stimulus recognition and localisation Flashcards
give two weird animal examples of recognition systems
Echolocation - bats use this to detect the smooth surface of water (to drink from), and to detect their prey
Electroreception - some fish have an electric organ, and create an electric field around themselves
They have electric sensors on their body that can detect changes in the electric field due to presence on objects in the field
what are two properties of visual recognition?
Orientation invariance - an object is still recognisable in different orientations
Scale invariance - an object is still recognised as the same thing at different sizes
how is object recognition hierarchical?
The further along the visual pathway/ventral stream, the neurons detect more complex aspects of an object
detection of edges and contours - object parts - POV - particular person a car - categorisation etc…
what increases along the ventral stream?
The receptive field size of neurons also increases along the ventral stream
what was seen when there were lesions in the inferior temporal cortex?
Lesions in inferior temporal cortex decrease ability to recognise objects -
removal/damage of this area causes mice to not recognise objects that they could recognise before, and they cannot learn to recognise new objects
information from the retina goes to?
describe the structure of it (input of the different layers)
lateral geniculate nucleus
contains 6 layers
the four outer layers receive input from parvocellular cells as these are responsible for fine spatial details and object recognition more so than magnocellular
the inner two layers receive input from magnocellular
contralateral (receives input from eye on the other side) vs ipsilateral
from outer to inner, it goes C, IP, C, IP, IP C
the LGN is deep in the brain so is hard to test on and image its activity…
what do we know about how it works?
It acts as a ‘thalamic’ relay station
The ganglion cell axons make 1:1 connections with the neurons here in the LGN - no high computation going on or integration of signals
Receptive fields of the LGN neurons are similar to receptive fields of the retinal ganglion cells - suggests the LGN doesn’t do much
in the LGN, the ganglion cell axons make 1:1 connections with the neurons, indicating low levels of computation.
what two things contradicts this?
Presence of local interneurons suggests there is some computation going on, we just don’t know what it is yet
60% of synaptic input at the LGN comes from the cortex (feedback), not the retina - the brain is putting a lot of effort in to regulate the activity of these neurons
where do signals go after RGCs - primary visual cortex (V1) - V2?
splits into two
dorsal pathway (where pathway)
to V3 – medial temporal cortex – parietal cortex
ventral pathway (what pathway)
to V4 to inferior temporal cortex
what are the dorsal and ventral pathways responsible for?
dorsal = ‘Where’ pathway - responsible for processing info on the object’s position, speed, direction of movement
ventral =
‘what ‘ pathway - object recognition, terminating in the inferior temporal cortex
the cortex is arranged in two ways - what are they?
it is organised into layers, don’t need to know them just appreciate the idea
also organised into four different kinds of columns…
explain how ocular dominance columns were found?
Injected radioactive proline into one eye, it was uptaken by ganglion cells and transferred to the brain, and then labelled the V1 cortex
saw dark columns showing that these areas receive input from one eye - the eye you injected the dye into, and the areas not showing radioactivity must receive input from the other eye
alternatively inject cortex with radioactive glucose and stimulate only one eye
what are possible reasons for having the ocular dominance columns?
May be due to estimating depth perception, though there is minimal proof of this, some papers claims this organisation does not have a particular function
note - LGN has contralateral vs ipsilateral layers, so these columns the columns must receive input from the particular layers of the LGN
what are Blobs and how were they found?
Idk like columns within the ocular dominance columns
Cytochrome oxidase was used to stain a sample - it stained certain neurons better than others, resulting in little blobs
Turns out these neurons receive input from the parvocellular cells of the LGN, and respond to/process info on colour
what are orientation columns?
so you’ve got the ocular dominance columns, then going alongside the other dimension (like along the length) e.g. a left eye column is split further into columns that then respond to different orientations
these would be horizontal, then slightly tilted etc… to vertical
aside from blobs, orientation columns and ocular dominance columns
what is the other?
motion/direction columns
what are simple cells of the V1 cortex?
One of the main kinds of cell receiving input from the LGN
they respond to a specific orientation (shown by Hubel and Wiesel measuring activity from these cells in a cat while it views different orientations of a bar)
where are simple cells located?
what are their receptive fields like?
located at layers 4 and 6 of the cortex
Their receptive fields are elongated/bar shaped (within a circle)
so like the circle is the area in which an object must be for the cell to fire, but it must be in the orientation of the bar shape (anywhere else in the circle will stop firing)
these simple cells don’t respond when the entire receptive field is lit - it requires an EDGE (same for complex cells
what are complex cells in the V1 cortex (input, what they respond to?)
These still respond to orientation like simple cells
In fact it is thought that several simple cells provide the input to a single complex cell. They should respond to the same orientation of a bar, but complex cells should have a larger receptive field (but slightly shifted?)
located in layers 2, 3 and 5
what is the key difference between what complex and simple cells respond to?
complex cells will respond to a bar in any part of their receptive field (simple cells only respond to a bar in the centre of the receptive field)
but complex cells are direction selective
what are hypercomplex cells of the V1 cortex?
responds to a bar in a certain orientation, as long as it is in the receptive field - if the bar is longer and enters the inhibitory surround area the cell stops responding
UNLESS
The bar changes orientation between the inhibitory surround and the centre of the receptive field - in this situation the neurons fire again (so responds combination of orientations)
downstream of V1…
Increase in complexity of the stimulus
Receptive fields increase in size
what does the temporal lobe do? talk about how this was shown
Detects more complex objects - the study tested this with familiar complex objects - faces
Even very basic faces still got the same neurons responding
Progressive removal of hallmark features (e.g. eyes) reduces neuronal response
explain the experiment behind the Jennifer Anniston neuron
Showed humans a large number of images of faces
Found one neuron responding to loads of different images of jennifer anniston, but not other faces
Whether other neurons would work with it, wasn’t confirmed, but it did show that neurons can be super specific in the object they recognise
