h a l f b a k e r yThis ain't rocket surgery.
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
| |
500,000 sounds like one of those out-of-the-air
figures to me. How could anyone have tested it? |
|
| |
I defy anyone to conceive of a training/testing
program which could demonstrate a dog's ability to
discriminate 500,000 smells. |
|
| |
Also, how does the dog use this olfactory language?
As far as I know, dogs can only produce two odours,
and neither of them is under voluntary control. |
|
| |
//out-of-the-air figures// Well, obviously! What
other sort of figure would you use to quantify
smells? |
|
| |
Reminder to self - find link to video of Episode 1 (or maybe 2) of Series 1 of Red Dwarf, in which the ship's cat is revealed to have evolved over millions of years into a culture who developed olfactory books deep in the ship's hold. |
|
| |
Olfactory books were also developed right here on
Earth - see link. |
|
| |
Wasn't there a Pratchett novel that discussed the complex olfactory language of an alien race during a sensitive diplomatic mission between humans and the aliens during which the human kept outgassing synthesized insults about them? |
|
| |
//I defy anyone to conceive of a training/testing program //
Surely all you need is a multiple choice questionnaire & a pencil? |
|
| |
I'm guessing that dogs can be said to distinguish that many smells in much the way that humans can distinguish a similar number of colours - based on the ability to tell apart closely similar values. |
|
| |
However, it's one thing to be able to tell similar smells apart, but another to be able to recognise that many as individual, absolute values, another another bit of fence to associate those values with concepts, and another another another to deliver them to the dog's olfactory centres with sufficient accuracy and precision (although a tube fed with odorents carefully mixed into a metered air flow, into which the dog places its nose, could do that.) |
|
| |
//I'm guessing that dogs can be said to distinguish
that many smells in much the way that humans
can distinguish a similar number of colours - based
on the ability to tell apart closely similar values.// |
|
| |
Yes, but that doesn't answer the question of how
you'd test that ability in a dog. We're talking
about 500,000 different smells, and they don't lie
on a spectrum (so, you can't say "a dog can
distinguish smells that are 0.0002% apart,
therefore it can distinguish 500,000 smells"). |
|
| |
The only way to know would be to train the dog to
give a response if two smells were different. You
would then need to test all 500,000 test-
substances against each other, for a total of
250,000,000,000 pairwise tests (and that
presupposes you know *which* 500,000 smells the
dog can distinguish). |
|
| |
250 billion tests, performed at the rate of one per
second, would take 7,927 years. That's a good
lifespan for any dog. |
|
| |
Exactly - just as for colour discrimination, it's a bit meaningless to present a figure like that without qualifying the methods used, and what is meant by "different smells". However, you should be able to get some sort of order-of-magnitude estimate by using a lot of interpolation and arm-waving (wearing suitable activated carbon armpit pads to prevent confounding smells, of course). |
|
| |
//You would then need to test all 500,000 test- substances against each other, for a total of 250,000,000,000 pairwise tests// [troll] There's this new thing called 'statistics'. You should look into it some time. [/troll] |
|
| |
But, with colour discrimination, you can rely on
the fact that colour is a continuum. If you take a
simple colour gamut represented in a plane (as an
approximation), you can pick, say, 1000 points on
that plane and ask how far away from each of the
those points a second colour has to be, in order to
be discriminable. Thus, you can map colour
discrimination fairly accurately and get a
reasonably valid answer. (If someone can
distinguish two shades of green which are 5 units
apart on your gamut; and can distinguish two
shades of blue which are 8 units apart; then you
can infer reasonably that they can distinguish two
shades of turquoise which are 6 or 7 units apart on
the same gamut.) |
|
| |
However, I don't see a way to do that for smells,
even with arm-waving. Suppose a dog can
distinguish peppermint from spearmint, and can
distinguish cow's milk from sheep's milk; how do
you deduce from that whether it can distinguish a
pair of similar smells which are "midway" between
milk and mint? There's no gamut of smells, and
hence no way to extrapolate, interpolate or
intrapolate. |
|
| |
Even if you decided to arrange known
smells into some sort of "small gamut", it would
be completely human-centric and arbitrary. |
|
| |
Well, colour vision has approximately 4 dimensions (because the rod cells influence the perception of light quality at low levels), so the task is, as you say, relatively simple. |
|
| |
I suppose you would estimate the total number of primary smells available to a dog, then find out how precisely it can distinguish between concentrations of a representative sample of those, and how that precision is affected as the number of primary smells increases. You would have to model the sensitivity curves (I assume they would be logarithmic). The result would still be rather meaningless, though. |
|
| |
I have some experience in this sort of thing. My honours project partly involved studying scent discrimination in mosquitoes, by observing their behaviour. Another research group, meanwhile, was attaching electrodes to mosquito antennas and exposing them to particular molecules, as well as natural scents that were analysed by mass spectrography. |
|
| |
Ah, but what's the number of primary smells?
Whereas the three primary colours (plus intensity
as per the rods) can be mapped logically to their
receptors, the same is not true of odorants
mapping to their receptors. One receptor may be
activated by many odorants, and one odorant may
activate several different receptors. My guess is
that there may be on the order of 100 different
odorant receptors, and you could argue that 100
different levels of activation could be
discriminated for each receptor, giving 100^100
distinguishable smells, which is clearly a lot more
than 500,000. Or you could argue that
qualitatively different smells would differ in a
binary way, giving 2^100 (or about 10^30) different
smells - again, a big number (far more than the
number of brain cells in a dog, which seldom
exceeds 6). |
|
| |
In other words, any estimation based on a realistic
number of different receptors and combinatorics
gives ludicrous answers. |
|
| |
[EDIT - a quick Google shows that roughly 1000
different canine olfactory receptor genes are
known; this makes estimates based on simple
combinatorics and reasonable assumptions even
more implausible.] |
|
| |
(In contrast, the vision thing works quite
plausibly: four receptors, discriminable to within
1%, gives 100^4 colours, which is not many orders
of magnitude away from the "millions" which are
commonly quoted.) |
|
| |
Yes, we are in agreement - that 500,000 figure is highly dodgy at best. |
|
| |
Obviously, light sensitivity is also roughly logarithmic; we can detect the difference between the light from one candle and two, or a thousand candles and two thousand candles, but not a thousand candles and a thousand and one candles. The same applies to colours and smells. |
|
| |
I'm always curious about "facts" like the 500,000
smells. It's a bit like the folk-fact that the Welsh
share 43% of their genome with Neanderthals -
potentially true, but on what basis? |
|
| |