Flavor Chemistry and Gastronomy: Sensory, Instrumental, and Multivariate Approaches
By
ARIELLE JURCHAK JOHNSON B.S. (New York University) 2009
DISSERTATION
Submitted in partial satisfaction of the requirements for the degree of
DOCTOR OF PHILOSOPHY
in
Agricultural and Environmental Chemistry
in the
OFFICE OF GRADUATE STUDIES
of the
UNIVERSITY OF CALIFORNIA DAVIS
Approved:
______Susan E. Ebeler, Chair
______Hildegarde Heymann
______Roger B. Boulton
Committee in Charge
2014
i Acknowledgements
I would be remiss not to acknowledge the many people who have made this work possible, more or less chronologically:
My parents, Martha Jurchak and Larry Johnson, for encouraging and pushing my intellectual, scientifc, and creative development from an early age; my brother and sister, Eban and Adriana Johnson; my Grandmother, Marie Jurchak, for planting the seed of bon vivantism and supplying the younger me with plenty of cookbooks to read; to my de-facto in-laws Alice Meadows and Pete, Grace, Jack and Joe Pold for welcoming me into their family. From the early days of what would turn out to be my career, Kent Kirshenbaum, for letting me stand in the back of the room at the frst Experimental Cuisine Collective meeting and giving me the space and support to start developing ideas about how to do scientifc research on culinary questions; and Dave Arnold for taking my half-developed ideas seriously and giving me a lot of challenging questions about smell and chemistry to fgure out. At UC Davis, my advisor, Susan Ebeler, for providing the tools and space to develop as an analytical chemist and learn the hard work of a writing, lecturing scientist, and putting up with my scientifc wanderings; Hildegarde Heymann, for teaching me everything I know about sensory science and experimental design, cultivating my ideas about how science should be dealing with questions of favor, chemistry, and cuisine, and freely ofering profound words of encouragement when I needed them. Carolyn Doyle, for invaluable instrumentation help. My lab and ofce mates, for their support and friendship, Larry Lerno, Helene Hopfer, Ellie King, Maya Hood White and especially Anna Hjelmeland, a stalwart friend both in and outside of work. Roger Boulton, for introducing me to the technology of distilled beverages and being a valuable and enjoyably challenging presence on both my qualifying exam and dissertation committees. Alyson Mitchell and Doug Adams for their presence on my qualifying exam committee and help with academic issues and conference travel. Harold McGee, frst as a role model, then as a QE committee member, and now as a dear friend and colleague.
ii Related to the University of Copenhagen, Line Holler Mielby for being an early scientifc role model and encouraging me to go to Denmark, Michael Bom Frøst, for enabling my coming to the Nordic Food Lab and KU, my sensory panel at KU, and Wender Bredie, Belinda Nielsen, and Charlotte Dandanell of the sensory science group. My collaborators, all of which I thank primarily for giving me interesting and challenging questions (gastronomic and otherwise) to think about-Rene Redzepi for founding the Nordic Food Lab and welcoming me into the noma/mad family; David Chang for inviting me to explain olive phenolics Harvard; Rachel Dutton and Ben Wolfe; Josh Evans, Guillemette Barthouil, and Mark Emil Hermansen with their help making it happen at the Nordic Food Lab; Dan Felder; Rosio Sanchez for giving me interesting things about the psychological, cultural, and hedonic aspects of dessert to think about; Lars Williams, for taking my ideas and expertise seriously, teaching me how to throw a punch, and fundamentally changing the way I think about the production and development of knowledge and ideas.
Paramount among the thanked is Tom Pold, the most encouraging, supportive, intellectually challenging, and wonderful signifcant other I think could possibly exist. Tank you for your patience, your vigilance, your encouragement, your love. I truly could never have done this without you.
