Why do we move our eyes?
- Image stabilization in the presence of
body movements.
- Information acquisition - bring objects of
interest onto high acuity region in fovea.
Retinal structure
Cone Photoreceptors are densely packed in the central fovea
Visual Acuity matches photoreceptor density
Oculomotor Muscles
Muscles innervated by oculomotor, trochlear, and abducens (cranial) nerves from the
oculomotor nuclei in the brainstem. Oculo-motor neurons: 100-600Hz vs spinal motor
Neurons: 50-100Hz
Types of Eye Movement
Information Gathering
Voluntary (attention)
Stabilizing
Reflexive
Saccades
vestibular ocular reflex (vor)
new location, high velocity (700 deg/sec),
ballistic(?)
body movements
Smooth pursuit
optokinetic nystagmus (okn)
object moves, velocity, slow(ish) – typically
whole field image motion
up to 35 deg/sec
Vergence
change point of fixation in depth
slow, disjunctive (eyes rotate in opposite directions)
(all others are conjunctive)
Note: link between accommodation and vergence
Fixation: period when eye is relatively stationary between saccades.
Acuity – babies
Acceleration
Depth-dept gain, Precision in natural vision
Velocity
https://www.youtube.com/watch?v=KSJksSA6Q-A
Latency of vestibular-ocular reflex=10msec
Demonstration of VOR and its precision – sitting vs standing
Miniature eye movements
Slow drift
Micro-saccades
tremor
It is almost impossible to hold the eyes still.
Step-ramp allows separation of pursuit
(slip) and saccade (displacement)
Saccade latency approx 200 msec, pursuit approx 100 – smaller when there is a context that
allows prediction.
“main sequence”: duration = c Amplitude + b
Min saccade duration approx 25 msec, max approx 200msec
Factors That Control Gaze.
- TASK
Defines behavioral goals, what information is relevant.
- REWARDS
Oculomotor circuitry sensitive to reward/subjective value
of those goals.
- UNCERTAINTY
REDUCTION
Get information. Peripheral resolution/ working memory
decay etc
- PRIORS/ Memory Gaze targeting reflects stored knowledge.
- IMAGE
Salient properties eg high contrast/ spatial outliers
Brain Circuitry for Saccades
1. Neural activity related
to saccade
2. Microstimulation
generates saccade
3. Lesions impair saccade
Brain Circuitry for Pursuit
Eye Tracking Methods
Developments in Eye Tracking
Difficulty: optical power of eye + observer movement
Head fixed /restricted:
Contact lenses: mirror / magnetic coils
Early infra-red systems
Dual Purkinje Image tracker
Head Free:
Head mounted IR video-based systems
Remote systems with head tracking
Scene camera
Why eye movements are hard to measure.
A small eye rotation translates into a big change in visual angle
Visual Angle
x
18mm
a
d
tan(a/2) = x/d
a = 2 tan 1 x/d
1 diopter = 1/focal length in meters
0.3mm = 1 deg visual angle
55 diopters = 1/.018
Measuring Eye Movements
Early Methods:
“Barlow photographed a droplet of mercury placed on
the limbus. Translations of the head were
minimized by having subjects lie on a stone slab
with their heads wedged tightly inside a rigid iron
frame”
Kowler, 1990
Early methods:
“The eye is first cocainized, then the lids should be
propped apart by some form of eye-lid fastener, of
which the best is probably that in form of a wideopening spring with tortoise-shell grooves for the
lids.”
Delabarre, 1898
Monitoring Eye Movements; Yarbus
Mirror mounted on eye using suction.
Light bounces off mirror and is recorded on film
Non image-based Eye Trackers
• Non image-based eye trackers
– Electrical/analog
– Limbus
– Magnetic search coil
EOG
The eye is a ‘dipole’
with ~millivolts
voltage difference
between the retina
and the cornea.
EOG
ElectroOculoGram (EOG)
Use in clinic – head not fixed
Limbus Trackers
By monitoring the ‘whites of the eye’ below
the iris, it is possible to determine eye
position.
Vertical eye movements cause
both signals to increase (up)
or decrease (down).
Horizontal eye movements cause differential
illumination between the right and left sensors.
Limbus Trackers
Limbus
EOG and Limbus trackers
Good temporal resolution.
Lousy spatial resolution
High noise, drift
Mostly useful in clinic
Magnetic Search Coils
Used for much animal work, though less so recently. Very high precision and
accuracy (few minutes of arc). Used in older human em literature.
Can use similar methodology for head and hand (see Hayhoe lab)
Skalar search coils
Image-based Eye Trackers
• Image-based eye trackers
– Dual Purkinje
– Video based
Dual Purkinje Trackers
The ‘gold standard’ in eye trackers
Multiple reflections from the cornea and lens vary in a very welldefined way as the eye moves. By tracking the 1st and 4th
reflections, the tracker can determine eye position with very high
precision.
Bill Geisler lab has a binocular tracker.
Dual Purkinje Trackers
Precision: < 1’ (~1/100 deg)
Accuracy: a few min arc
Update rate: > 500 Hz
Dual Purkinje Trackers
Usually requires bite bar but
theoretically can get away with
head rest.
Video-based Eye Trackers
• Video-based eye trackers:
– Head mounted
– Remote
Head mounted
Camera on head views scene, another camera views eye.
Video-based Eye Trackers
Infra-red video camera finds center of pupil and corneal reflection.
Advantages: unconstrained viewing.
Disadvantages: temporal resolution may be as low as 30 Hz
Accuracy never better than 0.5 deg.
