

the spatial frequency of the stimulus was .33 cycles/0. Also stimulus motions

were either 30 or 60. Thus, as in Experiment 2, the stimulus patterns will be

identical before and after the end of the eye movement and/or stimulus

motion so only motion and not displacement information should be able to

lead to sway. Moreover, motions the size of the eye movement are interesting

since they match the extent of motion caused by the eye movement. Motions

the extent of the eye movement were subthreshold in Experiment 2 during

saccades but not during fixations, so the choice of 30 motions was made. The

6 motions was chosen because motions twice the size of the eye movement in

Experiment 2 tended to be close to threshold even when the motions occurred

during saccades.



Eighteen subjects (12 female and 6 male) participated in the present

experiment as volunteers. The ages of the subjects ranged from 22 to 40

years of age. None of the subjects reported any previous history of postural

problems, unusual dizziness, or diseases that affect any sensory organ

associated with balance. Any prescription for corrective lenses was worn by

subjects during the conducting of this experiment.


The apparatus to amplify and generate trigger pulses from the EOG was

the same as the one used in both Experiments 1 and 2.

The apparatus to measure body sway and generate the stimuli utilized a

position detection/stimulus presentation system that has been described in

detail elsewhere (Krantz, 1985; and especially Shuman, in press). Only a

brief description of the apparatus will be presented here.

The entire system is referred to as the position sensor system (PSS)

which may be broken into four subsystems: data acquisition subsystem (DAS),

data processing subsystem (DPS), stimulus control subsystem (SCS), and

stimulus presentation subsystem (SPS) (Shuman, in press). The localization

of head position in space is accomplished by the DAS online, while the DPS

stores the data onto disks for offline analysis. The DAS performs the

localization of the head with two acoustic click sources, attached to a light

weight head gear which the subject wears, and 4 microphone detectors

positioned above the subject's head. A microcomputer (Challenger CIP) sends

a signal to each click source in alternation to emit a brief sound pulse and

simultaneously starts four timers, one associated with each microphone.

When a microphone detects the click, it stops its associated timer. These

timers have then recorded the propagation time of the click from the

acoustic source to the microphone, and this information allows the

determination of the position of the subject's head in space (Krantz, 1985).

Although more degrees of freedom can be resolved, presently only lateral or

left/right, and front/back translation movements and rotation of the head

about the mid-saggittal axis relative to a plane parallel to the floor are

resolved by the DPS.

The SCS consists of another Challenger CIP microcomputer connected

to a galvanometer driven SPS through an A/D converter. The SPS displays a

square-wave grating of 0.33 cycles/0 subtending over 1800 horizontally and

1150 vertically in the form of a half cylinder in which the subject stands

(Shuman, in press). The contrast of the grating was 46%. Fixation points

were also controlled by the SCS and were rear projected on the same stimulus

screen by a 12" black and white monitor with a visual angle of 160. The DAS

controlled the operation of the SCS, and also the presentation of all stimulus

events, by signalling the beginning of a trial (Krantz, 1985). To present

motion of the grating during saccade and record the samples during which a

saccade or grating motion occurred required modification to the PSS

configuration described by Shuman (in press) and Krantz (1985). These are

as follows: The trigger pulse generated by the onset of a saccade was fed into

the SCS microcomputer. On appropriate trials, this pulse signalled the SCS to

generate a stimulus motion as described below. To record the sample when a

saccade occurred (or on some trials when a grating motion occurred), the SCS

toggled a communication line to the DAS which stored the number of the

most recently collected sample. When data collection was finished, the DAS

sent the number of the samples during which saccades or grating motions

occurred to the DPS for storage and offline analysis as described below.


The EOG was recorded by electrodes placed bitemporally with a ground

electrode placed on the left mastoid as in Experiments 1 and 2.

