1.

ANOTHER NEON-COLOUR SPREADING ILLUSION - reason is reduced spatial resolution of

the colour system and the fact that you fill colour information upto a luminous defined boundary.
It is a sensory illusion because it is based upon the limited resolution of the colour system
compared to the achromatic system.

2. How do cells attune to different stimulus dimensions and how they pull information
ADDITIVE COLOUR MIXING - our colour system works by having 3 broadband
photoreceptors so each one sensitive to a wide range of different wavelengths and our perceived
colour is based upon the ratio of the activity and the way we encode that is via those 3 opponent
mechanisms. This means that different combinations of wavelengths will be perceived identically
if they result in the same relative activity of those 3 cone types.

● We can account for these illusions based upon how the L, M & S cones encode wavelength
(colour) information
● So by knowing what physically distinct stimuli are perceived to be identical can tell us a lot about
how the cells attuned to different stimulus dimensions and how they pull information. (Metamers)

Perceptual illusions -
How we interpret the retinal image. It's not about how the information is encoded by the sensory
mechanism but it's about how we read out that sensory representation.
- We look for global patterns
- We try to combine things to look for objects in the background.
- Sensory representation is being actively interpreted by the brain

3. Is seeing an active or passive process?

Think about it in relation to the brain generating a percept of the outside world
● The aim of the brain is to make internal percept of the outside world
● We want that percept to be an accurate representation/model of the world

4. But in terms of creating that perception, what information does the brain have to go on? -
The only direct information is the retinal image

● What are the problems with the retinal image?

5. How would you perceive the retinal image? - Sensory representation happens when the
cells early in the visual system encode that information. So you get the representation in the brain.
What you then have to end up with - is a perception of the outside world.
The percept must correspond to that object in the outside world. So if the percept nearly corresponded to
the retinal image, you would perceive that retinal image.

Perceptual constancy - the brain actively interprets that retinal image to try to perceive the constant
objects out of the world, not the constantly changing retinal image.
The retinal image is the only direct information that the brain has to produce the percept.That is, the
actual information about the world that our brain has access to
• But, the retinal image is constantly changing
• When we move in relation to an object, the shape and size of its retinal image changes
• Consider the size of the retinal image of your hand as it moves closer to you

If our percept just corresponded to the retinal image we would see the world constantly changing
• We want our percept to correspond to the unchanging (constant) objects • Not the ever changing retinal
image • Means that the brain has to actively interpret the retinal image in order to produce a percept that
(hopefully) corresponds to the outside world • Process called Perceptual Constancy

The perceived physical attributes of an object typically do not change with changes in viewing conditions
• e.g. perception of shape, size & colour stay substantially constant
• If perceptual constancy did not exist, objects would appear to change as your view of them
changed
Means that the visual system must actively interpret the changing sensory retinal image in order to
construct a percept in which objects are constant
• Imagine how difficult & confusing it would be if your hands constantly changed size as the retinal
image size changed

• Sensation : Representation formed by our sense organs

• e.g. response of our retina to light and how those retinal signals are encoded, and hence transformed, by
brain cells
• Can think about it in terms of the ‘hardware’ of the visual system

• Perception : How we ‘interpret’ the sensory input, i.e., the neural representation in the brain
• In other words, what we actually see
• Can think about it in terms of the ‘software’ of the visual system

TWO TYPES OF VISUAL ILLUSIONS

• Sensory : Due to limitations in how the information is encoded • e.g. due to the tuning properties of
visual cells • Can think of as a hardware fault

• Perceptual : The sensory representation is fine

• Mistake made in how that representation is interpreted
• Many are due to perceptual constancy getting it wrong
• Can think of as a software ‘fault’
• But not represented at a ‘cognitive’ level
• Implemented via the connections between cells in the brain

WHAT CAN THE ILLUSIONS TELL US ABOUT HOW THE BRAIN WORKS
• Sensory illusions : What brain cells are tuned to and what information cortical areas process
• How the brain cells/areas interact with each other
• How information is transformed in going from one brain area to another
• Perceptual illusions : How the brain interprets the sensory representation
• What assumptions it makes about the outside world
• Note: these assumptions are encoded by brain cells, e.g. how they are wired together (patterns of
excitation and inhibition between them)
• Types of perceptual constancies • Size • Shape • Colour • Luminance

PERCEPTUAL CONSTANCIES
The perceived physical attributes of an object typically do not change with changes in viewing conditions
• e.g. perception of shape, size & colour stay substantially constant
• If perceptual constancy did not exist, objects would appear to change as your view of them changed
• e.g. a circle would be perceived as elliptical as viewing distance increased

Perception of shape, size, colour staying substantially the same - i.e it is happening in the outside world,
rather than what's happening at the retina, where these things are actually changing due to the differences
in viewing conditions.

