The Naked Scientists

The Naked Scientists Forum

Author Topic: Feedback: can we really discriminate 1 trillion smells?  (Read 4330 times)

Offline thedoc

  • Forum Admin
  • Administrator
  • Sr. Member
  • *****
  • Posts: 511
  • Thanked: 12 times
    • View Profile
Dalton asked the Naked Scientists:
   I wrote you a few days ago taking exception to the 1 trillion smells claimed by Dr. Andreas Keller. At first, it was just a gut response and rejection without qualification. I have since given my negative response more thought so that I could justify my position. The following is my justification for this position.
 
Sensory Perception and Discrimination
 
The guest has claimed that the original estimate for olfactory discrimination of 10,000 odors is really more like 1 trillion odors. This claim set off my exception detector by making it seem impossibly large (combinatorial explosion problem). After all, we only have 80-100 billion neurons in our brains, so the downstream resources are far too few to support such a large number of recognizable features.  
 
I believe it is necessary to take into consideration that the olfactory sense is not our primary sense and therefore has evolved to have greater limitations than either vision or hearing. In most other species, the olfactory sense is a much bigger deal and I've seen it quoted that half of the dogs cortex may be devoted to the sense of smell.
 
According to Dr. Andreas Keller, we have something like 400 different types of olfactory receptor cells. I suspect that the problem with Dr. Keller's claim stems from an interpretation based on the mathematical extrapolation. The number of possible combinations for 400 different odors is 6.4034e 868. It's obvious that the characterization of smells in our olfactory sense can't be using every conceivable combination, therefore, it must be wired to constrain the number detectable odors.  
 
Mother nature has seen fit to constrain potential combinatorial explosion via wiring and feature collection. If we step back and compare the sensory transducers and their bandwidths between different senses (as in 20K for auditory and 400 for the olfactory sense), I see that phoneme detection has limited sound perception to just a triplet of formants in the time slice of a sound stream.

In English, this scheme has limited perception to the detection of just 44 phonemes. Since what we are talking about here is discrimination and recognition of concepts, I feel I can further extrapolate a comparable potential based on audition and language. Out of those 44 phonemes, English has combined them in sets to generate over 1,000,000 words or concepts. One can't help but realize that the down stream cortical resources to support the upstream sensory field grows exponentially if such a scheme isn't employed.

Circuitry based on the triplet appears to be very popular in nature. DNA uses the triplet to generate codons and vision employs a triplet of colour photoreceptors, and such a triplet scheme may be used to constrain odour perception as well.  
 
If I hold perception to a constant based on the triplet, then I can say that "if 20K auditory transducer circuitry employing triplets can generate 1,000,000 concepts",  
and if, 400 olfactory transducers employing a triplet for identification and recognition is only 20% of what audition has, then 20% of 1 million is only equal to 200,000
 separate smells (not taking into consideration variations in intensity).  
 
Our sense of smell may have a potential greater than 10,000 odours. It may be more like 200K, but not 1 trillion. Fortunately, this number allows for multimodal
 relationships and learned associations. Was this paper ever peer reviewed?
 
Sincerely,
Dalton Seymour
What do you think?
« Last Edit: 12/04/2014 15:30:01 by _system »


 

Offline CliffordK

  • Neilep Level Member
  • ******
  • Posts: 6321
  • Thanked: 3 times
  • Site Moderator
    • View Profile
Re: Feedback: can we really discriminate 1 trillion smells?
« Reply #1 on: 15/04/2014 15:18:48 »
One may not be able to name 1 trillion smells.  However, if the question is whether given any two distinct smells out of a trillion possibilities, could we determine whether they are the same or different, it may well be possible.  Or, perhaps could our favorite pet dog determine if they are the same or different.

Consider a binary multiplexer.
40 receptors.

240 = 1,099,511,627,776, so with just 40 distinct receptors, one could detect over a trillion distinct patterns of activation. 

You can think of the neurons in your brain as multiplexing and combining data. 

Vision is interesting in that one can recognize a vast array of colors using basically 4 types of receptors, 3 cones + the rods, with any individual receptor either 100% firing or not at any moment in time.
 

Online Bored chemist

  • Neilep Level Member
  • ******
  • Posts: 8654
  • Thanked: 42 times
    • View Profile
Re: Feedback: can we really discriminate 1 trillion smells?
« Reply #2 on: 18/04/2014 18:33:08 »
I can distinguish two 20 digit numbers, even though I plainly couldn't remember all the possible 20 digit numbers.
 

Online chiralSPO

  • Global Moderator
  • Neilep Level Member
  • *****
  • Posts: 1872
  • Thanked: 143 times
    • View Profile
Re: Feedback: can we really discriminate 1 trillion smells?
« Reply #3 on: 19/04/2014 04:13:49 »
It isn't obvious to me that "only 80 to 100 billion neurons" is not enough to process 1 trillion scents. While there may only be 10^11 neurons, there are over 10^14 synapses in the average human brain. It is my understanding that the number of synapses is a better measure of processing power than the number of neurons.

That said, distinguishing 1 trillion (10^12) smells does sound somewhat fantastic.
« Last Edit: 19/04/2014 04:17:12 by chiralSPO »
 

Online Bored chemist

  • Neilep Level Member
  • ******
  • Posts: 8654
  • Thanked: 42 times
    • View Profile
Re: Feedback: can we really discriminate 1 trillion smells?
« Reply #4 on: 19/04/2014 11:38:42 »
Is it credible that, of our 10^14 synapses we use 10^12 just for the sense of smell?
 