Procedures that were identical across trial type were: 1) Every subject

filled out an informed consent which requested the subject to inform the

experimenter of any history of problems with balance, unusual dizziness or

problems with their inner ear. 2) The subject was positioned in the center of

the half cylinder formed by the screen. 3) An initial fixation light was

presented at the center of the screen for 500 msec at the beginning of the

trial. 4) Subjects stood on one foot, and the same foot was raised just prior to

each trial. When the subjects stabilized, the experimenter began each trial

and placed himself in a position to assist the subjects in case they lost their

balance. None did. 5) At the end of each trial, the subjects lowered the raised

foot so that fatigue was minimized. And 6) the subjects had a break after half

of the trials. There were four different trial types with different types of

stimulus/eye movement events. They were: baseline, saccades only, stimulus

motion prior to saccades, and stimulus motion during saccades.

During the baseline trials, the stimulus surround stayed still and the

subjects were instructed to keep their eyes positioned towards the location of

the initial fixation point. In all other trials, 30 saccades were executed as

signalled by the occurrence of fixation lights rear projected onto the screen.

Saccades were executed in pairs. The beginning of a saccade pair was

signalled by the presentation of a fixation light in the position of the initial

fixation light for 500 msec. About 50 msec after the initial fixation light was

extinguished, a second fixation target was presented 30 to the right of the

first target. The second target was on for about 150 msec, or shorter than the

typical delay between target onset and saccade onset. Then no fixation target

was presented for a random period of time of more than 1.2 seconds and not

more than 2 seconds to allow a sufficient period of time for the averaging of

sway responses (Krantz, 1985). Leftward saccades were then signalled by the

next presentation of the first fixation light. In all trials, other than baseline

trials, subjects executed 10 saccade pairs. The saccades to the right were the

eye movements relevant for the presentation of motion and data analysis.

In the saccade-only trials, no stimulus motion occurred. In the

motion-prior trials, the stimulus surround grating was moved either 30 or 60

during the interval between the extinguishing of the first fixation target

and the presentation of the second target. Both motions were approximately

ramp like and had a velocity of 2000/sec so the 30 motions lasted 15 msec and

60 motions lasted 30 msee. A higher velocity more comparable to that used in

Experiment 2 was preferred but was unobtainable because higher velocities

caused instabilities in the SCS that lasted much longer than the intended

motion. During a trial, only one motion size occurred, half (5) were to the

right and half were to the left. In the motion-during trials, the occurrence

of a stimulus-surround motion was triggered by the onset of a saccade to the

right following the extinction of the second fixation stimulus. Again,

depending on the trial, either 30 or 60 stimulus motions occurred. Typical 30

saccades last 30 msec (Bahill, et al., 1975).

Trials lasted 20 seconds with a sampling rate of 10 Hz. Only 10 stimulus

motions occurred, 30 stimulus motions represent .75% of a trial and 60

motions 1.5% of a trial. The twenty saccades occupied only 3% of the trial

time. Since both 30 and 60 motions were used, there were six specific types of

trials, with 4 replications of each trial type giving 24 total trials. The first

twelve trials, consisting of two trials of each type were randomized across

subjects, with the second 12 trials being the reverse of the first 12 for the

same subject. The EOG and trigger signal were monitored on every trial to

ensure that triggers were well aligned with the rightward saccades. Because

an individual false trigger could not be prevented from affecting the data

stream trials, the monitoring was for overall accuracy. On about 90% of all

individual saccades the trigger was aligned accurately with the saccade


Data Analysis

The samples of head position over time that are measured by the PSS

can be thought of as a wave form. This wave form can be submitted to a fast-

Fourier transform (FFT) which analyzes the wave form into the amplitude

and phase relationships of sine wave components that would make up this

wave form based on the assumption that the observed wave form represents

one cycle of an infinite wave form. In the present experiment, only the

powers, the square of the amplitude, were used for data analysis. The power

at a frequency is the square of the amplitude at that frequency. For purposes

of the present experiment, the resulting frequency components were

subsequently grouped into frequency bins to give greater stability to the

data. These frequency bins were then averaged within subjects across

replications of trial types. Comparisons across conditions could be made by

dividing the power at one frequency bin in one condition by the power at

the same frequency bin of another condition. The resulting ratio gives the

gain in sway for that frequency component of one condition relative to the

other. Taking the logarithm and multiplying by 10 converts the ratio to

decibels (dB's) where an increase in sway for the condition in the numerator

is given by a positive gain and a decrease is indicated by a negative gain.