RIT Wearable Eyetracker
RIT Wearable Eyetracker
Build-up neurons in the intermediate layers of the SC are active
prior to a saccade.
Cell in the superifical layers get input from the retina. This may mediate
Very fast saccades – sometimes called “express saccades”
Extent of buildup neuron activity reflects stimulus probability.
Express saccades might also reflect activity in buildup neurons.
Posterior Parietal Cortex
reaching
Intra-Parietal Sulcus: area
of multi-sensory convergence
grasping
LIP: Lateral Intra-parietal Area
Target selection for saccades: cells fire before saccade to attended object
Visual stability
Model of saccade generation: target selection depends on expected value
Area LIP contains a reward expectation signal which modulates the gain of visual
neurons in LIP.
Reward modulation of saccadic eye movements originates from dopaminergic input
to caudate nucleus.
Trommershauser, Glimcher, Gegenfurtner, 2009
Relation between saccades and attention.
Saccade is always preceded by an attentional shift
However, attention can be allocated covertly to the
peripheral retina without a saccade.
Pursuit movements also require attention.
Visual Stability
Figure 8.18 The comparator
A cross seen through an aperture that moves clockwise around the boundary.
Alternatively, the aperture may be stationary, and the cross move behind it.
Individual views, shown on the right, are ambiguous.
Observers have no trouble with this if they have an “internal model” or schema that
readily allows interpretation of the sequence.
Supplementary eye fields
-Saccades/Smooth
Pursuit
-Planning/ Error
Checking
-relates to behavioral
goals
A subset of SEF neurons and LFPs exhibited strong modulation following erroneous saccades to a
distractor. Altogether, these results suggest that SEF plays a limited role in controlling ongoing
visual search behavior, but may play a larger role in monitoring search performance.
Nearby Anterior Cingulate also involved in performance monitoring.
Pre-motor neurons
Trochlear
Motor neurons
V
Oculomotor
nucleus
Abducens
H
Motor neurons for the eye muscles are located in the oculomotor nucleus (III), trochlear nucleus (IV), and
abducens nucleus (VI), and reach the extraocular muscles via the corresponding nerves (n. III, n. IV, n. VI).
Premotor neurons for controlling eye movements are located in the paramedian pontine reticular formation
(PPRF), the mesencephalic reticular formation (MRF), rostral interstitial nucleus of the medial longitudinal
fasciculus (riMLF), the interstitial nucleus of Cajal (IC), the vestibular nuclei (VN), and the nucleus
prepositus hypoglossi (NPH).
Pulse-Step signal for a saccade
Brainstem circuits for saccades. Omnipause neurons (OPN) in the nucleus raphe interpositus (RIP) tonically
inhibit excitatory burst neurons (EBN) located in the paramedian pontine reticular formation (PPRF).
When OPNs pause, the EBNs emit a burst of spikes, which activate motor neurons (MN) in the abducens
nucleus (VI) innervating the lateral rectus muscle. The burst also activates interneurons (IN) which activate motor
neurons on the oculomotor nucleus (III) on the opposite side, innervating the medial rectus. Inhibitory burst
neurons (IBN) show a pattern of activity similar to EBNs, but provide inhibitory inputs to decrease activation in
the complementary circuits and antagonist muscles. Long-lead burst neurons (LLBN) show
activity long before movement onset, and provide an excitatory input to EBNs.
Brain areas involved in making a saccadic eye movement
Behavioral goal: make a sandwich (learn how to make sandwiches)
Frontal cortex.
Sub-goal: get peanut butter (secondary reward signal - dopamine - basal
ganglia)
Visual search for pb: requires memory for eg color of pb or location
(memory for visual properties - Inferotemporal cortex; activation of
color - V1, V4)
Visual search provides saccade goal. LIP - target selection, also FEF
Plan saccade - FEF, SEF
Coordinate with hands/head
Execute saccade/ control time of execution: basal ganglia (substantia
nigra pars reticulata, caudate)
Calculate velocity/position signal oculomotor nuclei
Cerebellum?
RF reticular formation
VN vestibular nucle
PN , pontine nucleii
Cerebellum
OV oculomotor vermis
VPF ventral paraflocculus
FN fastigial nucleus
otoliths
Rotational (semi-circular canals)
translational (otoliths)
Function of Different Areas
monitor/plan
movements
target selection
saccade decision
saccade command
inhibits SC
(where to go)
V
H
signals to muscles
(forces)
Brain Circuitry for Pursuit
& Supplementary
Smooth pursuit
Velocity signal
Early motion analysis
Eye Movement Research
Consequences of image motion on visual acuity: stabilized images
Metrics of saccades/ pursuit/ vergence/vor
Constancy of visual direction
Eye movements in reading/ Cognitive role of eye movements
Active Vision/Natural tasks: Fixation patterns, eye/head/hand
coordination
Language comprehension in visual context
Paradigm Differences
Natural Tasks:
Study small segments of behavior
Multiple visual operations: transitions between
operations
S in control of agenda
complexity of scene
Standard approach: Repeated observations of a small time slice
Single visual operation or movement
Limited complexity
Remote
Bright pupil
Dark pupil
Video-based Eye Trackers
Video-based Eye Trackers
Bright-pupil; coaxial illumination
Dark pupil
Bright pupil
Two signals; pupil and corneal reflection - fairly robust
to tracker movement wrt head.
Lower temporal and spatial resolution than eg coils/DPI
1 deg, 60-120 Hz
Embed eye image in video record to monitor quality of track.