In their cultural and artistic creations and practices, human beings grapple with – colours, sounds, aromas, flavours, rhythms, timbres, temperatures and textures. In the production of paintings, musical works or, for that matter, recipes, creators and practitioners select some elements rather than others. They reckon with the physical properties of these elements, as well as with existing principles and conventions for ordering them. Such grappling and reckoning has held much fascination for scientists and social scientists for over a century, from Hermann von Helmholtz in the 19th century and Carl Stumpf at the turn of the 20th century to Max Weber and Lévi-Strauss in the 20th: how are these sounds and those flavours selected ‘among the infinite number no doubt available’ (Lévi-Strauss 1994: 21); are different corpuses of sounds and flavours mobilised in different cultures and periods, and do the cultural possibilities for their combinations recognize some form physical relationship between them?
A recent ‘discovery’ of ‘networks’ of flavours by a team of physicists and computer scientists has brought such questions to the fore, further bolstering the ‘food-pairing hypothesis’ made by a few scientifically minded chefs and food scientists over the last two decades (Ahn et al. 2011). They applied a network-based approach to a corpus of 56,498 recipes (from two repositories in North America and one in Korea) and more specifically to their combinations of ingredients. As the molecular composition of the flavour of ingredients (‘flavour compounds’) is becoming increasingly well known, they were able to identify where ingredients seemed to be paired, when they shared flavour compounds and where there seemed to be an ‘avoidance’ of such relation. Intriguingly they found that food-pairing, i.e. the sharing of flavour molecules, characterised the North American, Western European and to a lesser extent Latin American cuisines, whereas Southern European and East Asian cuisines tended to ‘avoid’ food pairing through flavour sharing. Intrigued by the parallel between this cultural contrast and the comparison between the organisation of sounds in ‘Western’ and ‘other’ musical systems, as laid out by Max Weber in his study of music, I decided to take a closer look.
Obvious objections to the study immediately sprang to mind. Considering the immensity of the range of flavour compounds overall and in each ingredient (a ripe fruit would typically contain several thousands of aroma compounds) and the complexity of the chemical transformations and combinations operated in cooking, one might doubt the status and relevance of the relationships put forward by Ahn and colleagues. However, in the same way as the physical properties of sounds and their relations can be recognised and traced back into the use of the modes and tones in a particular piece of music (though the latter are never reducible to the former), the flavour effect of a recipe may arguably be traced back, without being reduced to, the aroma compounds of its ingredients.
Another objection concerns the quality of the data used, or what is sometimes referred to as ‘dirty data’, namely the lack of completeness and accuracy in large-scale data sets. A study of 4,133 medieval recipes from twenty-five different source texts from England, France, Germany, and Italy, from the years 1300 to 1615 (Varshney et al. 2013), found opposed results regarding flavour pairing of ingredients, depending on the flavour compound database used, thus suggesting that the flavour pairing hypothesis was still inconclusive.
However, I am interested in yet another, more ontological, questioning of the network approach to the analysis of flavours, for networks and nodes, and connections between them, same or different, may not capture human creation in different cultures.
One dimension omitted is hierarchy: should the relations of all known molecules with each other be computed or should the more powerful or defining aroma of some be recognised? Whether flavour sharing or non-sharing takes place between relatively bland aroma molecules or between very ‘suggestive’ ones surely matters. It is very significant that contemporary ‘cutting-edge chefs [may] truly think in terms of pairs or triplets of ingredients when coming up with new recipes’, the most famous one being Heston Blumenthal’s association between caviar and white chocolate (Varshney et al. 2013: 2). However such pairing is likely to hinge on prevalent flavor compounds in the ingredients and take less account of others; such implicit or explicit ordering is missed in an account centered on pairing
Indeed, in the words of food scientist Harold McGee, ‘flavors are something like chemical chords, composite sensations built up from notes provided by different molecules, some of which are found in many foods… Usually just a handful creates the dominant element of an aroma, while the others supply background, supporting, enriching notes’ (McGee 2004: 3, 272). Cuisine, as well as musical composition, has played with these phenomena, which enhance perception. Such hierarchical conception, drawing on the musical metaphor of chords and their harmonics, sits uneasily with a network ontology simply predicated on counting the number of molecules shared by ingredients.
Let’s take the musical comparison beyond a purely metaphoric sense, and imagine what is entailed by a structural analogy between the organization of sounds and that of flavours. A different way of thinking about the sharing of aroma molecules is to compare it to the way in which sound material is shared in musical scales. In the Western harmonic system, related tonalities (i.e. those immediately related through the cycle of fifths) share the same sound material except for one note. This is what the 19th century German physicist Herman von Helmholtz had called the principle of ‘affinity’ (Verwandtschaft). This affinity is due to tonal relations relying on physical properties of sounds. Similarly flavour affinity between ingredients seems to be derived from sharing molecules. In both cases, physical relations are given meaning by a significant cultural ordering. And, importantly, talking about affinity rather than similitude allows for the combination of the notion of connection with those of hierarchy and significance.