Response averages could be calculated on trials with either saccades or

stimulus motions because the sample where a motion or eye movement

occurred was stored by the DPS. Offline the movement of the subjects were

averaged both within and across subjects for the 2 seconds following the

stimulus motion or eye movement. For the motion-prior trials, the averages

were taken relative to the onset of the stimulus motion. For the motion-

during, the averages were also taken from the onset of the stimulus motion,

which also coincided with the onset of the eye movement as signalled by the

EOG. In these two trials, averages were done separately for both left- and

rightward stimulus motions. The saccade-only trials were averaged relative

to the onset of the rightward saccades following the second fixation target.


Observations of the EOG indicated that the subjects had no problem

making eye movements according to the instructions. Baseline trials rarely

had any eye movements, and during the other trial types saccades occurred

as indicated by the fixation targets. Also important was the lack of any

evidence indicating that the stimulus motions lead to any type of eye

movement, reflexive or voluntary. At least, no eye movements reached the

level discernible on the EOG of approximately 0.5.

FFT Analysis

The typical power spectrum contains the majority of sway power in

the frequency components below I Hz. As gain measures are relative

measures, this general shape should be kept in mind. Also, it is relevant that

stimulus events occupied only a small proportion of any trial.

For all FFT analyses, the power spectra from individual subjects were

grouped into 0.5 Hz bins with center frequencies from 0.25 Hz to 4.75 Hz.

Then repetitions of the same trial type were averaged before the gains were

calculated. After these gains were calculated within subject, then the gains

were averaged across subjects.

In the first analyses, the gains of the five stimulus event conditions,

saccade-only, 30 and 60 motion-prior, and 30 and 60 motion-during, relative

to the baseline conditions of eyes still with no stimulus motion were

calculated. The average gains across subjects are presented in Figure 8a, b, c.

In all cases the gains were small as expected from the small amount of

stimulus motion used to improve response averaging. No strong spectral

signature was found for any condition, as is seen by the relatively flat and

smooth profiles. Still some important trends in the data appear.

As found by Krantz and White (in preparation), the saccade only

conditions uniquely tended to show negative gains across all three axes. This

trend toward negative gains was significant using the binomial test,

collapsing across all axes of sway measured (p(27130) = 3.7 x 10-6). The n of 30

comes from collapsing across the 3 axes with 10 frequency bins each. The

one difference in the stimulation between saccade-only and baseline trials

was that during baseline trials the fixation lights did not flash on briefly, but

remained off.


0 = / > --. .--A -
0.25 1.25 2.25 3.25 4.25
Frequency Bin (Hz)


-O- Prior 3 deg
.0- Prior 6 deg
**- Saccade Only

-~- During 3 deg
-- During 6 deg

1 +- ----l---'--- l-l- -l
0.5 N

-1.5 T,
-2 I I i I I I I I
0.25 1.25 2.25 3.25 4.25
Frequency Bin (Hz)


3. 0
2.5. / a..^.^u
1.5 / _
01 //o, \o

-0.5o i ,- -
. .


1.25 2.25 3.25
Frequency Bin (Hz)


Figure 8. FFT gains for the various stimulus conditions listed in the legend
relative to the baseline condition. a) Gains for the lateral sway axis. b) Gains
for the fore/aft sway axis. c) Gains for the head rotation axis.




_ 1 1

One of the most notable features evident from Figure 8b of the fore/aft

axis is the fact that the gain appears to be closely related to the size of the

stimulus-surround motion and not when the motion occurred relative to a

saccadic eye movement. The smallest gain (actually negative) is with the

saccade-only condition. Both 30 motion-prior and motion-during trials have

very similar gain spectra with gains larger than saccade-only gains. The 60

motion conditions have the largest gains, irrespective of whether the motion

occurs prior to or during the eye movement. This initial FFT analysis does

not give a clear picture of whether there was any sway suppression or not

because it does not directly compare the motion-prior and motion-during


To get a better measure of any suppression of body sway in response to

stimulus surround motions during saccades, gains were calculated for the 30

motion-prior trials over the 30 motion-during trials, and similarly for the 60

motion-prior over the 60 motion-during trials. These results are presented

in Figure 9. Figure 9a,b presents lateral and fore/aft sway respectively.