→ Means that the visual system must interpret the (changing) sensory information (retinal image) in order
to construct a percept in which objects are constant
• Imagine how difficult & confusing it would be if your hands constantly changed size as the retinal
image size changed
• Reading material refers to this as an illusion
• Debatable, in terms of representing the retinal image its an illusion, but that isn’t the aim of the visual
system

Size constancy -
• Indicates that perceived size does not simply correspond to the retinal size of the image
• Perceived size depends upon: • The size of the retinal image
• (Estimate of) viewing distance
• We scale the size of the retinal image by the an estimate of viewing distance in order to generate our
percept of size
• Note, perceived size can also be a cue to distance
• Will return to this concept with the moon illusion

Visual angle is the angle that the object produces on our eye.

DEPENDENCY OF PERCEIVED SIZE ON PERCEIVED DISTANCE - an experiment

• Stimuli - Test physical disk of various sizes at various distances
• Always produced the same sized retinal image (they were arranged that way so that even at different
distances, they produced the same size visual angle.)
• so disk size increased with viewing distance
• Comparison disk at a fixed distance. So they got people to vary the size of the comparison disc to match
what they thought the perceived size of the test disc was.
• Task : Match the size of the comparison disk to that of the test disk

- As the viewing distance was increased, the disc got bigger, such that the visual angle of the disc
always stayed the same.

DISTANCE ESTIMATE
Subconscious inference • At a subconscious level, the visual system takes into account the distance to the
object, and scales the retinal image accordingly

1. How can you use an after image to show that an estimate of distance is used to scale the size of
the retinal image?
2. What pattern of results would support the idea that -
• Viewing distance is not used to scale perceived size?
• Viewing distance is used to scale perceived size?
3. How can we use a retinal afterimage to show that we use an estimate of viewing distance to
generate the percept of size?
4. Size illusions: Due to?
5. These two vertical bars are the same physical length but they are perceived to be different. Why?
6. People & cars look very small (almost like toys) when viewed from a very tall building/mountain
or a low-flying aircraft. Why?
7. Do we learn size constancy (e.g. interaction between vision and touch) or is it innate? • How
would you test?

Size illusion - misestimation of distance.

Relative size • Perceived size also depends upon the size of surrounding objects • e.g.
Ebbinghaus/Titchner illusion • The orange dots are the same size

SHAPE/OBJECT CONSTANCY
• The retinal shape varies with viewing distance, however perceived shape doesn’t change (view
invariance) • Can view as a special case of size constancy
• Parts of the object further away create smaller images than those closer • Foreshortening: distances
along the line of sight get shorter with increasing viewing distance more rapidly than orthogonal distances
• frontal plane projection

REGRESSION TO THE REAL - shape constancy basically

Tendency to perceive the actual shape of the object, rather than the retinal shape
→ instead of perceiving the distorted retinal image, what you perceive is the interpretation of that retinal
image. In other words, the object out in space.
• Perceive a circle on a table at a distance as a circle, not an ellipse
• Percept is powerful, makes it hard to ‘see’ the retinal image
→ due to perceptual constancy we don't perceive that distortion, we see the object as it really is, rather
than the distorted retinal shape.
• Have to learn what the retinal shape is - Children’s drawings

CONSTRUCTIVISM - Notion that the the aim of visual processing is to build an internal model (percept)
of the outside world
• However, thought that the retinal image is intrinsically ambiguous
• Therefore visual system has to make a series of assumptions in interpreting the retinal image when
forming a percept
One of the aims of Ames was to determine what these assumptions are and their hierarchy, i.e. which
assumptions are more important than others
• Consider two possible assumptions: • Rooms are rectangular • People stay the same size as they move •
Which one do you think would be more important ?

• Ames room • Room is trapezoidal, but arranged so that under monocular viewing conditions, it produces
the same retinal image as a rectangular room
• If a person moves from A to B, the increase in the retinal image they produce is greater than that
generated by a rectangular room. When people move around, their distance from the observer changes.
And therefore, the retinal size changes.
• Given viewing conditions, no change would be expected
• Two possible interpretations: • Room is rectangular, person changed size • Room isn’t rectangular,
person didn’t change size

• Seems curious that the assumption of rectangular rooms overrides that of constant sized people
• What is the reason for this assumption ?
• Innate or learnt assumption ? • How would you test which is true ?
• Why does the rectangular assumption override the one that people don’t change size as they move?

