Cog Sci of Smell - Neural Coding and Smell

Anything in the world can be represented using anything else

• Basically: Anything that can be represented one way can be represented another way 

• such representations are alwsys imperfect

• Anything that can be represented one way can be represented another way 

 

• You can represent ideas, concepts, etc 

Examples of representations:

• Words represent objects, actions, ideas, etc at some level of detail/abstraction 

• A physical photo, digital photo, and written name might all represent the same person 

• Neurons firing action potentials, releasing neurotransmitters, making certain connections, expressing certain proteins, etc. might all represent an object, action, ides, etc. 

We don't know what neurons are doing to let use smell these certain things - nuerons firing action potentials

• We don't know level of detail neurons operate on to make representation

Representations enable

storage of information and transformation of information via computation

Equations 

• Way of using logic (abstract, logical principles) to understand relationships 

  • When you put x = “three cats" into an equation and solve it, you are transforming a representation of three cats 

Your brain is a machine

• understand the brain - means we can understand the mind better

• Enables you to store and transform information about the world

• We can study how it works to

  • A - understand minds better 

B - figure out how to fix it when it breaks

Given that 

• Sensory organs feed information to the brain 

• Brain damage can disrupt sensory function 

• Sensory activity drives neural activity in the brain 
  
  

You would expect that… 

• A particular sensory stimulus could evoke a particular pattern of neural activity in the brain that corresponds to a particular perception of the stimulus  

Neural representation of sensory stimulus

The pattern of neural activity corresponding to a sensory stimulus 

Neural encoding of a sensory stimulus

The process by which a sensory stimulus gets represented in the brain. This is highly variable across the senses.

Encoding

• When it gets coded in brain 

• Diff codes - light, sound, smell 

• We have to figure out codes and understand how they are mutually intelligible so you can use all info you have in world at once 

Encoding - the neurons that respond to a stimulus might represent

• the location of a light in a visual field, or the frequency of a sound, or the identity of an odor. (simple examples)

Encoding - how the responding neurons act might represent

• brightness of the light, or the intensity of the sound, or the concentration of the odor. (also simple examples) 

One way to figure out - encoding

Look across neuron  Look at how neuron responds  - May be firing a lot when stimulus is really bright or fire less when stimulus really bright - tell  - Stronger odors induce more action potential, or inhibit action potential

Tuning curves of auditory neurons 

Neurons responding - sound intensity and frequency of stimuli 

• activity of neuron

Change tone of sound

• see what happens/ which neurons respond

Asking - How loud does sound need to be to make a particular neuron respond?

Activity of neuron

Change tone of sound

• see what happens/ which neurons respond

Asking - How loud does sound need to be to make a particular neuron respond?

• As they changes the frequency noticed they could use the quietest sound and still get response 

  • Loudest sound it responded to: - 20 dB

This experiment demonstrates 

• The which and how of neural code 

Even if you are only looking at one neuron - can know something ab frequency of sound out there - (also sound intensity) 

Sound frequency - by watching the cells 

- red - 1.3 kHz

- green - 22 kHz

- Yellow - 10.2 kHz

Actual recordings of hippocampal region - can watch across sounds to see as well 

Neuron polarizes than depolarizes 

• Classic action potential - changes of voltages across membrane

recordings of hippocampal region

What voltage looks like on the inside 

• Use Black dye to measure  

• Tree of neuron

• Can record for intracellural electrode

If you poke a hole in neuron you kill it basically 

• Cant really do that if you want to measure it, so you keep cell on outside of brain - cut it open and have exposed 

Tertrode

- type of electron recording used for electrophysiological recordings - assess neural activity 

the difference between inside and outside of cell

• polarization and hyperpolarization

• 2 problems - voltage - difference potential between 2 points 

Electron on outside - doesn't see voltage on outside of brain 

If the electron is only on the outside

• 0 between 2 places 

• Take wiring and screw in skull some place - Skull counted/defined as zero (earth)

• Similar to plugs - one positive and one negative (+ and -) wires coming from there 

• Wire running from everything in your house to the earth - ground copper/cotton? wire in the middle - 

Parts of electrodes recording the hippocampus - intercellular electrode

- inside cell to outside of cell 

• In cell - acc voltage change measured/percieved? 

