Two flavour pairing case studies

In previous posts and comments I have suggested that flavour pairings based on key odorants could be explored by looking at odor activity values (= ratio of volatile compound to it’s threshold). If two foods share one or more key odorants, chances are that they will go well together. It is also reasonable to assume that the more key odorants are shared, the more similar the flavours will be and the more likely it is that the foods will blend well and match each other.

Having initiated the TGRWT event I figured I should try to see if there was any OAV data available for coffee, chocolate and garlic. I was lucky to find OAVs for coffee (both arabica and robusta beans) and cocoa. To compare coffee and cocoa I sorted the flavour compounds in a descending order based on the OAV, keeping only the 20 first compounds. I turned out that 7 out of 20 key odorants in coffee and cocoa are shared, corresponding to 28/25% and 39% respectively of the total “odor activity” (= sum of OAV of top 20 odorants). Here’s the whole list:

(I hope the authors stuck to the IUPAC naming conventions as I did not take the time to check if synonyms were present in the compounds lists)

To compare this with a random pairing I search for more OAVs and found data for parmigiano reggiano and mango, so I repeated the excercise. Among the 20 odorants with the highest OAVs respectively for coffee and mango there was no overlap. A neglibile overlap was found between cocoa and mango: one odorant (linalool) was present in both with OAVs corresponding to 0.03% and 0.05% of the “odor activity” respectively. The fact that there is no overlap between coffee or cocoa and mango does not imply that they don’t go well together, only that their key odorants don’t match. Parmigiano reggiano and cocoa however had a lot in common, as seen from the table below. In fact 6 out of 20 key odorants, representing 36% and 89% of the “odor activity” for parmigiano reggiano and cocoa respectively.


The degree of overlap between parmesan and cocoa is in fact better than for coffee and chocolate when judging by the percentages (albeit with one less odorant), so this pairing will certainly be included in a future TGRWT event! A quick google search revealed that chef Masaharu Morimoto has come up with a recipe combining cocoa and parmesan:

Chocolate Carbonara with Parmigiano Reggiano Cream

Chocolate Pasta:
1 pound all-purpose flour
4 eggs
½ cup cocoa powder
1 Tablespoon olive oil

Pasta Sauce:
2 cups cream
4 egg yolks
½ cup sugar
½ cup Parmigiano Reggiano

For the Chocolate Pasta:
Sift flour and cocoa powder together and knead in the eggs and olive oil for 15 minutes. Rest for another fifteen minutes then roll and cut in a pasta machine. Heat up a pot of lightly salted water and boil pasta until al dente.

For the Pasta Sauce:
In a medium sauce pot scald the cream. In a separate bowl, whisk together egg yolks, Parmigiano Reggiano, and sugar. Temper this mixture into the hot cream and bring to a light simmer, whisking constantly to prevent curdling.

Unfortunately I couldn’t find any OAVs for garlic, so I haven’t been able to verify the triple pairing forming the basis for TGRWT #1. The claim was that coffee has dimethyl sulfide in common with garlic, and methyl pyrazine in common with chocolate. The table above confirms that coffee and chocolate have several methyl pyrazines in common, but dimethylsulfide is not among the 20 key odorants in coffee. This puzzles me, but there could of course be other volatile compounds that garlic shares with coffee. There should also be quite a difference between raw garlic (not to mention between whole, crushed and possibly even minced) and roasted garlic. If I overlooked something (or perhaps a paper with OAVs for garlic), please drop me an email about this. The OAVs of garlic could easily be calculated if data on volatile compounds in garlic and threshold concentrations are available.

I did a search on coffee, cocoa and garlic on The Good Scents Company website as described previously and found the following compounds either naturally occuring or used for recreating the aroma of coffee, cocoa and garlic:

  • 5-methyl furfural (found naturally in all three, used for coffee and garlic)
  • benzothiazole (found naturally in cocoa, used in all three)
  • 2-furfuryl mercaptan (found naturally in coffee, used in coffee, garlic cocoa)
  • isovaleraldehyde (found naturally in coffee and cocoa, used in all three)
  • ethyl methyl sulfide (found naturally in coffee and cocoa, used in coffee and garlic)
  • bis(2-methyl-3-furyl) disulfide (used in all three)
  • butyraldehyde (found naturally in all three)
  • S-(methyl thio) butyrate (used in all three)
  • isopropyl mercaptan (found naturally in garlic, used in coffee and cocoa)
  • So there are obviously similarities similarities between coffee, chocolate and garlic, but the question is whether these compounds are key odorants or not.