iii Contents Abstract 1 Chapter 1: Introduction and Literature Review 3 What is Flavor? 3 Smell and Chemistry 4 Analyzing Flavor 7 Analytical Flavor Chemistry 7 Descriptive Analysis and Sensometrics 11 Flavor and Multivariate Statistics 13 Principal Component Analysis 13 Partial Least Squares Regression 15 Multivariate Statistical Techniques on Non-Continuous or Nontraditional Data 16 Applied Flavor Chemistry: Science, Food, Culture 17 Flavor and chemistry of wine and other beverages 17 Cuisine Research and Development 19 References 27 Chapter 2: Perceptual Characterization and Analysis of Aroma Mixtures using In-Instrument Gas Chromatography Recombination-Olfactometry1 39 Abstract 39 Introduction 40 Materials and Methods 42 Instrument: 42 Sampling and Chromatographic Conditions 43 Sensory Conditions 44 Data Analysis 45 Results and Discussion 45 Conclusions 51 References 52 Acknowledgments 54 Supplementary Information: 55 Chapter 3: Volatile and Sensory Profling of Cocktail Bitters 59 Introduction 59 Materials & Methods 64 Samples 64 Chemical Analysis 64 Sensory Analysis 65 Statistical Analysis 67 Results & Discussion 67 Sensory Analysis 67 Gas Chromatography-Mass Spectrometry 70 Flavor Chemistry of Bitters 77 Conclusions 85 References 87 Chapter 4: GC-Recomposition-Olfactometry (GRO) and multivariate study of three terpenoid compounds in the aroma profle of Angostura bitters 96 iv Introduction 96 Materials and Methods 99 Bitters: 99 Sample Preparation and Extraction: 99 Instrument and Conditions: 100 Sensory Conditions: 101 Statistical Analysis: 101 Results and Discussion: 102 References 110 Chapter 5: Aroma Perception and Chemistry of Bitters in Whiskey Matrices: Modeling the Old- Fashioned Cocktail 113 Introduction 113 Materials and Methods 114 Whiskey: 114 Bitters: 114 Model Old-Fashioned: 114 Sensory Analysis: 116 GC-MS: 117 Analysis of Variance: 118 Means and Signifcant Diferences: 118 Principal Component Analysis (PCA): 118 Partial Least Squares Regression (PLS): 122 Results and Discussion: 122 References 132 Chapter 6: Sensory Attributes and Flavor Chemistry of Acetic Fermentations with Novel Plant Ingredients 136 Introduction 136 Materials and Methods 137 Acetifcation: 137 Juice Vinegars: 138 Tea Vinegars: 138 Wine Vinegars: 139 Sensory Analysis: 139 Volatile Analysis: 140 Chromatographic Conditions: 140 Compound Identifcation and Relative Quantifcation: 142 Capillary Electrophoresis: 142 Statistical Analysis: 142 Results and Discussion 143 Sensory Analysis: 143 Chemical Analysis: 146 PLS 152 References 159 Chapter 7: Correlating Labeled Sorting Sensory Analysis and Volatile Analysis of Malt Vinegars with Novel Ingredients 165 v Introduction: 165 Materials and Methods: 167 Vinegar Production 167 a. Base Beer Fermentation: 168 b. Acetifcation and favoring: 169 Sensory Analysis: 169 GC-MS: 169 Statistical Analysis: 170 a. Sorting: 170 b. Labels: 171 c. Volatiles: 171 d. Multiple Factor Analysis (MFA): 171 Results and Discussion: 171 References 186 Conclusions 191
vi List of Tables and Figures Figure 2.1 Conceptual schematic of the In-Instrument Gas Chromatograph Recombination Olfactometer (GRO) instrument. Volatiles are extracted onto a solid phase (via solid-phase microextraction or SPME) from the headspace of a food, beverage, or other sample, in this case, lavender fowers, and initially they are separated conventionally on an analytical capillary GC column. In-line with the GC column, a pneumatic Deans Switch followed by a cold trap allows the experimenter to build a mixture of these separated volatiles that is held until the cryotrap is rapidly heated, releasing the mixture for a subject to smell at the olfactory port and evaluate. 41 Figure 2.2 Schematic of (a) standard GC-MS; (b) GC-MS with splitter at end of column for olfactometry; and (c) In-instrument Gas chromatograph- Recombination Olfactometer or GRO with Deans switch, splitter, cryogenic trap and olfactory port. Abbreviations: i-inlet; c-column; d-detector; o-oven; olf-olfactometry port; sp-splitter; sw-Deans switch 1; w-waste; cr-cryogenic trap; and cb-switch 2 on control box. 42 Table 2.1 Experimental GC-O conditions and aroma descriptors for mixtures of volatiles from the lavender chromatograms. 43 Figure 2.3 Top aroma descriptors for mixtures of sections of the lavender chromatogram by cut time and chromatogram composition. Abbreviations correspond to Experimental Conditions described in Table 1. As chemical complexity and number of components per mixture approaches the makeup of the whole chromatogram (W) mixture, there is evidence of perceptual additivity as increasing cross-utilization of terms from simpler mixtures, masking as reduced use of dominant terms for simpler (P1-P6) mixtures, and synergistic efects as new complex or composite terms like “fresh lavender” become important. 45 Figure 2.4 Correspondence Analysis of (A) lavender volatile mixtures; and (B) lavender volatile mixture descriptors. Abbreviations for mixtures correspond to those in Table 1. Terms generated by the panelists to describe the perceived odor of from each Experimental Condition described in Table 1 were tabulated by frequency of use and used for the Correspondence Analysis. 30.57% of variance explained by dimension 1 (x), 22.84% of variance explained by dimension 2 (y). 