SUBCONSCIOUS INFERENCE
The inferences made in interpreting the sensory input to create the percept occur at a subconscious level
• We are not aware of the assumptions, and the hierarchy of those assumptions • e.g. Ames Room, the
assumption (constraint) that rooms are rectangular overrides the assumption that people do not change
size
• Can be encoded via patterns of excitation and inhibition between cells
• Will look at that with the Marr-Poggio model to solve the correspondence problem in stereo-depth
perception

THE MOON ILLUSION

• The full moon just off the horizon appears to be much larger than when it is high up in the sky
• The size of the retinal image doesn’t change as it moves up in the sky
It is an illusion because if you look at the size of the retinal image of the moon as it goes up in the sky, it
does not change size. So there is no distortion due to the atmosphere that affects the size of the retinal
image, nor does it go further away as it goes up such that the size of the retinal image gets smaller.
The size of the moon stays the same.

DEPENDENCY OF PERCEIVED SIZE ON PERCEIVED DISTANCE -

• Experiment by Holway & Boring (1941)
• Experiment to determine whether we use distance to scale the retinal image to produce a percept of size
- This has all to do with perceptual constancy.
- Why does the visual system need to do perceptual constancy? If we didn’t do perceptual
constancies, we would directly perceive - in terms of generating a percept size, the effect on
distance in terms of direct information, the visual system has to go with in relation to size to the
retinal image.
- One way for the brain to percept the outside world - light reflects off them or they generate their
own light, they cast an image onto our eyes, you have sensory encoding of that information,
you’ve got cones, ganglion cells, etc - they encode that information, you’ve got sensory
representation of the retinal image. At the end of this process, you create a perception of the
outside world.

If our percept was a direct correlation of the retinal image, what would be the drawback with that? In
relation to the perceived size of the object, what happens to the retinal image as you move in relation to
the object?
- If you put your hand in front of you, and then you have the distance to that. So if you draw it
closer to the size of the retinal image, the surface area has increased by a factor of 4. Its like
having 4 hands there, but you do not see that change in perceived size.

The rationale behind perceptual constancy -

Due to differences in viewing distance and angle, our retinal image is constantly changing. So if we
directly perceive the retinal image, the size, the shape, the colour, when we go from artificial light to
outdoor light, the light coming off that object changes quite markedly because of the different
wavelengths making up the luminance by the light hitting it. So if we directly perceived the retinal image,
our perception of the outside world would be constantly changing. People would be perceived to be
bigger when they moved towards us and get smaller when they walked away.
→ we want to perceive the constant object out in the world, not the continuously changing retinal image.

• Stimuli - Test disk of various sizes at various distances

- Always produced the same sized retinal image
- so disk size increased with viewing distance
- Comparison disk at a fixed distance
→ as the distance to the disk got greater, the actual physical size of the disc increased and what it means
is that the size of the retinal image, as defined by the visual angle of that object (the angle it produces as it
hits the eye) stays the same.
• Task - for the different sized disks at different distances, they had to pick a disc that they thought was the
same size as that disc.

So what would happen if perceived size did not depend upon viewing distance? (it depends purely upon
the size of the retinal image)

- We are assuming here that you are not using distance to scale the retinal image. So the only thing
you can go with is the retinal image. The retinal image does not change so the perceived size
should be the same. But if you are using distance, you have to scale that distance. So scale that
retinal image with distance and you will perceive it as being bigger.

Main finding from the experiment -

As the disc got further and further away, the participants saw that disc as being bigger. So even though it
was creating the same size retinal image, because it was further away, the visual system was scaling the
size of the retinal image by an estimate of distance. So if the visual system estimates the distance to be
further away, for that same size retinal image, it means that the disc must be bigger.

If viewing distance is used to scale perceived size -

• How could you reduce the ability of the brain to do this?
- Remove the cues to distance. If people are scaling the size of the retinal image by an estimate of
distance, if you are removing cues to distance, the visual system cannot accurately estimate
distance. So you’re only left with the size of the retinal image. So now the perceptual size
constancy is lost and you perceive directly the retinal image size.

• What would be the result of this, i.e. what would perceived size be based upon?
- The perceived size would be bigger for that same size retinal image. As the disc gets further and
further away, its got to get bigger and bigger in terms of generating a percept of those discs.