    It’s only fair enough to add that the concept of odor activity values has it’s limitations. Some are related to matrix effects, because thresholds are not necessarily recorded in a matrix mimicking the food product. Possible synergies between flavour compounds are disregarded (examples are known where sub-threshold concentrations are detected in the presence of other volatile compounds). Also, the underlying assumption that the odor intensity increases linearily is not quite correct. The typical intensity vs. concentration curve is more ‘S’ shaped with an expansive, linear and compressive region as shown below. At low concentrations (expansive region) synergism (also known as hyperadditivity or mutual enhancement) is observed. At high concentrations (compressive region) antagonism (or subadditivity or mutual suppresion) is observed. This means that a high OAV overestimates and a low OAV underestimates the impact of the individual compounds. This also means that the odor activity percentages calculated for the pairings above should be take with a pinch of salt. In between these extremes normal additivities are observed.


    Even though OAVs are not phsychophysical measures of the perceived odor intensity, they compare quite well with models that take different aspects of sensing into accout. The validity of the found OAV can also be tested by a recombination of the flavour compounds to see how good it imitates the original product studied. I can recommend the freely downloadable article “Evaluation of the Key Odorants of Foods by Dilution Experiments, Aroma Models and Omission” (DOI: 10.1093/chemse/26.5.533) for those interested in reading more about the science.

    Despite the drawbacks and limitations I think OAVs can and will be helpful when studying the flavour pairing hypothesis.

    Tips: You can read more about OAVs in books which are (partly) available through Google books.

    Tags: , , , , , , , , , , , ,
    Filed under: flavor pairing, molecular gastronomy, recipe


    1. Fred Says:

      It’s a very intersting subject indeed. Matching products odorant and adding concentration as parameter may then be, one day, the key to find out about new cooking experiences ??

      By the way : Parmeggian and chocolate is actually a great match, have you tried these : it’s awesome.
      If you don’t read french I’ll be glad to translate it for you :)

    2. Martin Lersch Says:

      Would be great if you could translate the recipe!

    3. Trig Says:

      Reading about Masaharu Morimoto it’s obvious that he’s a very creative and talented chef with a great story to tell. His recipe for chocolate pasta sounds very intriguing and one I’d love to cook and eat.

    4. Jurgen Says:

      The central question is this, I think: Why do we percieve a flavor as pleasant?
      My thoughts; Obviously, all food consists of a mixture of flavors. I think we learn that these mixes are pleasant. Children in general prefer bland food. Our taste is in my opinion therefore a psychological phenomenon rather than biology or chemistry.
      If this is true the flavor p[airing hypothesis is right, combining similar flavors will yield dishes that reminds us of things we know. And this will taste well. Strong overwhelming flavors will overrule any similarities of course.
      On the other hand, the biological perception of flavor will not be the same for everybody. As our taste is generated by 350 different olfactory receptors, differences in expression levels of these receptors will give different taste perceptions. Some polymorphisms in taste are well described, bitterness perception for instance.

    5. Fred Says:

      Here’s the recipe translated, have fun !! Fred :)

      30 g parmeggiano (0,5 Oz or 0,33 lb ??)
      100g chocolate (70%)
      1 pinch of Guerande sea salt (or better : fleur de sel or Maldon salt)
      2 or 3 pinch of pepper, freshly ground

      Grate parmeggiano and spread if on a baking sheet. Pre-heat oven to 150°C (300°F) and bake it for 10 minutes until color glodens. Let it cool at room temperature.
      Melt chocolate (if you want to have it shiny and more resistant to temperature modifications you need to follow the temperature curve in order to “temper” it : heat to 55°C then stir to cool the chocolate to 28°C and re-heat it to 31°C but if you don’t care just melt it gently till 55°C max)
      Mix with salt, pepper and parmeggiano. Taste and add more salt/pepper… to your taste.
      Pour on a baking sheet, put another baking sheet on top of it and press it in order to get a thin sheet of chocolate of about 2mm thick. Let it cool until it hardens.
      When the chocolate is hard enough, break it in triangles. You can keep it for 3 days.

    6. Mirko Junge Says:

      You are persuing a great way of analysis. But after having read Luca Turins ‘The Secret of Sense’ you featured a while ago. Your 4 point summary is very much to the point: (point 3 of 4) we smell functional groups. Your more ‘semantic’ analysis of the names of the chemical ‘structures’ seems to be a bit much of a short cut, doesn’t it? Should you not ‘disect’ the compounds into their functional groups and redo the comparision?