46 Figure 2.5 Te rated representativeness of the aroma of samples W, O1-O3, and P1-P6 as compared by panelists to the aroma of whole fowering lavender. Letters a, b, c refer to the mixture’s Signifcant Diference from each other- if two samples do not share a letter, they are signifcantly diferent. Samples P1, P5, and P6 are signifcantly less representative of the aroma of fowering lavender than sample W, which incorporates all the volatiles in fowering lavender. 48 Supplementary Figure S2.1 Te chromatogram of mixture O2. Compounds eluting between 16 and 25 minutes were vented to waste by the Deans Switch and were consequently excluded from the smelled mixture and not sent to the mass spectrometer. 55 Supplementary Figure S2.2 Te chromatogram of mixture P5. Compounds eluting between 0 and 25 minutes and 32 and 40 minutes were vented to waste by the Deans Switch and were consequently excluded from the smelled mixture and not sent to the mass spectrometer. 55 Supplementary Figure S2.3 Alternate views of correspondence analysis (fgure 4) incorporating the frst 3 dimensions of variation. 30.57% of variance explained by dimension 1 (x), 22.84% of variance explained by dimension 2 (y), 14.03% of variance explained by dimension 3 (z). 56
vii Supplementary Table S2.4 Tentative identifcation of lavender volatile compounds. Volatiles were identifed by matching their mass spectra to the NIST 05 Mass Spectral Library (National Institute of Standards and Technology, Gaithersberg, MD) and to chemical standards, as noted. Te table is divided by cut time for perceptual mixtures P1-P6. 57 Supplementary Table S2.4 continued 58 Table 3.1: Ingredients used in historical and contemporary recipes for bitters, listed by taxonomic name and literature source 61 Table 3.2 Bitters samples used in the study, with historical sources and precedents, and style noted. 63 Table 3.3 Sensory terms and references used in the descriptive analysis on bitters. 66 Table 3.4 Mean values of sensory qualities for each sample 68 Figure 3.1 Principal Component Analytsis (PCA) of bitters. PC 1 (X-axis) explains 40.8% of variance in sensory data; PC 2 (Y-axis) explains 30.2% of variance. Sensory descriptors are in red italicized text. Samples are in bold capital letters, coded by style: citrus in orange, aromatic in purple, tiki in blue, mole in brown, New Orleans-style in dark red, and celery in green. 69 Table 3.5 Compounds identifed by GC-MS in samples of bitters, by name, CAS number, Retention Time (RT), and Calculated (CRI) C8-C20 Retention Index (RI). ni=not identifed 71 Table 3.6 Volatile components reported as literature headspace abundances (% of total peak area) in ingredients used for historical and literature recipes for bitters. Ten most abundant compounds reported where available. T=Tomas 1862; P=Parsons et al. 2011; W=Wondrich, 2007; O=Orange bitters (Regan 2003); J=Jerry Tomas’ Own Decanter Bitters, (Tomas 1862); S=Stoughton’s bitters (Wondrich 2007); B=Boker’s bitters, (Wondrich 2007). 76 Figure 3.2 Plots of Partial Least Squares Regression (PLS) analysis of bitters volatile composition and sensory qualities by descriptive analysis. Sample names from Table 3.2; Sensory Attributes from Table 3.3; Compound Identifcations in Table 3.5. 78 3.2A: Positions of samples 78 3.2B: Biplot of sensory descriptors (red) and volatiles (blue) 78 3.2C: Exploded view of compounds 78 Supplementary Table S3.1 Headspace concentrations of volatiles in bitters, in ug/L 2-undecanone equivalents. ni=not identifed, nd=not detected 91 Figure 4.1 Chromatograms of (top) sample A, the control, uncut sample; and (bottom) sample E, with linalool, terpinyl acetate, and caryophyllene excluded from the reconstitution 99 Table 4.1 Aroma properties of GRO reconsitution mixtures listed by volatiles excluded and calcuated odor activity values (OAV) of excluded compounds. “intensity” is average overall aroma intensity rated by three panelits from 1-10 for each mixture. Aroma descriptors expressed as overall counts for each descriptor for each sample across three panelists and four replicates. 103 Table 4.2 Diferences in aroma qualities of samples with volatiles excluded compared to the control sample. Decreases in descriptor count for experimental conditions highlighted in yellow, increases in descriptor count highlighted in magenta. 104 Figure 4.2 Correspondece analysis biplot of samples (blue circles, blu text) and descriptor counts (red triangles, black text). Dimension 1 (x) Variance explained=37.2%, Dimension 2 (y) variance explained=27.6% 108 Figure 4.3 Multiple Factor Analysis (MFA) individual factor map comparing panelists’ individual map positions with consensus map positions 109