These sway axes do not show any evidence for sway suppression. All of the

gains are near 0 dB. The finding for fore/aft sway was clearly anticipated by

the gains relative to baseline shown in Figure 8b, where the size of the gain

depended more on motion size and not when that motion occurred. Yet, the

head rotation axis does show evidence of sway suppression for both 3 and 60

motion sizes. The gains tend to be positive, with the 60 gains tending to be

quite large given the small proportion of the trial during which motion


-0- 30 Motions

-0- 60 Motions


Gain 1.5
(dB) 0.5





Gain 1
(dB) 0.5 3 -m




1.25 2.25 3.25 4.25
Frequency Bin (Hz)

Gain *
(dB) 0.5

1.25 2.25 3.25
Frequency Bin (Hz)

Figure 9. FFT gains for motion-prior conditions relative to motion-during
conditions, a) Gains for the lateral sway axis. b) Gains for the forc/aft sway
axis. c) Gains for the head rotation axis.

1.25 2.25 3.25 4.25
Frequency Bin (Hz)

Response Average Analysis

In addition to the gain analysis, the data were submitted to an response

average analysis. This analysis was accomplished by averaging the sway

position of the subject relative to the position of the subject at the onset of a

same motion (in the same direction).

The averaging was done across repetitions of the trial type and subjects, as

well as within motion size (30 or 60), trial type (saccade-only, motion-prior,

and motion-during) and the same motion direction (left vs. right). Figures 10

and 11 display these response averages for left and right motions,

respectively. The saccade-only average responses are repeated on both left

and right motions even though averaging was only done for the rightward-

going saccades.

The rotation average movements are the clearest (Figures 10c and

llc). For both left and rightward stimulus motions, the 30 and 60 motion-

during trials show average movements that are very nearly identical to the

average movements for the saccade-only trials. There are slight but highly

reliable motions in the direction of the eye movement (the standard errors of

the peaks are in excess of 10 z units), this motion is similar to the motions

observed by Krantz and White (in preparation) for head rotations during

saccade only trials. The motion-during response average movements are

distinctly different from the response averages obtained for the motion-

prior trials. In this case, regardless of the direction of the stimulus motion,

the subject rotated slightly to the left first and then to the right. The

observation that the motion-during response averages are much more

similar to the saccade-only response average movements than to the motion-

prior response average movements suggests that the sway response "evoked"

1.50 AOooo00 o- Prior 3
1.00 o .a L rAi& -0- Prior 60
Relative 0.50 i. Sacade O
Position 0 *E- Saccade Only
(mm) 0.00 t ,.,'-,"* **** ^ ^*'
,(mm) 0.00AA ...... .-- During 30

-1.oo0 -'- During 60
0 500 1000 1500 2000 2500
Time After Event (msec)


1.00 mEu
Relative 0.50. A S . I'..
Position AA.AA A, -,--____
( -0.50 "gm* 0 33
-1.00 00003
0 500 1000 1500 2000 2500
Time After Event (msec)


0 .50, A.AAAAAA '-AA.A
0.40 "O i i1l0mi n
0.30 ] Do-
Relative a00 o
Position 0.10 (0
(0) 0.10-

-0.10 E3 -
0 500 1000 1500 2000 2500
Time After Event (msec)

Figure 10. Response average movements following the onset of either a
saccade (Saccade-Only) or rightward stimulus motion (all other conditions).
a) Response average movements for the lateral sway axis. b) Response
average movements for the fore/aft sway axis. c) Response average
movements for head rotation.

1.50 A -o- Prior 3*
1.00 A A -0- Prior 60
Relative 0.50 Saccad***-Ue
Position 0 =Iu*..AA.A.* E-.-.. "-Saccade Only
( 0.50mm) w '-4E-R'

-1.00 -A- During 6'
0 500 1000 1500 2000 2500
Time After Event (msec)

Relative 0.50 0 00-
Position M ',A'

-0.50 1 ===='
0 500 1000 1500 2000 2500
Time After Event (msec)


0.50- AA A.AABt
0.40 ,A N' .
0.30 ..u0
Relative 0.20 Mo,
Position 0.10 i .
(o) ooi0.00 ii ,
-0.100 b
0 500 1000 1500 2000 2500
Time After Event (msec)

Figure 11. Response average movements following the onset of either a
saccade (Saccade-Only) or leftward stimulus motion (all other conditions), a)
Response average movements for the lateral sway axis. b) Response average
movements for the fore/aft sway axis. c) Response average movements for
head rotation.

by the stimulus-surround motion was "suppressed" during the saccade for

both the 30 and 60 motion sizes.