ANOTHER WAY TO SHOW THE USE OF DISTANCE -

• How can we use a retinal afterimage to show that we use an estimate of viewing distance to generate the
percept of size?
- For afterimages, you look at a stimulus for an extended period of time, the cells that encode that
information become adapted, so that their response decreases, so you get an afterimage. To use
this to show that we use distance to scale perceived size -
- If you looked at the dot for an extended period of time, then you looked to the left or right, you
would perceive the same size dot. If you then look at a blank page, what will be the perceived size
 of that after image? In terms of perceptual size constancy, we use an estimate of distance to scale
the retinal image because frame of reference does come into play when we talk about the moon
illusion - one potential explanation for that. But it's wrong as it's using size comparisons on the
horizon.
- When you look at the page, you will perceive it smaller because the estimate of distance now is
smaller. This is because the brain is assuming that the after image is a real object as it does not
realise that it's been generated by an after image. So the estimate of that same size retinal image is
a lot smaller. It is being scaled by a smaller amount than if you look at a greater distance.→
EMMERT’S LAW →

• The size of the retinal afterimage is fixed • However, the perceived size of the afterimage depends upon
the distance to the test field/screen • If the test distance is closer than the adaptation distance the perceived
size of the afterimage will be smaller than that of the adaptation stimulus • If its longer, it will be large
- You want your percept to correspond to the actual object in the world, not the retinal image. So
you scale accordingly.

ILLUSION OF SIZE - • These two vertical bars are the same physical length but they are perceived to be
different. • Why?
- From a size constancy perspective?
- If you’re thinking about depth, which one would be closer?
- The reason for it is perspective distortion.
- MULLER-LYER ILLUSION - Use of size constancy perspective cues to account for the illusion
- The arrow heads on top of the lines gives the impression of perspective distortion.
- Based upon perspective information, the one that is further away is scaled up accordingly.

Why do we need to impose size constancy? - because we want the percept to correspond to the object out
in the world, not to the constantly changing retinal image. The size of that image constantly changes, and
gets bigger. If our perspective directly corresponded to the retinal image, things would be constantly
changing size, shape, luminance, brightness, colour etc.

FAILURE OF SIZE CONSTANCY • People & cars look very small (almost like toys) when viewed from
a very tall building/mountain or a low-flying aircraft • Why?
- You can’t really estimate the distance/ have cues to distance when you’re that far away.
- You also don’t have perspective cues, in an open area like that, you don’t have those monocular
cues.
→ • Limited cues to distance, so underestimate it • Results in a too-low scaling of retinal-image size •
Perceived size based more on the actual retinal-image size • Recall the study by Holway & Boring, 1941

IS SIZE CONSTANCY INNATE?

Do we learn size constancy (e.g. interaction between vision and touch) or is it innate ? • How would you
test ?
• Use a habituation study - Infants preferentially look at a new stimulus
• Habituate the infants to a specific stimulus size and distance combination - Either a large cube far away
or small cube closer
- Cubes identical except for size and distance
- The combination of size and distance gives the same visual angle
• Two test stimuli - One the same size as the habituation stimulus • The other, a different size • However,
distances such that their retinal images sizes were the same
→ Present two identical images to the child, apart from distances in size at different distances such as the
size of the retinal image being produced is the same. One is big, one is small.
- If the baby has size constancy, will they perceive those two stimuli as being the same or as being
different? - Different.
- Even though the size of the retinal images are identical, given how you scale, distance and size, if
they have perceptual size constancy, and they can estimate distance, they will perceive them as
being different.

• What would indicate size constancy? - Infants preferentially look at the new cube (B)
- If the child sees the two objects as identical they will look at each one about the same. They look
at each object about 50% of the time. if they see a new object, they will preferentially look at the
new one. And if this happens, it means they have scaled that by that identical retinal image by
distance, and now they’re perceiving it as a new object. If they are not doing that, they think its
identical, you won’t get the preferential looking.

IS SIZE SCALING CORRECT?

• What about cases where perceived size and distance are in the opposite direction ? • i.e. perceive a large
size combined with the perception of a short distance
• Will consider in relation to the moon illusion

USE OF PERSPECTIVE CUES IN PAINTINGS -

Why did it take artists so long to use perspective information (retinal shape distortions) in painting (took
until the Renaissance)?
• Due to shape constancy
• We have a poor ability to directly see our retinal image • We see the image after it has been modified by
perceptual constancy mechanisms • First painters to use perspective used 2D projections, i.e. view
through a glass window or images on a mirror

- If you’re looking at a circular disc on a table, as that circle gets further and further away from
you, what is happening to the shape of that retinal image?
- This is foreshortening.
- Along your line of sight, distances get shorter and quicker than distances at right angles to your
line of sight. So if you’re looking at a circle, it becomes more and more elliptical. The major
access of that ellipse will be at right angles to your line of sight. It will get smooshed along your
line of sight.
→ it is a manifestation, a consequence of the things that lead into the need for size constancy i.e these
distances changing as a function of viewing distances.