viii Supplementary Table S4.1 PLS1 analysis positions in biplot of compounds and Angostura sample compared to specifc descriptors (positions of descriptors is appoximately 0.2 along each PC), from bitters descriptive analysis data, chapter 3. 112 Table 5.1 Whiskeys and bitters used to prepare the 16 samples in the study 114 Table 5.2 Descriptors used in the sensory analysis and their aroma references. Signifcance by ANOVA denoted by superscripts: b= signifcant by bitters; w=signifcant by whiskey; o, b*w =signifcant bitters x whiskey interaction. References were presented as described, in opaque, lidded, black glasses. 115 Table 5.3 Mean values of descriptors rated in model Old-Fashioneds calculated as signifcant by ANOVA, by sample, bitters, and whiskey. Means by bitters (or whiskey) were only calculated when attribute intensities for model Old-Fashioneds difered signifcantly between the model Old-Fashioneds by bitters (or whiskey) in the 4-way ANOVA. Letters next to mean values are results of Tukey’s HSD test; attribute intensities with diferent letters are signifcantly diferent from each other for the samples studied. 119 Figure 5.1 Principal Component Analysis of model old fashioneds descriptive analysis data. Samples are in bold, all caps; descriptors are in red italic. Top: PC1(31% Variance explained, X-Axis) and PC2 (23% variance explained, Y-Axis); Bottom :PC1 (X-axis) and PC3 (14% Variance explained, y-axis) 121 Figure 5.2 Plots of mean intensities of oak aroma (A) and cola aroma (B), showing signifcant interaction efect between bitters and whiskey. 123 Table 5.4: Compounds identifed in samples, identifed by retention time (RT), Mass Spectra, and Calculated (CRI) and literature C8-C20 retention indices (RI) 124 Figure 5.3 Partial Least Squares Regression Analysis (PLS) plots of model old-fashioned volatile and sensory profles. Variance Explained PC 1: 40% X, 21% Y; PC 2: 21% X, 19% Y; PC 3: 13% X, 14% Y 126 5.3A: Positions of samples, PC1 vs PC2 126 5.3B: Positions of samples, PC2 vs PC3 126 5.3C: PC 1 vs PC 2 plot of sensory descriptors 126 5.3D: PC 1 vs PC 2 plot of headspace volatiles 126 5.3E: PC 2 vs PC 3 plot of sensory descriptors 126 5.3F: PC 2 vs PC 3 plot of headspace volatiles 126 Supplementry Table S5.1 Headspace concentrations in samples (listed by type of bitters and type of whiskey) in ug/L 2-Undecanone equivalents. ni=not identifed; nd=not detected. 134 Table 6.1 Samples used in the study, their substrates, and source of alcohol for acetifcation 138 Table 6.2 Descriptors used in the sensory anaysis, their references, and their signifcance for product by pseudo-mixed model ANOVA 141 Table 6.3 Mean intensities for sensory descriptors for each sample. Signifcantly diferent values determined by a Tukey’s Honest Signifcant Diference test have diferent letter groupings. 144 Figure 6.1 Principal Component Analysis (PCA) biplot of sensory descriptive analysis data. PC1 (X) explains 50.6% of variance, PC2 (Y) explains 22.9% of variance. Samples (scores) are in bold black text, descriptors (loadings) are in red text. 145 Table 6.4 Volatiles identifed in vinegar samples, by chemical name, Chemical Abstracts Service number (CAS), retention time (RT), C8-C20 retention index (CRI), and literature
ix retention index. ni= not identifed. 147 Table 6.5 Organic acids in g/L determined by capillary electrophoresis 149 Figure 6.2 Partial Least Squares Regression (PLS) analysis of sensory and chemical data on vinegar samples. 150 6.2A: Positions of samples 150 6.2B: Biplot of descriptors and compounds 150 6.2C: exploded view of positions of compounds 150 Supplementary Table S6.1 Peak area of volatiles in vinegar samples determined by GC-MS, normalized to 2-undecanone. ni= not identifed. nd=not detected. 163 Table 7.1: Samples used in the study, their base compositions, botanials added, and codes 167 Figure 7.1: Consensus plots from DISTATIS analysis of sorted groups from sensory analysis by 16 panelists on 16 vinegars, with bootstrapped confdence ellipses. 172 7.1A: Dimensions 1(X-axis) and 2(Y-axis) 172 7.1B: Dimensions 1(X-axis) and 3 (Y-axis) 172 Figure 7.2 Correspondence analysis biplot of descriptors used for malt vinegar samples. Dimension 1 Explains 30% of variance, dimension 2 explains 19% of variance. Descriptors are represented by red triangles and samples by blue squares. 174 Table 7.2 Volatiles identifed in malt vinegar samples, their retention times (RT), C8-C20 Kovats retention index (CRI) and literature retention indices (RI), ni=not identifed, tentative identifcations included where possible with unidentifed peaks 176 Figure 7.3 Principal Component Analysis (PCA) score plot of malt vinegars based on volatile composition. PC 1 explains 45% of variance, PC2 explains 21% of variance, PC3 explains 8.5% of variance. 178 7.3A: PCs 2 and 3. 178 7.3B (inset): PCs 1 and 2, included to show overall distribution of scores along PC 1 dominated by one sample. 178 Figure 7.4 Multiple Factor Analysis (MFA) of Sorting, Label, and Volatile data for 16 malt vinegars. 180 7.4A Individual Factor Map shows compromise positions of samples in the consensus space with positional disagreements plotted by dataset. 180 7.4B Factor Map for the Contingency Table shows the positions of samples and labeled sensory descriptors in the consensus space. 180 Supplementary Table S7.1 Relative quantifcation of volatiles in malt vinegar samples, in μg/L 2-undecanone equivalents. ni= not identifed. nd=not detected. 188
x Arielle Jurchak Johnson June 2014 Agricultual and Environmental Chemistry
Flavor Chemistry and Gastronomy: Sensory, Instrumental, and Multivariate Approaches Abstract
One of the subjects in which the currently co-evolving interests in a scientifc understanding of food in a gastronomic context and applying scientifc knowledge and empiricism to culinary development is the development, chemistry, and perception of favor.