Offline evan_au

  • Neilep Level Member
  • ******
  • Posts: 4105
  • Thanked: 245 times
    • View Profile
Re: Feedback: can we really discriminate 1 trillion smells?
« Reply #5 on: 19/04/2014 13:32:38 »
Quote
The number of possible combinations for 400 different odors is 6.4034e 868.

As I recall, the guest speaker did say that their experiment used mixtures of a small number of chemicals, much less than 400; many of these chemicals would trigger multiple sensory nerves in the nose with its own unique pattern. They then used the experimental results to extrapolate the full number.

They weren't looking for a binary present/absent indication of particular chemicals, they were looking for a same/different comparison: "I have 3 vials: A, B and C. Two of them are the same, the third is different. Which one is different?" This test is able to estimate the sensitivity of the nose in picking up small differences in the concentration of the chemicals.

The brain summarizes multiple inputs into a significant concept, like "coffee", "mint", etc, even though there are many distinct types of mint or coffee. I do not think that the number of synapses will be a limit on the number of distinguishable types of mint or coffee - but it probably would constrain the number of concepts we could attach to combinations of smells.

The conclusion was that the nose could probably distinguish over 1 trillion odors - 6.4034x10868 certainly fits that description!
 

Offline alancalverd

  • Global Moderator
  • Neilep Level Member
  • *****
  • Posts: 4699
  • Thanked: 153 times
  • life is too short to drink instant coffee
    • View Profile
Re: Feedback: can we really discriminate 1 trillion smells?
« Reply #6 on: 21/04/2014 00:40:26 »
I think the question is absurd. What defines a "different smell"?

Suppose I make a mixture of 33.3333% each of three volatile chemicals as a starting point. If I alter two of them to 33.3334 and 33.3331% respectively, I may have found a way around a perfumier's patent but have I created a distinct smell? Scientifically, yes, but would an observer give it a different name?

There's no doubt that some animals (dogs, sharks...) can detect less than 1 part in 1010 of a fair spectrum of molecules in air or water. Humans seem particularly sensitive to H2S (5 parts in 109 in air). So if I add a broad hint, say 1 ppm, of H2S to any one of a million other mixtures, have I created another million smells? It would certainly be detected in the absence of any other sulfides or mercaptans.

Clearly the alternative interpretation, that we are able to name 1012 analytes from a single sample, is nonsense as nobody is likely to have encountered that many nasally active molecular species. So what exactly does it mean?

As an aside, some years ago I was working on perfume compounding for a new detergent. We set up a series of about 30 paired preference tests with a variety of standard synthetic olfactants. There was very little correlation of preference, but when we asked people to classify each pair as "high" and "low", without stating what those terms meant, we achieved almost 100% agreement. I  think the connection between smell and vocabulary is quite deep. 

« Last Edit: 21/04/2014 00:47:48 by alancalverd »
 

Online Bored chemist

  • Neilep Level Member
  • ******
  • Posts: 8654
  • Thanked: 42 times
    • View Profile
Re: Feedback: can we really discriminate 1 trillion smells?
« Reply #7 on: 22/04/2014 18:45:12 »
"I think the question is absurd. What defines a "different smell"?
This
"I have 3 vials: A, B and C. Two of them are the same, the third is different. Which one is different?"
 

Offline CliffordK

  • Neilep Level Member
  • ******
  • Posts: 6321
  • Thanked: 3 times
  • Site Moderator
    • View Profile
Re: Feedback: can we really discriminate 1 trillion smells?
« Reply #8 on: 22/04/2014 19:03:35 »
The notes I'm seeing indicate that mammals have up to about 1000 different types of olfactory receptors, of which Humans express about 400.

Any particular odor would activate a subset of the 400 different receptors. 

Let's assume each receptor is either 100% activated, or 0% activated.  This may not be the case, and it is likely that some chemicals would either strongly or weakly activate different receptors, and these fine details may be detectable. 

If every binary combination of receptor activation was indeed possible, that would give one 2400 different odors that could be detected, or a good deal more than a trillion.  If one can further differentiate between weak receptor activation and strong receptor activation, that would increase the numbers even further. 

Now, the question is what you can do with all of those odors.  You may well not have the ability to memorize a trillion different odors.  However, if presented with a novel odor, you could potentially memorize all the characteristics of that odor so that when you encounter it again, you'd be able to precisely recognize it, and differentiate it from all other odors, and this distinction can be extremely precise.

I think the study was about 100% vs 0% of any particular chemical odor.  To some extent, one can also judge proportions.  A little pepper vs a lot of pepper.

Many animals seem to be a little better at differentiating odors than humans despite generally having smaller brains. 
 

Online chiralSPO

  • Global Moderator
  • Neilep Level Member
  • *****
  • Posts: 1872
  • Thanked: 143 times
    • View Profile
Re: Feedback: can we really discriminate 1 trillion smells?
« Reply #9 on: 10/05/2014 20:53:34 »
Is it credible that, of our 10^14 synapses we use 10^12 just for the sense of smell?

I don't think it works like that. Presumably combinations of synaptic pathways are responsible for data storage and recall. In this way perhaps only a few hundred synapses can store and process all of the information generated from the combinations of a few hundred receptors...

A paper I found (open access! http://chemse.oxfordjournals.org/content/27/1/1.full) counts the number of neurons in the olfactory bulb required to distinguish smells. They found that a collection of only a few hundred active neurons in the olfactory bulb of a rat was sufficient for distinguishing the smells used in the study.
 

The Naked Scientists Forum

Re: Feedback: can we really discriminate 1 trillion smells?
« Reply #9 on: 10/05/2014 20:53:34 »

 

SMF 2.0.10 | SMF © 2015, Simple Machines
SMFAds for Free Forums