Why four’s the aroma limit

four aromas

I recently had an article published in Circle Update (the magazine of the Circle of Wine Writers). It concerned the number of aromas used in wine tasting notes. If you are interested you can view and download a PDF offprint of the article here: Why four’s the limit.

It draws heavily on a four-part series of blog posts I wrote late last year. Compared with the Circle Update article, these contain more words – not necessarily a good thing – but also, in “the science” post, there is considerably more information about the scientific basis for the notional limit of only four aromas being identifiable in blends.
How many identifiable aromas in a wine – the dilemma
How many identifiable aromas in a wine – the science
How many identifiable aromas in a wine – tasting experience
How many identifiable aromas in a wine – my conclusion

How many identifiable aromas in a wine – my conclusion

I have now done enough scene-setting and pussyfooting around: see my previous three blog posts (in chronological order 1, 2, 3). It’s time for me to nail my colours to the mast and say what I really think about tasting notes that mention many different aromas when science tells us we can only identify four in a mixture.

Jan_Brueghel_I_&_Peter_Paul_Rubens_-_Smell_(Museo_del_Prado)First of all, I cannot find any particular problem with the scientific evidence for our poor ability to identify aromas in a mixture, and I see no reason to doubt its applicability to wine. If anything I would expect it to be an easier task to identify aromas in the experimental situation than with wine, as in the experiment there were always subsets of the same 7 or 8 odours, as opposed to the much larger number that people find in wines.

There is a lingering doubt in my mind because the experiments presented odours only to the nose. With wine however, aromas are also detected when it is in the mouth. Does that cause a greater number of aromas to be detected in total? My personal experience suggests that happens only very occasionally, and to a small extent. However, in the tasting note example I gave in the introductory blog post for this small series, the aromas detected on the nose and palate are markedly different. I may return to this issue in the future, but my initial feeling is to go with my personal experience. And what about letting a wine develop over a few hours or days? Does that let additional aromas develop and become identifiable? It is possible, but again in my experience it is rarely the case.

I acknowledge that there is also experimental evidence that it is possible to tell if a single very familiar odour is present in mixture containing up to 12 odourants. Additionally there is anecdotal evidence of perfumers and chefs being able to detect single missing ingredients in complex familiar recipes. But these tasks are very different to identifying aromas in an unfamiliar wine.

For now, let us take the identification of an odour object in a wine literally, by which I mean that identification means there are key aromatic compounds in both the wine and the actual odour object. With this literal interpretation, I think it is fair to say that the limit of four correct identifications will apply. Indeed, the experiments suggest that even with fewer named aromas it is unlikely they will all be correct.

The literal interpretation of aromas I have just described is not totally unreasonable. Certainly in some cases it seems that the same chemical is responsible for the aroma in both the wine and the real aroma object. Rotundone, which is found in black pepper and Syrah, is one example. However, aroma objects mentioned in the tasting note may merely be reminiscent of the real thing. Or, as some less kind people might put it: imagined or made up. In these cases, there can clearly be no limit of the number of identifiable aroma objects, but by what criteria can we judge the value of such lists?

For me, the main criterion for a successful tasting note is its ability to communicate the experience of drinking the wine to the reader. And here I mean to communicate accurately; not just to give an impression of what the experience might hypothetically be.  When I am tasting, the correspondence with tasting notes independently written by others is usually minimal. I have seen no formal studies into how common this experience is, but we can also get hints by comparing tasting notes of the same wine written by different people. Usually there is little similarity, and sometimes the differences are huge. It is interesting to speculate about to what extent the differences are due to the subjective nature of taste, and to what extent it is imperfect communication; but differences there are.

Speaking personally, the tasting notes I find communicate best are those where the aromas listed are few and vaguely described. For example, it can often be accurate, and still helpful, to identify citrus fruit in a wine. But when someone else describes a wine as tasting of lemon, I often decide it is closer to lime. And does anyone actually care? It is difficult to imagine a disgruntled customer returning a bottle of wine to a shop because the type of citrus fruit was incorrectly described on purchase. The precision of description is linked the issue of the number of aromas: one person’s citric could be another’s lemon, lime and clementine. The level of detail we use in tasting notes is another interesting topic to which I might return.