Flavor is a complex phenomenon involving chemistry, biology, psychology, memory, and mixing- dependent perceptual efects.
Methods and analytical approaches to characterize the relationship between chemical composition and perceived favor must address this complexity, and the efects of synergy, masking, and gestalt perception in order to produce useful data and knowledge. In this dissertation, these methods include in-instrument reconstitution and omission experiments and multivariate statistical analysis of sensory and volatile data.
Te frst section of this dissertation focuses on directly evaluating mixing-dependent aroma perception efects using a newly developed technique called Gas Chromatography- Recomposition-Olfactometry (GRO) and lavender (Lavandula angustifola) as a model system. Using a GC-MS ftted with with a pneumatic fow switch, cryotrap, and olfactory port, selectively reconstituted mixtures of volatiles from a sample can be prepared and evaluated, omitting one or more compounds and describing the resulting sensory efects.
1 Te second section is a three-part analysis of the favor chemistry of aromatic cocktail bitters, a historically important product made from the extraction of aromatic plants in alcohol with recent resurgence in craf cocktails. Sixteen commercial bitters are analyzed using volatile and sensory profling and multivariate statistics including Principal Component Analysis (PCA) and Partial Least Squares Regression (PLS). Te mixing-related sensory roles of three terpenoid compounds in Angostura bitters are analyzed using GRO. Finally, as bitters are commonly used as a component in mixed drinks, the efect of a whiskey matrix on the favor chemistry of bitters is analyzed using model systems of “Old-Fashioned” cocktails.
Te third section is an analysis of favor and chemistry of vinegars produced by applying acetic fermentation to novel substrates, in the context of culinary research and development. In the frst part of this section, traditional descriptive analysis of vinegars produced from celery juice, strawberry wine, pine needles, and other substrates is correlated to volatile profles using PLS, as in the second section on bitters. Te second part of the section is an analysis of malt-based vinegars with diferent plant and fermented ingredients. A labeled sorting sensory analysis, a rapid method which has previously shown good comparability to descriptive analysis, is analyzed with DISTATIS and correlated to volatiles using Multiple Factor Analysis (MFA) and Correspondence Analysis (CA)
2 Chapter 1: Introduction and Literature Review
Tis dissertation is primarily concerned with two ideas: the complex nature of favor and its relationship to chemistry, and the favor chemistry of gastronomically interesting products and processes. Tese ideas direct the investigation of aromatic bitters, a product with a long cultural history and key use in craf cocktails, and applications of acetic fermentation to novel ingredients in the high-end restaurant Research and Development (R&D) kitchen.
Methodologically, this dissertation uses volatile and sensory profling and multivariate statistics to determine correlative relationships between perceived favor and chemical composition, and it develops new instrumental methods to causatively investigate specifc mixing-dependent sensory-molecular interactions.
Topically, aromatic bitters and acetic fermentation are chosen based on recent interest from restaurants, chefs, bartenders, and the media, and a lack of information on their favor chemistry as they are currently used.
Tis literature review will cover, as background, the psychology, chemistry, and analysis of favor; the history of gastronomy as a research subject in the sciences, and the developing demand for research, science, and technology in cuisine.
What is Flavor?
Flavor is a perceptual construct that primarily combines taste, retronasal olfaction, and oral somatosensation (Small et al 2007) with accessory information from the visual and auditory senses (Verhagen and Engelen 2006). Tis dissertation will focus primarily on aroma as a driver of favor and favor complexity, and a primary source of sensory diferences among samples.
It is important to note that while a food, beverage, or other type of sample may contain 3 molecules that impart favor, these cannot be said to ‘contain’ favor. Rather, these molecules have a bioactivity that stimulates receptors in the sensory system, and favor is the perception that arises from the processing of this stimulus (Auvray and Spence 2008).
Smell and Chemistry
Aroma1 is the perception of molecules that bind with olfactory receptors. Tese receptors are located, in humans, in the olfactory epithelium, an area of tissue located at the top of the nasal cavity. To reach these receptors, a molecule needs to be small (generally under 300 atomic mass units) and non-polar enough to be volatile, i.e. to have signifcant vapor pressure or partition into the gas phase (Rossiter 1996). As a shorthand, these ‘volatile molecules’ will be referred to throughout as ‘volatiles’. Humans possess around 500 types of olfactory receptors, (Axel 2005, Glusman et al 2001, Zozulya et al 2001, Young et al 2002) which enable us to detect tens of thousands, if not more, of diferent molecules. Te olfactory system is able to detect and diferentiate a larger number of volatiles than it has receptors through a process called combinatorial coding (Malnic et al 1999) in which each volatile interacts with and activates a number of receptors to various afnities (Floriano et al 2000). Likewise, each receptor can be activated by several diferent molecules. Each molecule will produce a relatively unique ‘code’ of receptors and levels of activation, and the processing of this set of signals is interpreted in the brain as an odor. When mixtures of volatile compounds are considered, which take on unique odor qualities of their own as discussed below, humans are estimated to be able to discriminate 1 trillion diferent odors (Bushdid et al 2014).