- In the painting, they are trying to get depth information, what they do have is occlusion, i.e
objects in front of others so you can’t see the one behind. One of the big mistakes they’re making
in the painting is - all the figures are the same size. They are not thinking, lets have these
perspective distortions in the image and the ones far away will be scaled up.
- Our brain is wired in a way to not directly perceive the retinal image.
- To get a 2D projection of a 3D world, Allbert looked through a pane of glass with only one eye
and painted what he saw. This was one way to work out what the perspective distortions are by
having that projection onto a pane of glass.

→ before photographs, all you had to go on in terms of what those distortions are is your retinal image.
But we don’t have direct access to our retinal image. The whole point of perceptual consistency is we
don't perceive the retinal image, we perceive our interpretation of it.

ALBERTI’S WINDOW (1436)

• Method for drawing pictures in proper perspective • View a scene through a plane of glass with one eye
in a fixed position • Trace all of the contours directly onto the glass with a grease pencil • If the window is
placed over a uniform white surface & viewed monocularly from the same relative position, recreate the
original pictorial depth information • Leonardo da Vinci (often credited with the applying perspective
information to pictures) advocated similar ideas about 50 years later

Would the sense of depth produced by such a drawing match that of the original scene? - No, contains no
stereoscopic cues

Ames was a constructivist and his idea was that the retinal image is intrinsically ambiguous i.e can be
interpreted in multiple ways. So his focus was - what assumptions must the brain make in order to
disambiguate/ interpret the retinal image?
→ gibbs thought we don’t need to create an internal perception of the outside world, that we could get
information directly from the retinal image.
• Two possible interpretations: • Room is rectangular, person changed size and distances to people stays
the same • Room isn’t rectangular, person didn’t change size
→ with the size of the retinal image is changing assumption - the size of the retinal image can be
changing due to either the person is changing size, or distance is changing. This is what happens in an
Ames room. It appears rectangular but its actual shape is trapezoidal.
- You only get the illusion if you close one eye.

EXPLAIN THE MOON ILLUSION

• Explain why the moon illusion occurs using perceptual size constancy
- Full moon, just off the horizon looks humongous. When it's up in the sky, it looks a lot smaller.
 - If you take a photograph of the moon, at different angles of elevation, the size of the image stays
the same. So its not that the image is actually bigger when its just off the horizon due to going
through the atmosphere.
- The retinal image size of the moon stays identical but it looks humongous just off the horizon, a
lot smaller up in the sky.
- Using perceptual size constancy, what does the brain do? - distance
- In terms of size constancy, the 2 factors that come to play are the size of the retinal image and an
estimate of distance.
- There are 2 different objects at different distances. They both produce the same size retinal image.
So what differs to account for the moon illusion using perceptual size constancy is the estimate of
the distance to the moon when it's off the horizon vs up in the sky.
- The moon at different locations produced the same size retinal image. So generate the percept of
the object out in the world, it's got to be making different estimates of distance.
→ the brain must be estimating the distance to the moon when it's just off the horizon to be greater than
when it's up in the sky. That way you can scale that retinal image by a greater distance when its off the
horizon vs when its up in the sky.

You have an infinite number of different size objects that can be at different distances. if they all produce
the same size retinal image, what it means is they get bigger with increasing distance. So therefore, to
create a perception of what you think the size of the object is, you want to correspond to the object out in
the world, not just on the retinal image. Therefore, the greater the estimate of distance, the bigger the
percept of size that you want.

INTERACTIONS BETWEEN CONSCIOUS (C) & SUBCONSCIOUS (SC) INFERENCES

- The illusions occur at subconscious levels
- These SC assumptions about distance lead to a percept of size and then at C level, you come up
with a conscious estimate that is the exact opposite of what the brain has actually done to create
the perception.
• Conscious reasoning may not have access to SC inferences & may actually draw the opposite conclusion
• The moon low in the sky is assumed to be further away than when its elevated, therefore seen larger
(Ptolemy, 200AD)
However, the moon looks further away when its high in the sky than when its on the horizon, therefore
above logic is wrong (Boring, 1940)
• No. C estimate of distance based upon perceived size, which is based upon SC inference of distance
(Kaufman & Rock, 1962)
• Decreasing distance cues decreases the magnitude of the moon illusion
• Conscious awareness of errors made in the subconscious inference process does not alter the percept •
The Ames Room: even when you know that the room isn’t rectangular, still get the illusion • These
assumptions can be wired into the visual system via the pattern of excitatory & inhibitory links between
neurons • Marr-Poggio neural model that accounts for how we perceive stereoscopic depth

What will be one way to minimise the difference in perceived size between the moon just off the horizon
and up in the sky? So impair the ability of the brain to estimate distance. And what you can do is to invert
a photograph. So this is conscious awareness.

