Smelling soundIs trilled smell possible? Ed Cooke and Erik Myin raise the question in the JCS.  Why do we care? Well, for one thing smell has always tended to be the poor relation in discussions of conscious experience. The science of vision is so much better developed that seeing generally looks a more tractable area to attack, but arguably the discussion is somewhat lop-sided as a result; ‘seeing red’ isn’t necessarily a perfect epitome of all sensory experience, so a bit of clarification around smells might well be useful.

But the main point of asking the question is to test what Cooke and Myin call the independence thesis: the view that the experienced character of sensations includes a ‘something it is like’ over and above the gross physics of the business: that there’s an ultimate smelliness about smell that has nothing to do with the details of the sensory process. I would say there’s a range of possible positions here. Hardly anyone, I think would say that the physics of perception is irrelevant to how the experience seems. We know that that the wave structure of light and sound determines some of the characteristics of the experiences of vision and hearing, for example, and we know that smell is vaguer about location than vision because it depends on gases wafting around rather than sharply defined rays of light.  But beyond that the consensus breaks down. Some would say that these physical characteristics are just the basics and the real excitement lies in the ineffable qualities of experience. Purple is a thing in itself, not a blank sensory token which would do equally well for the smell of coffee, they might say.

Some would go further and accept that the qualities of experience are very largely determined by the physics of the medium and sensory apparatus, but that there’s a certain something beyond that which doesn’t reduce to the simple physics.  Rigorous materialists will be tempted to go further still and take the view that however complex and indefinable our experiences may seem, they are fully determined by the qualities of the processes that give rise to them: this of course, amounts to denying the existence of ineffable qualia. (My own view, for what it’s worth, lies an infinitesimal distance short of this extreme.)

Cooke and Myin’s approach is to look at the consequences of the independence thesis. If it’s true then we ought to be able to transfer the forms of one sensory modality to another without it losing its identity. So, in sound we can have a trill, a very rapid alternation of two notes; if independence is true, we ought to be able to have trilled smells.

Before tackling the thought experiment in more detail, Cooke and Myin provide a brisk review of some of the relevant science, including some odd and interesting facts. The smell of pressure-cooked pork liver is made up of 179 different compounds; airflow is indispensable to smell (having your nose full of smelly stuff or your receptors stimulated produces no sensation unless there’s airflow); and so sniffing is more important than you may have thought. It turns out that human beings are pretty well incapable of identifying single components of a smell when there are more than three – so much for perfume designers – and perception of smell is also very heavily conditioned by previous experience (if you’ve encountered smell b together with lemony smells in the past, you’ll tend to think smell b has lemony notes even when the lemon smells are objectively absent).  It looks as if we might each be working with a typical vocabulary of about 10,000 known smells, out of a theoretical 400,00 that the nose can distinguish: best estimates suggest that smell-space has a minimum of somewhere between 32 and 68 dimensions (as compared to human colour vision’s paltry 3).

Now we come to the thought-experiment itself.  It seems that Jesse Prinz has denied the possibility that a sound could become a smell merely by changing the structure of the experience (could the sound of a fire alarm ever become the smell of smoke?), so with fine daring, that is the first transition Cooke and Myin propose to anatomise in a thought-experiment.

Thought-experiments are always a little unsatisfactory because they don’t really force people to accept your conclusions in the way that a proper argument does. In this case, moreover, it seems to me there’s a particularly difficult trick to bring off because for the experiment to convince, Cooke and Myin want the transfer of properties to seem plausible; yet the more plausible it seems the more plausible independence seems too.  They want us to believe that they’re doing the best possible description of a transfer that could plausibly happen, in order to convince us that once we understand it it’s not plausible that it’s really a transfer at all.

However, I think they do a commendable job. First, sounds have to become less distinct in their onset and direction; they have to be more like generalised hums which float around appearing and dispersing slowly (no good for rapid warnings any more). The we have to imagine that we use our noses to detect sounds, that they only become perceptible when we breathe, and that sniffing or breathing deeply affects their intensity. We must imagine that it’s now a little more difficult to pick out single sounds when there are several at once: we might have to think about it for few moments and take some extra sniffs.