It might also be noted that the response average movements for the

head rotations do not return to 0 or the starting location. It might be possible

that the head is simply making slow rotations to the right. Examination of

the sway averaged across the entire trial across all subjects reveals that this

is not so. The variability is quite large in the later periods of the response

averages. The fact that the response averages do not return to the starting

location is an artifact.

While the results are not as clear for the other sway axes, they are

generally similar. On the lateral axis, motion-prior trials show an initial

leftward movement. If the motion is to the right, the leftward movement is

followed by a rightward motion that is absent if the motion is to the left. The

pattern is rather unclear for fore/aft sway to rightward motions, but for

leftward motions the motion-prior trials show a trend towards moving

slightly towards the back of the subject that is absent during the motion-

during trials. The response average movements are more strongly

suggestive of suppression than are the gain analyses.


The present data present a somewhat contradictory picture. Looking

at the head rotation data, both from the FFT analysis and response average

movements, it appears that motion information is suppressed in the ambient

subsystem. A similar conclusion is tentatively reached from the response

average movements for the fore/aft and lateral axes. It is not possible from

the present data to know if suppression differed between 30 and 60 stimulus

surround motions. Yet, this conclusion is tempered by the observation, taken

primarily from the FFT lateral and fore/aft data, that sway depends on the

size of the stimulus surround motion and not when the motion occurred

relative to the execution of a saccade (Figures 8b, and 9a,b). Perhaps some

motion information is processed by the ambient subsystem, but not sufficient

information to lead to a time-locked response average movement.

Relevant to the main thrust of this investigation, the present

experiment suggests that suppression in the ambient subsystem for motion

may be somewhat similar to suppression of motion information in the focal

subsystem. In both experiments, motions the size of the eye movements were

suppressed. One of the questions about the generality of the second

experiment may have been somewhat answered by this third experiment.

The equipment used in Experiment 2 required low contrast (12% during the

motion) gratings to be used. In the present experiment, considerably higher

contrast was achieved (46%). Yet, the suppression observed in the present

experiment was still substantial, lending support to the contention that,

because of the smear of the retinal image during a saccade, contrast of the

grating may not be a very important factor.

Still, the suppression of motion information during saccades may not

be quite as complete for the ambient subsystem. Suppression was not

strongly indicated for either motions size for either the lateral or fore/aft

axis on the gain measures. It is interesting that the evidence for motion

suppression is clearly indicated by the head rotation axis. Perhaps why this

axis was so suggestive of suppression is related to the relative masses of the

head compared to the entire body. Given the small amount of stimulus motion

during a trial, perhaps there was not enough stimulus motion to cause the

entire body to move sufficiently to achieve a clear picture of suppression.

Despite this hypothesis, some motion information does seem to be processed

by the ambient subsystem during saccades as most clearly indicated by the

FFT data for the fore/aft sway axis (Figure 8b).

These results also show some interesting relationships to those of

Krantz and White (in preparation). First, the tendency for the saccade-only

trials to show less sway than the baseline trials is replicated. This might be

related in some fashion to the fact that not making saccades for more than

about a second (Bahill, Clark, & Stark, 1974) is an unnatural situation and the

concentration on this task could in some way lead to slightly more sway. The

tendency for saccades to lead to small head rotations in the same direction, as

seen in saccade only trials, is also replicated. One of the interesting

differences between the present results and the results of Krantz and White

(in preparation) is seen in the average movements in response motions

presented to the fixating eye. In the present experiment, there is not a

relationship between motion direction and the direction of the average

movement that was observed in Krantz and White. In the present

experiment, the motions always start to the left for stimulus motions in either

direction, for both lateral and rotation axes. In fact, the only indication of

directional sensitivity in body sway responses is that motions to the left cause

a longer lasting lateral leftward sway than motions to the right in the

motion-prior condition. In Krantz and White (in preparation) motion-only

trials were used without eye saccades during the experiment. Thus, the

response averages may be altered by the need to execute another saccade,

always to the right, shortly following the stimulus surround motion.