BAYESIAN APPROACHES TO VISION

- Fundamental assumptions about how to interpret the scene.
• Can use different cues to extract the same information, e.g. texture, disparity & size information to
determine the slant of a surface • These different cues can give different estimates • The visual system can
apply different weights to these cues and these weights can depend upon the expectations (assumptions )
the visual system has about them (priors)

OTHER THEORIES TO ACCOUNT FOR THE MOON ILLUSION -

• Angle of regard - The moon high in the sky looks smaller because you have to raise your eyes to look at
it so maybe the tilt of head plays a role in generating the perceptive size.
Test? • Size comparison to objects on the horizon at 2 different viewing conditions

Test by changing relative head & eye positions, both when standing & lying down & using mirrors to
offset the visual direction of the moon • Eye/head position don’t determine the existence of the illusion
- Size comparison to objects on the horizon - Still get the illusion on open plains and the sea

WEEK 2

Types of thresholds -
• Absolute - Detecting the presence of a stimulus
• Relative - Telling two (supra-threshold) stimuli apart

ABSOLUTE THRESHOLDS
• Minimum stimulus intensity at which the stimulus is ‘just’ detectable
• For example, how bright a dot of light has to be in order to see it.
Detectability of visual stimuli is explained in terms of the difference between the luminance of the dot
and the background, in other words the contrast. This is what the cells in the visual system are detecting -
the contrast level/ the luminance difference not the absolute luminance.

PSYCHOMETRIC CURVE
→ plotting performance, i.e percent of the time they can see it vs stimulus intensity.
• Plot of detectability vs stimulus intensity
→ for example if you’re asking a person - can you detect the dot of light and you’re varying the stimulus
intensity.
Basically, this is a plot of your ability to see the dot of light - so the percent of the time/ percent of the
trials that you can see the dot of light as a function of stimulus intensity.
A really low stimulus intensity to the left of the curve. The intensity of the dot of light is too low so we
can’t see it - 0% of the time. As you increase the stimulus intensity, you go from not being able to see it to
seeing it 100% of the time.
Slope/ shape of the curve is influenced by the noise/ variation (in terms of things outside of the
environment)

External noise - external of the participant or the person doing the experiment. For example - variation
in the stimulus. • e.g. variation in the light source over time
By noise we mean variability that affects the shape of the psychometric curve.

Internal noise - variation in the observer e.g. neuronal noise & attentional variation.
For example the cells in the visual system even if you showed the identical stimulus intensity to them
over trials, the response they would give would be variable. - this is neural noise

Noise can affect measured threshold values in two ways (can affect both external and internal)/ Different
ways in which noise can vary -
• Random Noise - The direction of the error/ variation varies randomly from trial to trial • e.g. the actual
light level is measured too high/low
• Systematic Noise - The direction of the error/ variation is constant. The offset is always in one direction
• e.g. the light level is always measured high

For both external and internal, the noise can either be systematic or random.

• What are the practical outcomes of • Random noise? • Systematic noise?

• That is, what do they do to the psychometric curve?
Random noise - Actual threshold (mean) largely unaffected but increased variability (variance) of the
threshold
• Increased standard deviation reduces the likelihood of getting significant differences between conditions
• Need to increase the number of times the threshold is measured in order to reduce the variance of the
data
→ increasing random noise would make the curve more shallow. The more random noise you have, the
greater the variation, the more shallow.

Systematic Noise - Measured threshold value will be changed and wrong • Experiment will give the
wrong results
→ you get an offset on the psychometric curve. Your threshold may be wrong and it will be worse if
systematic noise only affects one of your experimental conditions, not the other. Unless you can work out
the error, i.e if it is a measurement error, then you can correct it.

• Out of the three possibilities, which would be the optimal threshold? • A, B or C? • Why?
 - It is talking about your ability to just detect the stimulus but in terms of running an experiment,
choosing a threshold value, what is an optical one?