That’s not too bad, but there are bigger problems. We’ve noted that smell space appears to be huge; Cooke and Myin suggest we could enlarge sound space the same way by imagining that the differences in sound are like the differnces in timbre between musical instruments (though we have to suppose that we can readily distinguish the timbres of 10,000 or so different instruments). On the other hand, musical notes fit on an organised scale with perceptible relationships between different notes: smell doesn’t really have that, so we must drop it and assume that sounds are essentially monotonous. To round things off with behavioural factors, we should think of sound as no longer used for communication, but mainly for the evaluation of the acceptability of food, people and other biological entities; and we should imagine that sounds now have that characteristic of certain smells which allows them to evoke memories with particular potency.

If you’re still with the experiment, you’ll now have some intuitive idea of what it would be like if sounds had the structural and other characteristics of smells. But no, say Cooke and Myin: isn’t it apparent now that the sensations we’re talking about wouldn’t be sounds any more (in fact they would pretty much have become smells)? Isn’t it clear, in short, that in order to be trillable, smells would have to cease being smells? They go on to a further thought experiment in which smells become colours.

This is a valuable exercise, but as I say, thought experiments are not knock-down arguments, and I am willing to bet there wil in fact be plenty of people who are prepared to go along with Cooke and Myin’s transition but insist at the end that the sensations they’re imagining are still in some way sounds, or at least have a core soundiness which makes them different from echt smells. (You notice how I criticise the weakness of thought experiments and yet here I am doing something worse – a kind of third-person thought-experiment where I invite agreement that in certain odd circumstances other people would think in a certain way.)

Personally I think some of the most interesting territory revealed here is not so much at the ends of the transition as in the middle. The experiment raises the possibility of mixed modalities never before imagined, chimerical experiences with some of the characteristics of two or more different standard senses. Not just that, either, because we can invent new physical constraints and structures and develop possible sensory modalities which have nothing whatever in common with any real ones, if our imagination permits.

This gives Cooke and Myin some possible new ammunition. Do all these imaginary new modalities get their own essence, their own qualia? If we mix smell and hearing in different ways, do we have to suppose that there are distinct qualia of, er, smearing and hell?

For that matter, what if we took a subject (all right, victim) through the transition of sound to smell; and then separately gave him back sound? has he now got two distinct experiences of sound? Then if we move the new sound2 though the transition to smell, does he have two smells? And if we then give him back a separate sense of sound again? And so on.

I can’t help thinking it would be quite a Christmas present if we could have a sense with the spatial distinctness of vision, the structured harmonics of sound, and the immense dimensionality of smell. There would be some truly amazing symphonic odours to be painted.

Merry Christmas, all!


  1. 1. Vicente says:

    Merry Christmas !!

    have you realised that this year Xmas carols come beautifully perfumed ?

    Very appropriate post for the season, a bit disturbing though. That’s the point, since there is no strict and univocally defined mapping between neurological processes and their phenomenal counterpart… fantasy has a big room to operate.

    Maybe evolution will bring your Xmas present, if a mutation makes it possible, it seems that it would provide several advantages in order to gather information from the environment. I believe all the constraints lie on the neurological field, I don’t see limitations on the other side.

    “My own view, for what it’s worth, lies an infinitesimal distance short of this extreme”

    Nice one. Sometimes there is no difference between an infinitesimal distance and an infinite one, in what to the quality of an explanation concerns… ;)

  2. 2. Kar Lee says:

    Merry Christmas! And happy holidays to those who celebrate other traditional holidays in the same season.

    Interesting Article. I remember when I was a teenager when the dawn of personal computer was about to begin, a geek type classmate of mine pulled me aside and asked me if I would like to hear some computer “music”. What he had was a CPU board, which was hooked up to a cassette tape player/recorder and all his computer programs were all stored in this regular cassette tape which was usually used for storing music. So, he played the computer programs out as if they were music and it came out like squeaky noise, nothing like “programs” or “music”. But since the cassette player, being an analog device, could tell what it was playing, played whatever was there on the tape into sound. I am wondering our cross-modularity sense is a similar type of misuse of the senses: something that is used not as it is designed to.

    But on the other hand, we do have example of bats echo locating system that replaces vision in the dark. The more interesting experiment is the one involves using a pair of “eye glasses” with a built-in scanner which translates the line scan of objects in front to sounds that the wearer can distinguish. With training, over time, the wearer is able to make use of the sound, and the result is a description of a vivid picture in front but it comes through the audio channel.