In summary, as promised, here are my colours on the mast stated with an unjustified sense of certainty. There are two reasons why I am suspicious of tasting notes with a long list of aroma objects:

  1. If you take a more literal interpretation of aromas in tasting notes, it is impossible to produce correct lists containing more than four aroma objects .
  2. I am not convinced about how useful long lists are anyway. I favour a shorter tasting note that contains only the dominant aromatic components, and one that is not over-specific in its aroma descriptors.

In the words of Carveth Read: It is better to be vaguely right than exactly wrong.

How many identifiable aromas in a wine – tasting experience

I continue to investigate the number of aromas we can detect in a wine. There seems to be a conflict between scientific research, which has shown that we are incapable of identifying more than four odours in a mixture, and the testimony of wine tasters who name as many as 10 or more aromas. For the story so far in more detail, see my previous two blog post here and here. Now I will look more closely at how wine tasters come to identify so many aromas, and what it means when they do.


I know that some tasters sample their wine over the course of an evening, with and without food, and possibly even continue with the same bottle over two of more days. That is potentially going to lead to longer tasting notes, and the naming of more odours, because the wine, the context, and even the taster might change over that period, and the changes could result in different proportions of various aromatic chemicals, different chemicals becoming available to the nose in the form of vapour, and different taster sensitivities.

Other tasters however are perfectly capable of reeling off a list fruits, herbs and spices within the space of a few minutes. The people I have seen doing this have all been Americans with some level of sommelier qualification, and my preliminary conclusion is that it is related to their training and culture. In the case of professional wines critics who name many odours in a tasting note, I suspect those lists are also usually produced quickly due to pressures of time. This is a very different situation to the tasters who make their notes over an extended period, as nothing has chance to change much after the first nosing of the wine.

But do the odours named in these long lists actually exist in any real sense in the wine? Here, by “in any real sense” I mean as chemical compounds in concentrations that could stand any chance of detection. Note that Laing’s limit of four applies not to the taster’s imagination, but to the correct identification of odours physically present in the mixture as chemicals. To answer the question about whether the odours exist, we would need to perform a detailed chemical analysis of the wine.

Even if the odours do not really exist, there may be understandable reasons why they may be perceived. In Avery Gilbert’s book What the Nose Knows, there are several examples of how suggestible we are when it comes to our sense of smell, and suggestions of what we might find in a wine can come from many sources. The best known example is perhaps that the addition of red food dye to white wine prompts people to find aromas usually associated with red wines. In real life all manner of things might suggest what aromas should be in the wine, not least any hints as to what type of wine is in the glass. There is no shame in being suggestible in this manner. It is simply the way we humans work perceptually, and as we also are the ones that drink the wine does it really matter?

We must also acknowledge that odour perception is a complex multimodal process, and seemingly unrelated stimuli can affect our sensitivity to odours in ways that are not explicable by suggestibility. For example, a drop of a sweet substance on the tongue has been found to increase the sensitivity of Westerners to an almond aroma. Effects like that could also cause us to identify more odours than Laing’s experiments suggest we should.

Finally, I am convinced that some tasters fabricate flavours because they feel, for whatever reason, a few more are needed. I too would do that under certain circumstance – if I were taking some sort of test for example, and I was required to list a certain number of flavours. If you have a good idea what the wine is, it is very easy to throw in a few extra flavour descriptors that would not raise eyebrows. Claret? OK, that will be blackcurrant, pencil box and French oak then. In fact, come to think, I have done it, in a tasting competition a few years ago when I really thought the wine was so mediocre there was little to say about it.

On the subject of how many identifiable aromas there are in a wine, that is now all the detail and preliminaries out of the way. So far I have tried to be as objective as possible in describing the evidence, but in my next post I promise will stick my neck out and say what I really think. I just have to first decide what that is.