Parts of electrodes recording the hippocampus - extracellular electrode

End of electrode next to cell and skull/earth itself 

• Seeing ions leaving cell to go into neuron and returning when it repolarizes again - process of polarization

• Outside of cell - watching current change which makes stuff happen in the cell 

The actual recordings from the olfactory bulb show

olfactory system - the action potentials in olfactory system represent the change in a cell’s activity

Extracellular recordings - reference to skull and earth

• green line - noise

• each line/dot - an action potential

breakdown of Raster plot

• Raster plot - each dot is a spike, each line is 1 odor presentation 

• Summary histogram - basically showing it graph style 

• each dot - spike

• each line - 1 odor representation

Can look at how nose responds to smell to see what type of smell it is

Electron getting close to cell - start to hear tapping 

• Pick out individual spike sound 
  

When odor gets presented to nose 

• Before odor presented - neuron tapping back and forth 

• Not when odor presented - goes brrr  - fires lots of action potential 

• Cell also slows down even though odor is till there 

• Then it goes quiet slowly - to original rhythm and fire plate 

Neuron Pattern of activity - isoamyl acetate

• When odor stops - increase in action potential 

Neuron Pattern of activity - butyl acetate

• fires, slows down, stops when it goes away 

Neuron Pattern of activity - methyl valerate

• Hesitate - fire a bunch 

• when methyl valerate goes away, stops

By looking at the olfactory bulb we could see how and when the smell was out there 

Neuron does different things depending on odor, so by watching the neuron you could know what smell it was

Groups of neurons collectively firing code that actually means something - representation/code - Ex: Banana smell

Behaving differently before you present smell 

• Banana smell (Butyl acetate) when it started - neurons had maybe little bit of a pause 

• When it stopped  = they always started firing like crazy - responds to ending of stimulus - though sometimes it started early = responses to absence 

The brain does not know directly

what sensory stimulus is present

• it only knows the pattern of neural activity the stimulus induces

This is what we use to decode the neural representation

The brain/mind decodes neural activity to

• understand the world 

Neuroscientists decode neural activity to

• figure out how the representation works 

if you can decode neural signal into sensory stimuli - (the sensory stimuli that evoked it)

the information about the stimuli must be present somehow

• When you know neural activity but not reason for neural activity 

• Hit something with a baseball bat - could see size of dent after but not bat

Limitations of decoding approaches

• The fact that scientists can decode info from neural activity doesn’t mean the brain is using that info or using it the same way 

• (The brain isn’t recoding from itself with electrodes or fMRI) 

  • Not really learning about brain just engineering solution 

Often we actually do the decoding using machine learning approaches that provide no insight into how the neurons are actually represent the world 

• In this case all we are proving is that the sensory information is “in there somewhere” 

  • Which we already knew 

Very rarely do we have

• a system in which we can reasonably expect to be observing every neuron involved in a neural representation

or

• to know every possible stimulus the brain could be representing. 

Essentially all experiments are based on

• partial codes for things we already knew (what stimulus we presented). It will be amazing when we start finding things in the code that we didn’t already know were there 

Different physical aspects of stimulus represented separately - using different coding mechanisms or different neurons 

• Ex: Some neurons represent respiratory cycle in odor

- others only representing things recorded from brain 

parallel processing in neurons

•  multiple aspects of stimulus are being considered at once 

  • (instead of one after another, which would be slower)

specialization in neurons

•  different circuits or coding mechanisms are optimized for different things 

  • (ex: separate processing of shape and movement in the visual system)  

analysis of stimulus in neurons

•  the brain decomposes the stimulus into its constituent parts 

  • Literally! Sounds and images get decomposed in the mathematical sense into their frequency and phase information via Fourier analysis 

Different physical aspects of stimulus represented separately and using diff coding mechanisms 

• Parallel processing 

• Visual and spatial 

• Something is coming at you and you can get out of way before you know what it is 

• Speed, accuracy - emotion, certain smells - fine tune to needs and experience 

What is being represented by odoes

The chemical structure of the stimulus

• Compute something - what fractions of odor molecules respond that did certain things 
  

The response of odor receptors

• Timing of breathing- slow stimulus- more intense , Fast stimulus - less intense 

• Know both of those things - you can know what the stimulus is 
  

The intensity of the stimulus (measured or computed?) 

The chemical statistics of the world 

• Ex: odor molecules that are common  - brain doesn't pay attention to - your house 
  

The implications for the organism?

• Implications - danger, attraction, spoiled food 

Expected odors given the spatiotemporal context?

The difference between expected odors and actual odors 

Stimulus timing? Sniffing? 

• Timing of access to stimulus - determined by stimulus but also by you

How to discriminate possibilities of odor representations

Decoding

• Identifying odor chemistry - with which cell is responding, concentration, etc.  

• Odor information that narrows down possibilities 

Experimental manipulation of representations