    But for the information of sound to come through the channel of smell, it goes a little beyond the limit of my imagination, sort of. If I force myself, I still can imagine myself sniffing through some smells, which rapidly changing its intensity or quality which I was told to represent certain sound sequences. If there is a transducer of some sort that emits different smells according to what the piano is playing. Over time, I will be able to tell, perhaps, from the smells what the piano is playing and may even be able to enjoy this smelly music. I don’t know, but it is an interesting possibility.

  3. 3. Vicente says:

    Kar Lee,

    It is funny that you comment that. Your experience with the computer music has been shared by millions of teenagers all over the world, that in the eighties started playing with the Sinclair “toys” based on the Z80 processor, like the ZX81 and ZX-Spectrum (8 MHz). The programmes were loaded to the (8kb, 64kb for the lucky ones) on board 8pin memory chip, ha ha, from a cassette tape. During the loading process your could hear the squeaky sounds you refer to, and see strange lines on the tv set screen… it brings to me very good memories. You should have seen that ZX80 assembler code… the built in BASIC was crab, and people developed whole applications (and nice games) in assembler, believe or not.

    Anyway, I think it was a very good example of cross-sensing odd effects.

    What about the ferrets experiment in which the optical nerves were connected to the “audio-cortex” and the brain reconfigured itself allowing the ferrets to see….

    I believe it is more profound, it seems the brain is like an on the fly transducer, depending on the input data format it can be configured to produce an optimal image of the surrounding world.

    Maybe if you plug the nose to the visual cortex you end up seeing 3D blotches or clouds in different colours representing smells.

  4. 4. Shankar says:

    Wish everyone here a happy 2012!

    Actually a perfume harmonic progression is not totally outlandish. Same with tasting multiple ice cream flavors in some order. The pleasure of tasting strawberry, vanilla, chocolate, cherry, etc. flavored ice creams may differ depending on the order (similar to a chord progression) and different persons may have different sequences which they might prefer over others.

  5. 5. Ron Murphy says:

    “smell-space has a minimum of somewhere between 32 and 68 dimensions (as compared to human colour vision’s paltry 3).”

    Could you expand on this? What constitutes a dimension in smell space? Even if there may be more dimensions, aren’t they more ‘shallow’ than colour space dimensions? And when you add the binocular component that exposes 3 dimensions of space it seems our colour ‘space’ is made up of ‘deeper’ dimensions, even if fewer. So ‘paltry’ doesn’t seem to tell the whole story.

  6. 6. Peter says:

    Ron, each dimension represents one independently variable quality. For colour, we only need three independent variables (red, green and blue, or alternatively hue, saturation and luminance)in order to describe every possible colour. To describe every possible smell, it seems we need at least 32 different variables, perhaps as many as 68. The source is a 2004 paper by Mamlouk and Martinetz: there’s a pdf of it here if you’re interested.

    By shallowness, I suppose you mean we can discriminate fewer different levels on the smell axes than we can do for colours? Interesting suggestion, but I don’t know of any research that confirms it. (Now the question comes up I don’t know for sure that even colour space is necessarily equally “deep” in all dimensions, though I’ve always assumed it was.)

    Of course you’re right that there’s a lot more complexity to vision in other respects; even so, I think the complexity of “smell space” is striking.

  7. 7. Ron Murphy says:

    Hi Peter,

    I agree that the complexity of smell is interesting, and far more so that we might intuitively feel. Given the link between taste and smell then experienced wine tasters are possible evidence of this, though I’m not sure their acuity is as consistent as they often claim.

    It is interesting to wonder at the comparison of degrees of freedom and the acuity in any one dimension. Personally I think the special content of visual perception is important, even when dealing in monochrome light – just look at what you can read in this comment. Could you attain anything like this detail from smell?

    There is interesting work that isn’t usually associated with this issue. Tongue stimulation by electric signals from a matrix attached to camera source can help the blind ‘see’ through their tongues. This seems to cause a re-assignment of brain function to pay attention to this stimulus for the purposes of visual processing. Are they ‘tasting’ vision, or feeling vision? Is electrical stimulation of the tongue a sort of taste function?

  8. 8. Ron Murphy says:

    ‘spacial’ content, or course.

Leave a Reply