How many identifiable aromas in a wine – the science

This is the second post in a series that looks at the number of aromas we can detect in a wine. My first post explained that many tasters claim to be able to identify several aromas in the same wine, while scientific research has shown that we are incapable of identifying more than four. Here I will explain more precisely how the research was carried out, and what it found.

The research was carried out by David G Laing and co-workers in the 1980s and 90s. As far as I know it has not been replicated in other laboratories, but as will become clear the experiments were effectively repeated in the same laboratory, and as far as I know the results are not controversial. They are, for example, described in Avery Gilbert’s book What the Nose Knows without being questioned.

In Laing’s first published work on this subject[1], mixtures of odourants were delivered to the noses of the subjects in vapour form through a duct 150cm long and 55cm in diameter, and the subjects were asked to identify the odours present. Each of the 7 odourants was a single chemical compound with a distinctive smell, and introduced to the subjects under its everyday name: vinegar, almond, spearmint, fruity, cloves, orange or camphor. Odourant concentrations were chosen to be moderately strong perceptually, and roughly equally strong for all odours individually. The odourants and were also known to be identifiable in binary mixtures in the chosen concentrations. The mixtures were created in vapour form for delivery to the subjects’ noses. Vapour, rather than liquid, mixing was chosen to reduce the possibility that the odourants would react chemically. It seems to me that all reasonable steps were taken to give subjects the best possible chance of identifying the odours.

The subjects were first allowed to familiarise themselves with the individual odours for several minutes. Then they were given mixtures that varied in terms of the odourants used and the number present, and asked to identify the odours in the mixture. The results are summarized in the figure below[1], which is the basis for the assertion that no more than 4 odours can be identified.laing_1989A few possible reasons for the low number of identifiable odours spring to mind: maybe the subjects were not skilled enough, or maybe the odours chosen were particularly difficult to identify. The investigation of those two possibilities was the subject of two further studies [2, 3], which were based on the same experimental setup.

To determine the effect of training and experience[2], two different sets of subjects were used: experts, and trained non-experts. The non-experts were trained over the course of 5 days, and on day 5  they were tested to ensure they could reliably identify all 7 odours when presented individually. The other group were expert perfumers and flavourists; they were “highly familiar with the test odourants, had daily experience with odour discrimination, and had a prerequisite before entering their respective professions of an excellent ability to discriminate and identify odours”. The experts and trained non-experts performed slightly better than the subjects in the initial study, with the experts being the most competent, but nevertheless the number of correct judgements still dropped to negligible levels when more than 4 odourants were presented.

The next study[3] looked at the importance of the type of odours used in the mixtures. A panel of 10 perfumers and flavourists selected two sets of 8 single compound odourants: good and poor blenders. The poor blenders were judged to be perceptually very different, and thus easier to identify in mixtures than good blenders. The common names of the good blenders were rose, musky, cinnamon, coconut, fruity, orange, burnt caramel and almond, while the poor blenders were bad breath, mushroom, cut grass, Dencorub, garlic, antiseptic, aeroplane glue and mandarin. The results were as suspected in the sense that the poor blenders were easier to identify. However, the differences were not large, and for 5 or more poor-blender odourants the number of correct identifications was still very low indeed.

The fourth study[4] brought the experiments one step closer to the reality of wine. Here, none of the odourants were single chemical compounds, but complex mixes of chemicals usually identified as a single odour: smoky, strawberry, lavender, kerosene, rose, honey, cheese and chocolate. Again, the results were similar to those of earlier studies, suggesting that we identify the smell of familiar objects as a single gestalt odour, even if multiple chemical compounds are involved in creating it.

In the last article of Laing’s I will discuss here[5], subjects were trained to identify the test odours, and then required to detect a single highly familiar odorant in stimuli consisting of one, four, eight, twelve, and sixteen odorants by using a selective-attention procedure. Identification fell to chance level when sixteen odorants were present. Unfortunately I only have access to the abstract of this article, but I presume that the important distinction between this and earlier studies is that the subjects were attending only to one odour, and that the odour was highly familiar. It seems that it was those factors that enabled some people to identify an individual odour in blends of twelve, which is a big improvement on the limit of 4 found in earlier studies.

