Sensory systems and sensory input Flashcards
What 4 things does a visual neuron encode?
1) Contrast
2) Colour
3) Frequency
4) Orientation
What is contrast?
An increase or decrease in light intensity from a certain level
What is visual system interested in, rather that light intensity?
CHANGES in light intensity (contrast) in time/space
What frequency does the visual system encode?
Spatial/temporal frequency:
- How quickly neighbouring pixels change
- How quickly the intensity changes in time
Why is orientation important in an environment?
It encodes the EDGES of an object
What information does the visual system use from the environment in order to arrange itself?
- Contrasts are not random
- Colours are not random
- Orientations are not random
- Neighbouring pixels are similar in brightness and colour
Why is colour not random in the environment?
- Blue dominates the UPPER part of the visual field (sky)
- Green dominates the LOWER part of the visual field (grass)
Why do neurons adjust their sensitivities to contrasts?
To be able to process very bright and very dim pixels at the same time
What 2 orientations dominate the visual scene?
Why is this the case?
Vertical and horizontal
Due to gravity - stable orientations are either stood up (vertical) or lay down (horizontal)
Describe the distribution of pixel brightness in the environment?
Very wide
Describe the distributions of the key properties of the natural scenes
They are NOT random
They have certain statistics
How do the retinal neurons respond to the environment? (graphically, where X axis is property of the environment and Y axis is response amplitude of the neuron)
What does this show
Sigmoidal response:
- Most information is encoded in the middle (where small changes in the property of the environenment has large changes in the response amplitude)
- Curve saturated - very large changes in property has very small changes in response amplitude
What problem does the broad distribution of a property cause?
Problem for the encoding of the sensory neuron:
- The very edges of the distribution (very high contrast, very low contrast), which are more rare are still VERY important
- Most edges have very high contrast - visual system is tuned to process edges
- However, these lie of the flattened portion of the tuning curve (large changes in property, small/no changes in intensity)
How are the problems of the distribution/tuning curve matching overcome?
1) Tuning
2) Adaptation
Describe the tuning strategy
Different neurons tuned to different parts of the distribution:
- Some neurons - high sensitivity, respond to low stimulus intensity
- Others - intermediate
- Others - Less sensitive, respond to high contrast/intensity
Describe the adaptation strategy
Neurons have plastic curves, that can become more/less sensitive
How does the retina use the information from the environment (colour is not random)?
Green cones - located at the TOP of the eye
Blue cones - located at the BOTTOM of the eye
Where the ganglion cells that respond to approaching objects (increasing in size) present?
Why are they needed?
In the bottom of the retina
Sky is static - need to know if anything changes (eg. approaching bird)
Birds usually in the sky
Does the retina process negative or positive contrasts better?
Why?
Negative better
There are more negative contrasts in the environment
How does the retina better process negative contrasts?
More OFF cells than ON cells
Better to calculate decreases in light intensity
What calculation is used to calculate the amount of information that can be transmitted by a neuron?
C = B log2 (1 + S/N)
What is S/N?
What happens when this is increased?
Signal to noise ratio
When increase - more info can be transmitted
What limits the amount of information that can be transmitted by a neuron?
How is this calculated?
N - noise
Noise = individual response - signal
What is the main source of noise in a neuron?
- Spontaneous opening of channels (Na, K)
- Affects spike rate
- Spontaneous opening of Ca2+ channels that is NOT triggered by the release of membrane potential
What happens to the amount of information processed per spike when the spike rate of a neuron increases?
Increase spike rate - smaller amount of information that can be processed per spike
Why is it beneficial to keep the spike rate of a neuron low?
So that more information can be processed per spike
Each individual spike costs the neuron lots of ATP molecules
How is information in a neuron carried?
In ‘bits’
What happens to the cost per bit if more information is transmitted by a neuron?
Why is this a problem?
More cost per bit
Small number of spikes per second
What did Gollisch state?
Information is encoded by LATENCY of spikes NOT the frequency