Te ultimate drivers of the molecular feature-perceived odor relationship for a given molecule are not fully understood, however, we do know that features such as alkyl chain length (de Melho Castano Amboni et al 2001); rigidity arising from double bonds, rings and bridged rings (Hong and Corey 2006), aromatic rings (Rossiter 1996), heteroatom functional groups such as alcohols, carbonyl groups, aldehydes, ketones, organic acids, esters, and lactones (Laing et al 1 “Aroma,” “odor,” and “smell” will be considered synonymous for the purposes of this dissertation. Additionally, unless taste or other stimuli are referenced directly in context, “favor” will also refer to aroma. 4 2003, Sanz et al 2008); halides, sulfur, and nitrogen; as well as stereoactivity (Bentley 2006) all have impacts on the perceived aroma quality of a given molecule. Besides aroma quality, these factors may also afect odor threshold (Edwards et al 1991, Mihara et al 1988) (the amount or concentration of a particular molecule required to impart a detectible odor), odor pleasantness, and odor complexity (Kermen et al 2011).
Te foods and food products we use for nourishment and culinary enjoyment typically contain numerous molecules that contribute to aroma. Plant species (lavender [An et al 2001], pine [Domrachev et al 2011], elderfower [Kaack et al 2006]), product that have undergone thermal or other physical or chemical processing (chocolate [Afoakwa et al 2008], cofee [Ribeiro et al 2009]), products arising from the microbial processing of plant or animal species (wine [Ebeler and Torngate 2009], beer), and dishes, ingredients, or other food products created by combining several of these together (whiskey, gin [Pryde et al. 2011, Sanchez 2011]) have dozens or even hundreds of individual volatile components . Essentially, all foods are mixtures2, and so understanding food favor means understanding not only constituent aromas, but also how aroma functions in multicomponent mixtures.
In multicomponent mixtures, there is a complex relationship between chemical composition3 and perceived aroma (Wilson & Stevenson 2006). Understanding the perceived aroma of a mixture in relation to its chemical makeup requires investigating perceptual efects that arise from mixing, such as additivity (Berglund et al 1993), masking (Cain 1975), and emergent or synergistic efects. Perception of the characteristic aroma of a mixture is stored by the brain as a fairly unitary encoded pattern called a ‘smell image’ or gestalt (Livermore and Laing 1998), related to both the original receptor signals stimulated by a sample’s component volatiles and processing-related interactive efects (additivity, masking, synergy, etc). In other words, once a mixture of volatiles becomes familiar as the aroma of a particular entity or object (e.g. the
2 Due to chemical complexity introduced by biochemical processes from plants and animals themselves as well as mi- croorganisms, oxidation, reduction, heat related and other reactions yielding multiple types of molecules. 3 Which can be determined either by making a synthetic model mixture out of pure compounds, or analyzing the volatile composition of an existing or naturally-occurring mixture by Gas Chromatography-Mass Spectrometry [GC-MS]. 5 characteristic mixture of volatiles that makes up chocolate aroma), that object’s aroma is encoded and processed as a discrete entity rather than the conglomerate of the aromas of its component volatiles (Livermore and Laing 1998).To understand the scope of perceptual efects on aroma mixtures, it is illustrative to consider both “top-down” and “bottom-up” approaches. From the top down, the characteristic favor of many products (for example, wine, chocolate, or lavender) is not traceable to any one particular volatile. Rather, it arises from the characteristic mixture of volatiles that typifes that product. From the bottom up, model mixtures with as few as three components have been found to have aroma qualities not attributable to any of the aroma qualities of their three individual component volatiles (Le Berre et al 2008). Terefore, in the context of favor, mixing efects are both widespread, playing a role in many products, and pervasive, coming into play even for simple mixtures.
To expand on the concepts outlined above, the ultimate aroma of a mixture as it relates to chemical makeup may be impacted by additivity, whereby components that are undetectable individually become detectable when smelled together (Ryan et al 2008); by masking, where one component will suppress the aroma or intensity of another (Pineau et al 2008, Preston et al 2008, Hein et al 2009); and by synergy or other emergent qualities, where the perceived aroma (as a gestalt) of the mixture has characteristics not found in any of its components (Le Berre et al 2008). An aroma image or gestalt is encoded and stored in the brain through a combination of analytic processing—where components can be distinguished and do not interact—and synthetic processing—where information about components is lost to the unique and blended qualities of the mixture (Laing and Francis 1989). A mixture of as few as three volatile compounds can have aroma qualities that do not exist in any of the individual compounds when smelled on their own, providing evidence that aroma synergy and synthetic processing require very little in terms of chemical complexity to be signifcant (Le Berre et al 2008). However, a study of mixtures made up of one to fve compounds found that, while mixing impairs the ability of human test subjects to identify the components, these subjects were nonetheless able to name some (up to three) of the components involved (Laing and Francis 1989). A related example of gestalt processing is 6 Livermore and Laing’s 1998 study of multicomponent mixtures of familiar aromas—including chocolate, lavender, and honey—that are themselves mixtures of many odor-active compounds. Mixtures of these familiar smells (for example, a 5-component mixture of chocolate + honey + cheese + lavender + strawberry4) are perceived in very similar ways as mixtures of individual compounds (Livermore and Laing 1998). In both cases, subjects were able to identify some of the odorants present, a maximum of approximately three to four odors in an eight-component mixture. Mixtures of complex aromas (chocolate, lavender, honey), each processed as a gestalt, are perceived through a combination of analytic processing—as some component gestalts can be identifed—and synthetic processing—as some get lost to the aroma background or to interactive perceptions (Jinks and Laing 1999, Barkat et al 2011). Te association of a mixture of signals into a single gestalt is dependent on experience and learned associations (Auvray and Spence 2008).