For now, I shall let you draw your own conclusion on the applicability of these results to wine tasting notes.  I will return to that subject eventually, but in my next post I should like to take a closer look at how wine tasters arrive at long lists of aromas that seemingly break Laing’s four odour limit. In the meantime, if you found this blog post interesting you might also like to take a look at this one on the subject of olfactory white.

[1] Laing, D. G., and Francis, G. W. (1989). The capacity of humans to identify odors in mixtures. Physiology & Behavior, 46(5), 809–814.

[2] Livermore, A., and Laing, D. G. (1996). Influence of training and experience on the perception of multicomponent odor mixtures. Journal of Experimental Psychology Human Perception and Performance, 22(2), 267–277.

[3] Livermore, A. and Laing, D. G. (1998). The influence of odor type on the discrimination and identification of odorants in multicomponent odor mixtures. Physiology and Behavior 65 (2): 311– 320.

[4] Livermore, A. and Laing, D. G. (1998). The influence of chemical complexity on the perception of multicomponent odor mixtures. Perception and Psychophysics, 60 (4): 650– 661.

[5] Jinks, A., and Laing, D. G. (1999). A limit in the processing of components in odour mixtures. Perception, 28(3), 395–404.

How many identifiable aromas in a wine – the dilemma

Tasting notes with flowery language and long lists of descriptors divide opinion: Many wine geeks seem to expect them, and writers oblige, but on the other hand the wine-drinker-in the-street, when not ignoring them completely, will probably dismiss them as nonsense. Personally, I look at them quizzically, and ask myself if they are really communicating anything of value. There are a number of contentious issues in tasting note style and content, but in this and the next few blog posts I want to tackle just one: the number of aromas mentioned.

By aroma I mean something that is detected by the nose. It can be detected either ortho-nasally, by sniffing; or retro-nasally, through a passage between the back of the mouth and the nose. Because retro-nasal detection occurs when food or drink is in the mouth, most people get a strong but false impression that it is the tongue doing the sensing. Examples of aromas are orange, apple, vanilla, chocolate and coffee, as opposed to other non-aroma sensations like sweet, salt, acid and bitter, which are detected by the tongue.

The problem with the number of aromas in some tasting notes is that a series of experiments performed in the 1980s and 90s showed that people are incapable of identifying more than four in a mixture. So how can tasting notes meaningfully refer to more than four? For example, a tasting note taken from International Wine Cellar, lists ten aromas by my counting: fresh fruit aromatics of mandarin orange, black raspberry and grilled watermelon spring from the glass. On the palate, pretty nuances of rose petal, gardenia and oolong tea mingle with herbal notes of sandalwood, star anise, fresh thyme and fennel seed. So what is happening here? Was there a problem with the experiments that were performed? Or, after the first four most prominent aromas mentioned in the tasting note, are we merely reading the results of an overactive imagination?

I shall be weighing the evidence for these alternatives over the next few blog posts, starting with an examination of the science. I’m trying to keep an open mind, but am currently leaning a little in the direction of the scientific research.

The joy of brett

A recent seminar at UC Davis seems to have sparked a little flurry of discussion on brett.  These seem to be five of the PowerPoint presentations used at the seminar: 1, 2, 3, 4, 5.  And here are a couple of interesting follow-up articles I came across on wine-searcher and Palate Press.

I thought the brett debate amongst wine lovers was relatively straightforward.  Some think that no detectable level of brett is acceptable, not only for its unpleasant aromas, but because it suppresses more positive aromas in the wine.  On the other hand, a commonly expressed opinion is that, at low levels, bretty smells add interest to a wine – complexity if you like.  But even those who see low levels of brett as potentially positive seem to implicitly agree that the smells are basically rather unpleasant.

However, findings presented at this seminar (presumably they were known before, but certainly not by me) throw the debate wide open.  It appears that some aromas produced by brett are judged by some people to be unequivocally positive.  These include chili powder, red pepper, black pepper, cardamom, leather, cooked meat, smoked meat, coffee, mocha, graphite and cigar.  I don’t think it is being claimed that these aromas are always due to brett, but they certainly can be.  My guess is that no one really knows how they arise in practice. Another example is the bretty band-aid aroma, which apparently is recognised much more positively as five-spice by those more familiar with Chinese cooking than Western hospitals.