      I think this step is warrented, because of your dismissal of the chocolate-mango combination which I think works particuarly well in form of the ‘Chocolate Mouse’ and ‘Mango Puree’ (mango, sugar and mix in a blender). Of cause your dismissal of the chocolate-mango combination could be due to the differences in mangos or just a ‘failure’ to the theory.

    7. Martin Lersch Says:

      Mirko, that’s a very good observation! An interesting approach to flavour pairing would be to look at the vibrations in the aroma molecules. Perhaps one day cooking will be all about pairing the right vibrations (oohh – that sounds very esoteric…). But I think it will be a number of years until we’re there and able to extract any useful information from such an analysis. Too little is known about the vibration theory, how the receptors work and how we actually smell.

      So for the time beeing, flavour pairing based on key odorants will still be helpful when searching for foods that are quite similar from a chemical viewpoint due to key odorants present, but whose similarity many don’t think about or recognize due to our nose’s limited ability to analyze and dissect a smell sensation into it’s individual components

      I should quickly add that a lack of overlap between key odorants does not mean that two foods will not go well together! And I should point out that I did not dismiss the mango-cocoa combination, only state that there is a lack of odour overlap. I’m quite sure that they go very well together!

      All in all, we should not expect that the search for foods with similar key odorants will give result in a large number of new combinations. In fact, the flavour pairing “theory” can only explain a tiny amount of all the flavour combinations that excist. But I still think it’s worth finding those few combinations we otherwise would have missed.

    8. » Blog Archive » TGRWT #5: Chocolate and meat Says:

      [...] and roasted beef. All flavour compounds were ranked and compounds given the same color coding as before. As you can see, there is considerable overlap between chocolate and [...]

    9. » Blog Archive » TGRWT #7: Cocoa frico with cauliflower cream Says:

      [...] I wanted to make a cauliflower cream and serve it with something crispy. Considering the fact that cocoa and parmesan are also a good match I googled for parmesan crisps and found a nice recipe for “frico” – Italian parmesan [...]

    10. » Blog Archive » Foodpairing website launched Says:

      [...] of volatile compounds they have in common, not the actual key odorants. As I have elaborated on previously, pairings like these should preferably be based on odor activity values (OAV). Or to put it [...]

    11. TGRWT #9: Stinkmorchel und ranzige Butter? | molekularkueche Says:

      [...] macht den (ekligen) Geschmack. In geringen Mengen und in der richtigen Kombination ergeben die oben aufgeführten Geruchsstoffe nämlich durchaus angenehme Lebensmittel: Kakao und Parmesankäse. Das klingt doch schon deutlich [...]

    12. » Blog Archive » TGRWT #9: Parmesan and cocoa Says:

      [...] lamiacucina is hosting TGRWT #9 and the foods to pair this time are parmesan and cocoa. I’ve previously blogged about this combination and odor activity values (OAV) are available for both parmesan and cocoa. These are the molecules [...]

    13. » Blog Archive » TGRWT #9: Chocolate tagliatelle with parmesan cream Says:

      [...] Grissini Wrapped in Prosciutto di Parma) by Masaharu Morimoto which I’ve blogged about previously. I was quite intrigued by that recipe and wanted to try it! So here it is, converted to metric [...]

    14. » Blog Archive » Food pairing seminar in Belgium Says:

      [...] interested in the scientific considerations behind flavor pairings can find more info on this in previous posts. The topic however is controversial as the comments to this posts [...]

    15. Fördjupning i Brügge: Teorin bakom flavour pairing — Matmolekyler på Says:

      [...] att det inte bara är förekomsten av aromämnen utan också deras tröskelvärden för detektion, det vill säga deras aktivitet, som utnyttjas. Arrangörerna och konferencieren harangerade snarare än intervjuade de deltagande [...]

    16. TGRWT #17 - the “Days of wine and roses” edition — Matmolekyler på Says:

      [...] of an odorant’s intensity. A high OAV will likely overestimate the impact of an odorant, as explained here by Martin at Khymos – this is well illustrated by the debate on the role of beta-damascenone in wines. Tags: apple, [...]

    17. TGRWT #17 - med tema “Days of wine and roses” — Matmolekyler på Says:

      [...] OAV kommer antagligen överskatta luktämnets inflytande över “slutlukten”, vilket Martin förklarar här på Khymos (på engelska) – ett tydligt exempel är debatten om hur beta-damascenone påverkar aromen hos [...]