The fact that movement averages for fore/aft sway were not revealing

is also similar to what was found by Krantz and White. In both cases, motion

for fore/aft sway is signalled in both directions (fore and aft) for motions in

either direction (right or left). Thus, by and large, the present experiment,

where consistent, replicates Krantz and White (in preparation), extending

the notion that sway suppression includes motion during saccades and not

just still surrounds.

In summary, motion information in the ambient subsystem is

suppressed as it is for the focal subsystem. Some motion information does

seem to be preserved but it does not seem to be responded to in the same way

as equivalent motion during fixations. These basic findings seem at odds with

the findings by Hallett and Lightstone for saccades to stimuli flashed during

a prior saccade and by Hansen and Skavenski for abilities to strike objects

flashed during saccades. Yet, Hansen and Skavenski report small but

systematic errors to their targets during saccades. As a result, given the

small size of the responses in the present experiments, the two studies may

not be terribly inconsistent. Hallett and Lightstone do not give adequate

results to see if they could be interpreted in a consistent fashion. Still, some

motion information, including direction, is suppressed in the ambient

subsystem. As with Experiment 2, it is impossible to separate peripheral from

central mechanisms using the present results alone, but Krantz and White

(in preparation) found that only an extraretinal mechanism could account

for their results in the ambient subsystem. Since the same system is used in

both studies it seems logical to assume that extraretinal mechanisms are

responsible, primarily, for the results of the present experiment. The next

and final chapter will examine all of the present experiments in the context

of the question posed in the Introduction: Why does the world appear to

remain stable during saccades?



The three experiments reported here took a divergent attack on the

basic problem of why the world does not appear to move around the time of a

saccadic eye movement.

The first experiment explored directly what happens to the perceived

direction of an object around a saccade. The experiment was designed to

determine whether perceived direction shifted in a continuous manner

suggested by Matin's research (c.g 1972, 1976a,b) or whether the shift was

discontinuous as proposed by Hershberger (1987). The results clearly

supported a continuous shift of perceived direction, but the shift did not

appear at all like the trend observed by Matin. Matin observed that perceived

direction shifted slowly when it was inferred that only a corollary discharge

could cause the shift. Instead, the present study found a shift in perceived

direction that is rapid and only just precedes the eye movement. The

differences between the two patterns of results seem likely to be due to

memory factors. In Experiment 1, significant errors in perceived direction

were not observed in the data until after the eye movement was completed.

These errors are, also, likely to be due to memory factors. This interpretation

is strengthened because the judgements of direction were accurate and

consistent until after the change in eye position could act to perhaps disrupt

the remembered location of the reference stimuli. In other words, until

sufficient time and/or disruption by moving the eye occurred, judgements of

direction were quite accurate. In Matin's (1972, 1976) studies all judgements of

perceived direction were made at least a half a second after the offset of the

reference stimulus. This time period is much longer than the time between

reference and probe stimuli in Experiment 1.

The second experiment was designed to test if one reason the motion

caused by the shift in perceived direction found in Experiment 1 is not

normally observed was as a result of reduced sensitivity to motion during

saccadic eye movements. To get the best possible estimate of any motion

suppression, a grating was used as the moving stimulus. This grating was

moved in such a way so that the pattern seen by the eye before and after the

saccade was identical. In this study, suppression by a factor of about 3 was

noticed and thresholds for motion detection were larger than the 20 eye

movement used. It would be nice to interpret the results in terms of

suppression relative to the size of the eye movement, but the low contrast and

single size of eye movement used makes this interpretation untenable with the

present results.

The third experiment shifts response systems. This experiment

measured motion sensitivity during eye movements with body sway, a

response controlled by the ambient visual system (Trevarthen, 1968). The

pattern of results were complex but basically supportive of the interpretation

that motion during eye movements does not affect body sway as much as

motion during a fixation, and even when a response occurs, this response is

different from the response to the same motion during a fixating eye. These

results were similar to the results found for Experiment 2 using

psychophysical responses. This similarity occurred despite larger eye

movements and much higher contrast used in Experiment 3 than in

Experiment 2. Possibly Experiment 3 serves as an indirect extension of the

generality of the results of Experiment 2.