Analyzing Flavor
Analytical Flavor Chemistry
Picking the right tools for analyzing favor, and designing experiments that can actually answer the questions we want to ask about the relationship between perception and chemistry depends on understanding as fully as possible the nature of the interconnected systems that contribute to favor. As we know that volatile favor compounds interact in complex ways to determine aroma, we need methods that address mixing and interactions in order to fully understand the relationships between sensory and instrumental data on perceived favor and the molecules that cause it. Analytical favor chemistry necessarily deals with two questions and therefore tool sets – the chemical nature of the product, and the perceived favor of the product.
Identifying the chemical components potentially contributing to the aroma of a sample requires several steps. Generally, it is necessary to separate the volatile components from the non- 4 Which is processed perceptually as a 2-component mixture (of two odors) because of gestalt processing, but is chemically speaking a highly multicomponent mixture of volatiles from chocolate, honey, cheese, lavender, and strawberry 7 volatile components or matrix while preserving their relative composition through some form of extraction. Next, chromatographic separation of this mixture of volatiles is performed, and the separated components are analyzed qualitatively, semi-quantitatively, or quantitatively by a detector.
Extraction methods used for favor analysis include static headspace, dynamic headspace/ purge and trap (Kanavouras et al 2005, Ferraces-Casais et al 2013), solvent extraction (Caldeira et a 2007), solid-phase extraction (SPE) (Lopez et al 2002), simultaneous distillation-extraction (SDE) (Lee and Ahn 2009), solvent-assisted favor evaporation (SAFE) (Willner et al 2013, Engel et al 1999), and sorptive methods such as solid-phase microextraction (SPME) and stir bar sorptive extraction (SBSE or Twister) (Fan et al 2011, Bicchi et al 2002). Headspace-Solid-Phase- Microextraction or HS-SPME, in addition to being solventless and having a low margin for error (Wardencki et al 2004), has been shown to create extracts that are particularly representative of the aromas of the samples on which the extraction has been performed (Poinot et al 2004, Aceña et al 2010), and this method has been employed extensively in the literature and in this dissertation.
Gas chromatography (GC) has been available since the 1950’s as a method for separating mixtures of volatiles for individual detection (Ebeler and Torngate 2009). In this method, a fused silica capillary with internal diameter typically between 0.15-0.5mm and length typically between 15-60 meters has a stationary phase bonded or coated to its inner walls in a layer usually less than one micron thick. Volatilized analytes are introduced into the capillary or ‘column’ in a stream of gas, typically helium, hydrogen, or nitrogen, that also acts as the mobile phase. Analytes are separated by their afnity to the stationary phase, where those analystes that experience more, or stronger, interactions with the stationary phase are retained longer and therefore elute from the column later, and those with fewer or weaker interactions are less retained by the stationary phase and elute earlier. Te chosen column stationary phase chemistry can vary and is usually chosen to best separate the analytes of interest; dimethylpolysiloxane is a very common stationary phase
8 for aroma work, either in its pure form or 5% substituted with phenyl groups, as is polyethylene glycol and these phases were used for the work covered in this dissertation.
Once the volatiles are separated by the column, they can be detected individually. Detectors commonly coupled to GC include fame ionization detection (FID), heteroatom- specifc detectors such as the Nitrogen-Phosphorus detector (NPD) or sulfur chemiluminescence detector (SCD), or various types of mass spectrometer (MS)—single-quadrupole (employed in this dissertation) as well as tandem methods such as triple-quadrupole (QQQ) and quadrupole- time-of-fight (Q-TOF). GC-MS has the beneft of being applicable to both quantitative and semi-quantitative analysis, as well as providing structural information about each analyte through comparison to the mass spectra of lab-run standards or those deposited in libraries such as the NIST mass spectral database.
HS-SPME-GC-MS allows for the determination of the volatile profle of a sample—a chromatogram showing the volatiles it contains separated in time and with relative concentrations refected by peak area that can be tabulated for further analysis. Te practice of using a human volunteer as an “odor detector” is referred to as GC-Olfactometry or GC-O (Grosch 2001). In this method, part or all of the stream of volatilized compounds eluting of the column are diverted from the instrumental detector (e.g. MS or FID) to a port or cone from which the subject snifs over the course of the chromatographic run and records their perceptions of the aroma quality, and sometimes intensity, as it changes over time. By comparing these perceptions in time to the elution times (as determined instrumentally) of the volatiles, the individual odor contributions of each volatile in a sample can be analyzed.