The comparison with the organisation of the sound material prompts a further question, namely that of the number of principles presiding over such organisation. Helmholtz had criticised the kind of naturalistic ‘objective’ thinking about sounds developed by music theorists from Rameau to Hugo Riemann, and had shown that the principles governing the musical organisation of sounds could not be simply derived from the physical properties of harmonic consonance. Another principle was also at play: that of melodic distance. However he had only granted that principle a function of passage between two harmonically defined notes. It was Max Weber who suggested that the principle of ‘proximity’, or ‘distance’, (governing notes not through their relation on scales but through their proximity of pitches to the ear) was a fully autonomous one. Thus the organisation of the sound material in all musical systems could be understood as a way of operating through tensions between these two different principles (Darmon 2015).
The computational analysis proposed by Ahn et al. and other studies building on their approach considers only one principle, that of pairing, although this principle can play out positively (connecting sets of ingredients that overlap, more than the random overlap in a random recipe, in the cuisine under consideration) or negatively (connecting sets that overlap less than is randomly the case). In the same way, the Pan-harmonists had considered each note used in a piece through its harmonic provenance, and saw a sequence of notes as harmonically or non-harmonically grounded. Contrast could only be understood in relation to harmony. Following Helmholtz, and above all Weber, however, I suggest that the creation of flavour may well be structured by not one, but two principles, and that contrast should be understood not only negatively, as lack of affinity (less shared compound), but also positively, as distinctiveness. Spices may serve as clarification here.
Spices, as well as some herbs, have considerably simpler compound structures than other ingredients. Although they share many terpenes (e.g. pinenes, citral, linalool, cineole, caryophyllene) and phenolic compounds (eugenol, myristicin), some of them tend to be characterised by one distinctive compound giving them their distinctive quality, such as thymol for thyme, cuminaldehyde for cumin, cinnamaldehyde for cinnamon etc. (McGee 2004: 390-2). Breaking them down into their constituant compounds one both finds a high number of shared compounds between them and with other foods (what McGee calls the ‘bridging’ role of spices); as well as the one highly specific compound. The specific contribution of a spice is interpreted, in the flavour network approach, as minimising the sharing of flavour compounds with other ingredients in the recipe (Jain et al. 2015), i.e. as a contribution to ‘negative food pairing’. It can however also be interpreted in terms of distinctiveness, through their one distinctive compound. How can we compare this distinctive flavour with others if it is just one molecule? The food-pairing hypothesis in the network approach cannot provide any answer here. The principle of affinity cannot apply, only a principle of proximity or distance in aroma, assessed through fine gustatory and olfactory perception – in the same way as melody brings together notes which a trained ear hears as more or less proximate.
Two principles are at play, suggesting that a dynamic analysis of the making of creative decisions might see the process as one of dealing with tensions between principles. This is a very different understanding of cultural creation than that proposed by the network approach. Yet the very intellectually intriguing and aesthetically satisfactory character of the computational approach to flavour compounds need not be lost. Rather, it is worth harnessing it for the service of a more dynamic approach, making space for tension and hierarchy, and the combination of science and the cultivation of practical sense.
Such analysis is also predicated on a different approach to cross-cultural and historical comparison of cultural and artistic practices. Weber suggested that any musical system or genre is organised through the two principles of hierarchy/affinity and proximity/distance, though these could be defined in different ways. My conjecture in starting the research discussed in this article is that affinity/hierarchy and proximity/distance possibly apply also to culinary practices across cultures, thus prompting the more speculative question focusing on whether we should retrieve a dynamic form of structuralism and look for affinity and proximity as key principles structuring human creative practices more generally.
Ahn, Y.Y., Ahnert, S.E., Bagrow, J.P. & Baraba´si, A.L. (2011) Flavor network and the principles of food pairing. Scientific Reports 1, 196.
Darmon, I. (2015). Weber on Music: Approaching Music as a Dynamic Domain of Action and Experience. Cultural Sociology, 9(1): 20–37.
Jain, A. and G. Bagler (2015). Spices form the basis of food pairing in Indian cuisine. arXiv preprint arXiv:1502.03815.
Lévi-Strauss, C. (1994). The raw and the cooked: Introduction to a science of mythology. London: Pimlico.
McGee, H. (2004). On food and cooking: The science and lore of the kitchen. New York: Scribner.
Varshney, K. R., Varshney, L. R., Wang, J., & Myers, D. (2013) Flavor pairing in Medieval European cuisine: A study in cooking with dirty data. arXiv preprint arXiv:1307.7982
Isabelle Darmon is a Lecturer in Sociology in the School of Social and Political Science, University of Edinburgh and a member of the Food Researchers In Edinburgh network. Her interests are broadly in the domain of cultural sociology, sociology of music and art, social studies of food and social studies of the environment. Recent publications include a paper on Weber and music and various articles on the cross-national comparison of food practices.