However, even given this positive angle, there is still the issue of how to encourage bretty aromas to develop in the right direction – to improve the wine rather than reduce it only to the smell of shit.  We might well learn how to do this in the winemaking process.  Indeed we must do it already to an extent, even if we do not know exactly how it works.  But after bottling, the conditions to which the wine is subjected are out of the hands of the winemaker.  So unless any remaining brett is zapped with a pre-bottling injection of sulphur, each bottle will develop unpredictably depending on the precise content of the bottle and its environment.

A new Brett Impact Wheel of aromas was presented at the UC Davis seminar, and I am wondering if brett connoisseurship might become one of the next big things in wine, replacing minerality and terroir expression as subject to discuss.  If it does, remember you heard it first here.

Olfactory white, and complexity

For me, as far as winetasting is concerned, the most thought-provoking piece of research in 2012 was published an article where the concept of olfactory white, the smell equivalent of white noise, was mooted.

The researchers mixed unrelated aromas of equal intensity, and presented them to the noses of the participants.  In the words of the authors, ‘the chief finding of this study is illustrated in [the figure reproduced below]: The more components two mixtures have, the more similar they smell, even though they have no individual components in common. Moreover, odorant mixtures with ~30 components or more begin to smell alike, having a quality we call “olfactory white.”’


The usual assumption seems to be that the introduction of new aromas into a wine serves to increase its complexity.  In this article there is at least a hint that additional flavours may serve little purpose, merely taking it one step closer to olfactory white, or at least to the point where it merely has a generic wine smell that is indistinguishable from other wines.

Personally, I actually find that rather satisfying.  When ever I read, usually on the back label, that something, for example a small percentage of grape variety, is used to “add complexity”, my bullshit detectors start twitching.   It is just a pity that all those top Burgundy and Barolo producers never thought of adding a soupçon of Syrah to give their wine more complexity.   Adding Syrah to give insipid Burgundy more oomph is of course a totally different matter!

It also got me thinking about the concept of complexity, which is perhaps all too often taken for granted.  It is one of those winetasting terms that rolls off the tongue and pen a little too easily, and is almost synonymous with good.  Here are a few definitions:

Many-faceted smell and taste. The hallmark of a well-developed fine wine (Michael Broadbent: Winetasting)

Opposite of simple or one-dimensional; a multiplicity, intricacy, nuance of smells, textures. Quality of a high order. (Michael Schuster: Essential Winetasting)

Quantitative/qualitative term referring to the perceptible presence of many aromatic compounds, combining to generate pleasure; a highly desirable attribute. (Ronald Jackson: Wine Tasting – A Professional Handbook)

Refers to the presence of many, distinctive, aromatic elements, rather than one or a few easily recognizable odors. (Ronald Jackson: Wine Tasting – A Professional Handbook)

The final quotation from Jackson is interesting in that it includes the word “distinctive” – preempting the concern that the aromatic elements might be merely participating in a soup where the components cannot be identified.  Because that is a negative thing, isn’t it?  Maybe, but I have certainly seen others suggest that complexity can work at a subconscious level.  That is, you might not detect the oakiness (for example) but the subtle use of oak improves the wine regardless.  There is also the case where all the aromas may not be distinctive at the same time.  Here the aromas you get at any one time depend on many things – like temperature, degree of swirling in the glass, distance of your nose from the wine, etc.  Multiple clearly defined aromas, subtle aromas that work at a subconscious level, and aromas that come and go all sound credible and positive to me.

But soupiness is bad, and I usually associate it with cheap wines.  It is very tempting to associate that soupy quality with olfactory white, or at least a step in that direction.  If so it might be the case that the relationship between the number of aromatic compounds present and complexity is not as straightforward as you might think, and complex wines are actually distinguished by having fewer aromatic compounds.

You might like to take a look at the discussion on Wine Lovers Discussion Group.