• Think about it in terms of: what are the desired properties of the threshold?
• There are two desired properties • For one of them, think about what you would want if you ran the
same experiment several times - you want the same threshold value
- If your experimental manipulation, if your new therapy actually doesn’t have an effect, what are
you essentially doing? You are just running the same experiment time and time again.
- In other words, if you get differences in threshold measured between two conditions, you want to
be able to say that the difference is due to experimental manipulation.
- In terms of selecting a threshold value, there are 2 criteria that are really important to determine a
threshold value. - reliability and validity.
- Validity - it's got to indicate a stimulus intensity, a performance level that indicates that at least to
an extent you can detect the stimulus. At B, you are not seeing it everytime. So if you presented
the stimulus intensity 100 times, only 50 times the person can see it.
- C can be seen 100% of the time. So it is valid.
- But why is B and C not reliable? You can have a range of stimulus intensity that would give you
the same performance level. If you re run the experiment, you show them a higher stimulus
intensity, you would measure performance at 100%, that's your threshold.
- Point b is the steepest, you will have better ability to get that same threshold value if your
experimental manipulation is having no effect.
- The steeper the curve, the more effect then deviating from that stimulus intensity will have on
performance.
- Validity is - the stimulus intensity is too low, you can’t see it, its not valid.
- Reliability - the idea that in experiments you are comparing thresholds across different
conditions. If you are getting different threshold values across those different conditions you want
to be able to say the difference is due to what I have done to those conditions/ my exp
manipulation or the individual difference factor. It doesn’t have to be an experimental design.

• If you were running an experiment, what point on the psychometric curve would give you the
most sensitive measure? - The greatest ability to detect an effect of the experimental manipulation?

• Consider the example of conducting an experiment to investigate the effect of working-memory

(WM) load on the ability to perceive particular stimulus at low contrast • WM load may either
improve and impair contrast sensitivity
The question you want to investigate going in is what would be the best point on the psychometric curve
to run before your threshold value. And then you have to compare that across different working memory
load.
• What is the best performance level to use in the experiment?
• In terms of the sensitivity of an experiment, what are the issues with floor and ceiling effects? - In
other words, to detect or measure the impact of an experimental manipulation, what are the effects of
floor and ceiling effects and what part of the psychometric curve gives you the greatest sensitivity.
In other words - if you have 4 different experimental conditions over which you vary the working
memory load on the person, what performance level do you want them to be performing at that control
condition.
If working memory load has an impact on contrast sensitivity, it would be equivalent to changing the
contrast of the stimulus.

Everything talked about thus far is subjective technique - you are just asking the person if they can see the
stimulus or not. Then you get the plot of percent seen vs stimulus contrast.

PROBLEMS WITH SUBJECTIVE TECHNIQUES

• Subjective criterion - How confident the person needs to be in order for them to be willing to state that
they see the stimulus

• Get individual variation in subjective criterion -

• Some people may say they can see it only when they can easily see it
• Others may be willing to say they can see it with just have a slight impression of doing so
• Would lead to variations in the measured threshold

• For the same person, the criterion can vary from test to test - Leads to high variability, i.e. reduces
reliability // this can affect the measured threshold level.
• Validity - Don’t know if they can actually see it when they say they can
• Raises a question on the validity of the threshold

DIFFERENCES IN THRESHOLDS
You are in an experiment and you’ve got differences in thresholds between two groups. You want to
compare the ability of two groups to perceive a motion stimulus
• Highly-anxious group
• Control group
• You run the experiment, and the Highly-anxious group has higher thresholds than the Control group
• Do the results necessarily mean that they have worse motion ability than the control group?

OBJECTIVE METHODS: FORCED CHOICE METHODS

- There are problems with just asking the person if they can see the stimulus. So this has led them
to objective or forced choice methods. Therefore -
• No longer ask the person if the stimulus is there or not
• Tell them it is there and ask them something about it

• Give the participant a number of choices & force them to choose

• e.g. 2 Alternative Forced Choice (2AFC), is the stimulus moving left or right? - you have given them 2
choices.
• This removes subjective criterion
• They have to choose one of the alternatives even if they think they are guessing

CHECKING RESPONSE VALIDITY

• Unlike a subjective response (can you see it, yes/no), know what a correct response is with the forced
choice method therefore can determine if the observer can actually see the stimulus

• What if the person wants to deceive you into thinking they have worse ability than they actually do? •
Can you determine if someone is lying?
• Imagine that you are nefarious, and you want to give the impression that your vision is worse than it
actually is • For example, you are claiming to have had your vision impaired in a work injury

PSYCHOPHYSICAL TECHNIQUES
• Determining thresholds
• How would you determine the threshold value, e.g. how bright a dot of light must be in order to see it ?