In light of these results, the next section will describe in general terms

how perceptual stability might be maintained. Next, the mechanisms that may

be involved will be discussed.

Perceptual Stability: Description

A good way to discuss perceptual stability is in terms of the theory of

Local Signs (Hering, 1879; Lotze, 1889). This theory attempts to account for

how the brain interprets the direction of an object. In essence, the theory

states that each part of the retina is associated with a particular direction. For

example, an object will be associated with a direction 10 above the direction of

gaze because its image falls on the retina a specific distance below the fovea

determined by the optics of the eye. Thus, any stimulation of that receptor and

associated neural architecture directly encodes some information about

direction. In terms of retinotopic versus spatiotopic coordinates, each

retinotopic coordinate is directly associated with a spatiotopic coordinate. The

problem with this simple idea is that before and after a saccade, the same place

in the world stimulates a different position on the retina, that is, the

relationship between a particular retinotopic coordinate and spatiotopic

coordinate changes as the eye moves. Yet, the brain somehow recalculates the

local signs in such a way that directions relative to the body remain the same

but the eye is given a new reference direction, perhaps associated with the

fovea. This recalculation or recalibration of the reference direction is so

accurate that no part of the perceived world appears to move. This

recalibration can be thought of as establishing the new and correct

relationship between retinotopic and spatiotopic coordinates.

The results of the present experiments suggest that this recalibration of

the local signs begins just prior to the eye movement and concludes before the

eye movement is finished. This interpretation explains the present results,

including the informal reports of motion in the direction of the eye

movement. It also explains Hershberger's observation that saccading towards

a low level flickering light causes a pattern of flashes to be imaged on the

retina towards the eye (an observation replicated by the current author). This

observation, present with every attempt and supporting the results of

Experiment 1, suggests that the recalibration must be finished prior to the end

of the saccade.

Yet, this recalibration usually goes unnoticed despite the potential to

cause apparent motion of the world. The research typically associated with the

term saccadic suppression suggests one reason the apparent motion is

unnoticed: visual sensitivity is reduced by about a factor of 3 (Matin, 1974;

Volkmann, 1976, 1986). Above this general reduction in visual sensitivity,

motion sensitivity is further reduced by about a factor of three and perhaps to

such an extent that motions of the size of the eye movement are not typically

noticed (Experiment 2). Suppression of motion does vary according to the size

of the eye movement as observed by Bridgeman, et al. (1975) and Mack (1970),

as does suppression of detection of an increment (Volkmann ct al., 1981). Thus,

suppression appears to be adjusted to match in some way the perceptual

disturbance that will accompany saccades of specific sizes.

Perceptual stability during saccades is not only important to our

conscious perceptions, but perhaps even more to visual control over postural

stability. It is also likely that the type of visual information is different for

ambient versus focal vision. Suppression in the ambient visual subsystem

(that portion of the visual system that provides information for postural

stability) has been demonstrated to exist in a growing body of studies

(Experiment 3; Krantz & White, in preparation; White, et al., 1980). The studies

indicate that saccades do not create as much sway as would be expected from

comparable visual motion. In fact, it seems as if making saccades leads to

slightly greater stability than artificially holding fixation (Experiment 3,

Krantz & White, in preparation). Moreover, the calibration of suppression for

the size of the saccade does not seem to be present. Larger saccades are not

associated with as much sway suppression as are smaller saccades. But, as in

the psychophysical domain, motion during an eye movement appears to be

suppressed. Motion during saccades both has less of an impact on sway, and

when it does appear to affect stability the effects are different from those

produced by the same motion to the fixating eye. First, there does not appear

to be an effect of the direction of the motion on sway during the saccading

eye. Second, even if the FFT spectra are similar for motions presented to the

saccading and the fixating eye, the response average movements appear to be

different (Experiment 3). Thus, suppression in support of perceptual stability

exists in both visual subsystems, but suppression seems to operate differently

in each.

University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs