Psychology 1XX3 Notes – Form Perception II – Mar 12, 2010
The brain uses a division of labour; with each region along the visual pathway
processing relatively specific information and then passing it on.
Form recognition follows a similar strategy.
Magno and Parvo Cells:
First, magna and parvo cells in the retina transduce the light stimulus into a neural
Recall that magno cells are found mainly in the periphery of the retina, and are
used for detecting changes in brightness, as well as motion and depth.
Parvo cells, on the other hand, are found throughout the retina, and are important
for detecting colour, pattern, and form.
These ganglion cells, with their small receptive fields, are the crucial first step to
object recognition. From the retina, the axons of these cells exit the eye via the
optic nerve, travel to the LGN, and end up in the primary visual cortex in the
What is most striking is that cells here are very particular about what will make
them fire, and these cells are called feature detectors.
Hodgkin and Huxley:
In 1952, Hodgkin and Huxley recorded the electrical activity in an individual
neuron of the squid, and this paved the way for other researchers to use this
technology to see how individual neurons respond to specific stimuli.
Lettvin et al:
For instance, in 1959, Lettvin and colleagues discovered a neuron in the optic
nerve of a frog that responded only to moving black dots, and they called these
cells "bug detectors".
Hubel and Wiesel:
Hubel and Wiesel spent years extending this work in their studies of cells in the
visual cortex of cats and monkeys, eventually earning the Nobel Prize in 1981.
Beginning in 1962, Hubel and Wiesel began their exploration of the visual cortex
by trying to learn what type of stimuli the individual conical cells responded to.
They did this by putting microelectrodes in the cortex of a cat to record the
electrical activity of individual neurons as the cat was shown different types of
visual stimuli, such as flashes of lights.
The problem was that they weren't getting much response from the neurons, until
one day when they presented the cat with a slide that had a crack in it. When the
line that was projected from that crack moved across the cat's visual field, the
neuron started to fire like crazy! This was a light bulb moment for Hubel and Wiesel, who realized that neurons
must respond to stimuli that are more complex than diffuse flashes of light.
They began using lines of different orientations and thickness that moved in
different directions, and they found that each neuron is very specific about what
will make it fire the most. These cells fire maximally to stimuli of a certain shape,
size, position, and movement, and this defines the receptive field for that cell.
Def’n of Simple Cell: Responds maximally to a bar of a certain length and
orientation in a particular region of the retina.
For example, this simple cell responds the most to a horizontal bar but if that
same bar is moved outside that particular region and/or changes orientation, then
the cell will be inhibited, and actually fire less than baseline. (See image below.)
So the receptive field for a simple cell is organized in an opponent fashion,
making it sensitive to the location of the bar within the receptive field.
Def’n of Complex cell: Responds maximally to a bar of a certain length and
orientation, regardless of where the bar is located within the receptive field, but
unlike the simple cell, a complex cell does not care about where in its receptive
field the bar is located, and it will even continue to fire if the bar is moving within
the receptive field.
Some complex cells do care about the direction of this movement, such as the cell
in this figure that fires the most when the stimulus is oriented at a certain angle
and moving in a particular direction.
See image on next page. Hypercomplex Cell:
Def’n of Hypercomplex cell: Responds maximally to a bar of a particular
orientation that ends at specific points within the receptive field.
For example, this hypercomplex cell fires the most to a horizontal bar of light that
appears anywhere in the "on" region of the receptive field, but gives only a weak
response if the bar touches the "off" region. So these cells have an inhibitory region at the end of the bar making them
sensitive to the length of the bar.
These three types of cells should give you some idea about how specifically tuned
our visual cortical cells can be.
The layout of the visual scene is preserved in the visual cortex.
Neighbouring objects in your visual field are processed by neighbouring areas of
your brain, but this mapping from visual field to brain is not exact, because the
largest amount of cortex is devoted to processing information from the central
part of the visual field, which projects onto the fovea.
Nevertheless, each region of the cortex receives some input from a small piece of
the visual field, and within each region, there are cells that analyze specific
features of the scene.
For a particular part of the visual field, there are neurons that fire maximally if
there is something in the scene that has a line of a certain orientation, length, and
movement; other neurons respond maximally if there is something in that tiny
portion of the visual scene that is a specific colour; other neurons respond most
when there is a line that moves in a certain direction.
Cluster of cells in the region of the cortex right beside this region are doing the
same analysis for the neighbouring part of the visual scene.
An important benefit of this parallel processing strategy is speed. (See img below)
Combining Information in the Extrastriate Cortex:
The processing of visual input in the primary visual cortex involves specific cells
responding to relatively specific features from a small portion of the visual field.
But for the visual scene to make any sense, this information has to be combined to
form a meaningful whole. Subregions in Extrastriate:
This combination begins in the extrastriate cortex, also known as visual
association cortex, which surrounds the primary visual cortex.
The extrastriate cortex has multiple subregions that each receives a different type
of information from the primary visual cortex about the visual scene.
For example, one subregion of the extrastriate cortex will receive information
about the colours in the scene, another about any movement in the scene, and
another about different line orientations in the scene.
It is in the extrastriate cortex where the information begins to be segregated into
two streams according to the type of information that is processed.
One stream is the dorsal stream, also known as the "where" stream, which
processes where objects are located in the visual scene and how they are moving
within that scene.
The dorsal stream takes information from the primary visual cortex to the parietal
cortex, which processes spatial information.
The other stream is the ventral stream, also known as the "what" stream because
it processes information about what the object is, including form and colour.
The ventral stream takes information from the primary visual cortex and sends it
to the temporal cortex, where all the bits of feature information come together.
Columns in the Temporal Cortex:
The temporal cortex is arranged in vertical columns that are oriented
perpendicularly to the surface of the cortex.
Neurons in the temporal cortex respond to very specific stimuli that are much
more complex than the stimuli to which the neurons in the primary visual cortex
These stimuli include images like hands, faces, apples or chairs.
Within each cortical column, there are five layers of neurons, with each layer
responding to complex stimuli that come from the same category.
However, each layer responds to slightly different features within that category.
For example, one column of neurons might respond best to apples, and layer 1
might prefer red apples, layer 2 might fire maximally to green apples, layer 3 to
yellow apples, layer 4 to small apples, and layer 5 to big apples.
See image on next page. Although this coding at the neuronal level can seem quite specific, every object is
not coded by a specific neuron.
In fact, an object is represented by unique activity patterns across marry cells in
several different brain areas.
The individual cells are just one component of the overall representation, and
even these cells will respond to a range of stimuli.
Development of Pattern/Object/Face Recognition:
The infant visual development proceeds relatively quickly in the first few months
of life. But perception is quite different from simply being able to sense incoming
stimuli, and even if an infant has the necessary equipment to see an object that
does not automatically mean that they can perceive patterns, objects, and faces in
the same way that we do.
Are we born with the ability to detect patterns and objects, or do we have to learn
about edges and other signals of object boundaries?
Many researchers have used the preferential looking method to determine what
kinds of patterns Infants can perceive by measuring which of two patterns the
infants look at the most.
Infant Pattern Recognition:
It turns out that infants prefer to look at patterns more than plain stimuli. When
presented with different patterns, infants prefer the ones that have a lot of high
contrast with sharp boundaries between light and dark regions.
Infants will look the longest at the most complex stimuli that they are able to
perceive. Infants Prefer Patterns:
For example, if presented with a checkerboard pattern, a newborn will prefer to
look at the pattern with bigger squares; if the squares are too small, the newborn's
poor visual acuity will make it look lust like a uniform grey surface, which isn’t
By 2 months of age, when the infant's visual acuity has improved, they will now
prefer to look at a smaller-squared checkerboard because it is more complex.
Infant Object Recognition:
An infant’s initially poor visual acuity is obviously an issue for perceiving objects
as well as patterns.
In fact, some researchers believe that because infants under 2 months of age see
so poorly, and look at such a limited part of the object, they may be unable to
perceive whole forms at all.
For instance, if you show a young infant a triangle or a star and measure where
they look, they will tend to stare at one corner and not at the entire shape, whereas
infants over 2 months of age are beginnin