3 we will be looking at -
• Method of Adjustment/Limits
• Constant Stimuli
• Staircase Method

METHOD OF ADJUSTMENT/LIMITS
- It is such if the participant is changing the stimulus intensity
• Observer/Experimenter adjusts the stimulus intensity until it is just visible • e.g. increase dot brightness
until observer can just see it
• Threshold is the value at which the response changes, e.g. from can’t to can see it
• Need to run at least twice (ascending & descending) - i.e increasing stimulus intensity from threshold to
below threshold or decreasing.
• Otherwise get a systematic error

NEED TO VARY INTENSITY IN BOTH DIRECTIONS

• Habituation - Stimulus is gradually changed in the direction of the threshold over a number of stimulus
trials • Can be a tendency for the participant to develop the habit of repeating the same response, e.g. will
continue to make the same response even after the threshold has been reached • Different to how the term
habituation is typically used

• Expectation - Opposite to habituation

Increase stimulus intensity so the participant expects to see it
• The participant falsely anticipates the threshold stimulus value, so reports a change in percept before it
has actually occurred
• Advantages - Its quick • Needs no (limited) pre-testing
• Disadvantages - Not particularly precise

METHOD OF CONSTANT STIMULI

• Present a fixed set (typically 5 to 9) of stimulus intensities a fixed number of times in a random order
• Plot the frequency of responses as a function of stimulus intensity - Results in full psychometric curve
• Must be sure that the expected threshold value lies near the midpoint of the range of stimulus values
used • Need to run a pilot test (need to know both the estimate of the threshold but also an estimate of the
slope of the psychometric curve.

What you are doing is selecting a range of stimulus intensities. You present them a fixed number of times.
In a random order. You plot the percent seen for each of those stimulus intensities so its a function of
stimulus intensity. - you get a psychometric curve.

• Advantages - Produces full psychometric curve - the slope of the curve is an indication of noise, part
of that is internal noise.
• Can be useful
• Sometimes different groups differ more on the slope of the psychometric curve than the actual threshold
value - either in terms of direct neural noise or attentional variation more so than the actual threshold
value.
• Random order of presentation avoids a number of problems with the Method of Limits (habituation and
anticipation)

• Disadvantages - it produces a full psychometric curve

- Inefficient • If you don’t want the full psychometric curve • Many of the trials are way
above/below threshold value • Need to pilot test

STAIRCASE METHOD
- An adaptive method: the stimulus intensity shown depends on the responses of the participants.
- Its an extension of the method of limits
• or modification of the Method of Limits
• Start at a stimulus intensity way above/below threshold value
• How you change the stimulus intensity depends upon the observer’s response
• Positive response: Decrease intensity (they can see it)
• Negative response: Increase intensity • Called a 1 Down/ 1 Up staircase

• Reversal Point: • Where the direction of change of stimulus intensity reverses

• Stop when a specified number of reversals is reached

→ the point where you go from decreasing to now increasing stimulus intensity is your first reversal
point. The direction of change reverses. Now you go up stimulus intensity, and the person can see it - this
is your second reversal point.
So each up and down is your reversal point. To stop the staircase you need to specify beforehand how
many reversal points need to happen before you stop it. Once you hit the number - you stop it. Then your
threshold is the average of the last 2/ last 4/ whatever you wish.

• Advantages - Quick • Accurate

• Disadvantages - Spend most of the time around threshold • Some observers find that difficult • Very
sensitive to false errors • Not good to use on a group likely to give a high false-error rate, e.g. children,
some clinical groups

DIFFERENTIAL THRESHOLDS
- You have 2 stimuli. You perceive both of them. How different do they need to be to determine
which one is more intense.
• Differential Thresholds refer to the ability to discriminate between two suprathreshold stimuli
• e.g. which of two lights are brighter ?
→ Leads to the notion of a Just Noticeable Difference (JND)

WEBER’S LAW
• Just Noticeable Difference (JND) - Difference in intensities between two stimuli required to tell them
apart aka Discrimination threshold

• Imagine discriminating 2 weights • If could tell 10.3 kg is higher from 10 kg

→ your JND for this is 0.3 kg
• What weight could you discriminate from 20 kg ? • 20.3 kg, 20.6 kg, 20.9 kg or ….? • i.e. is the JND
constant, regardless of the reference stimulus magnitude ?

→ JND is not a fixed amount - it varies as a proportion to what you are comparing it against.

When you are looking at an image left eye right eye. Binocular disparity is an offset in the relative
location of those images. If you could overlay the two eyes/ retina, zero disparity would mean the images
would overlap. Non zero = there's an offset. We get offset bc of horizontal displacement of the eyes. Each
eye gets a diff view of the outside world. So relative location of left eye right eye can be different.