Commonly, a favor-chemical study of a product seeks to quantify the impact or contribution of each volatile compound in a sample to the sample’s overall aroma. One benchmark for the importance of an individual volatile compound in a sample is the odor activity value (OAV), defned (Patton and Josephson 1957) as the concentration of the compound divided by its experimentally determined odor detection threshold concentration in a model matrix or
9 solution. Any compound found in concentrations “above threshold” will have OAV>1, and any compounds “below threshold” will have OAV<1. A second benchmark for volatile importance uses GC-O in a methods such as Charm (Acree and Barnard 1984)5 or Aroma Extract Dilution Analysis (AEDA) (Grosch 1993); in the latter, an extract is successively diluted and olfactometry is performed on the GC-separated eluant of each diluted extract. Each compound or peak is then assigned a “dilution factor” value, the number of N-fold dilutions required to suppress its detectability by olfactometry. Te more dilutions required, the higher the dilution factor for an odor compound and the more important that compound is considered to the aroma of the sample.
OAV and/or dilution factor in these types of analytical studies are ofen used as a criterion for building reconstitution and omission experiments. In reconstitution and omission experiments, a model matrix is used to create a reconstitution of the volatiles in the sample. Each volatile is quantifed in the original sample, and dosed at this measured concentration in the model from a stock of purchased or synthesized pure compound. Commonly, a sensory- activity cutof is used to eliminate compound from the model reconstitution; for example, only compounds with OAV or dilution factor above a certain value may be included. Preparing the same reconstitution several times, each time omitting a diferent compound, allows for comparison of the omission models with the full reconstitution. Te underlying logic is that if the mixture omitting compound ‘x’ is detectably diferent from the full reconstitution, then compound ‘x’ is important for the aroma of the original sample. Tere are several limitations with this design – it requires every compound in a sample to be above the limit of quantitation for the instrumental system in use, to be available as a pure standard, and for the matrix used for threshold determinations and reconstitution to model the sample well. Furthermore, compounds in concentrations below their sensory detection thresholds may have a signifcant role in aroma perception through interactive efects, and examination of more than one sample, or any interactive efects within a sample, may extend the experimental design to a ponderous
5 Charm refers to an aroma quality and intensity at a given retention index 10 size6. Te existing instrumental and disciplinary toolset for analyzing favor chemistry causatively is undesirably reductive, given its inability to adequately address known mixing and interactive efects, and is hindered by the artifcial constraints of the OAV, thresholds, dilution factor, and model reconstitutions (Ryan et al 2008). Flexible, facile methods for producing representative reconstitutions will help bridge this gap between theory and practice in favor chemistry, and one such method is discussed in chapter 2.
Olfactometry-based approaches consider favor chemistry by investigating the aroma roles of compounds on an individual basis, ofen on one sample or a small set of samples. For analyzing perceived favor in a class or larger set of samples (for example, Malbec wines from diferent regions, or commercially available gins) and its relationship to chemistry, an approach of sensory profling followed by multivariate statistical analysis, ofen in tandem with data on chemical composition, is frequently used.
Descriptive Analysis and Sensometrics
Sensory descriptive analysis is a technique for generating quantitative data on the sensory characteristics of foods or other samples (Lawless and Heymann 2010). Since sensory qualities7 cannot be directly measured instrumentally, systematic evaluation of these attributes by human volunteers yielding numerical data allows for analysis that would otherwise be impossible.
Descriptive analysis (DA) was frst developed in the late 1940’s as a technique called
“Flavor Profle®” (Cairncross and Sjostrom 1950) and has evolved across a number of formats, proprietary and otherwise, over the decades since (Stone et al 1974, Brandt et al 1963, Civille and Lyon 1996, Powers 1988, Murray et al 2001). Generically speaking (Amerine et al 1965, Heymann et al 1993, Stone and Sidel 1974), DA uses a panel of judges to quantitatively rate the intensity
6 For example, in a sample with 10 volatile compounds with OAV>1, an experiment exam- ining only single-compound contributions would require the preparation of 11 reconstitutions (full model
+ 10 full-minus-1 models); examining 2-component interactions would require 10C8 = 45 additional reconstitutions. A reconstitution-based study of 3 such products in a category with 2-component interac- tions being considered would require 3(11+45)=168 reconstitutions. 7 For example aroma (generally) or astringency, pear-like smell, or sourness (to give just a few specifc examples) 11 of one or more sensory attributes in a product or set of products. Tese panelists are trained with (ideally) physical references for each attribute they are asked to rate, so that they agree on the meaning for each term used to describe an attribute and this meaning is grounded in a real object (Giboreau et al 2007). Where physical references are not available, specifc defnitions are agreed-upon by all panelists for each attribute (Lawless and Heymann 2010). For example, using the verbal concept of “a damp basement” to defne “mustiness” when it was not practical to have panelists visit a damp basement each time they were asked to rate mustiness (Wood et al 1995).
It is typical in a descriptive analysis to have the panelists generate the terms that they, as a group, feel are important for characterizing and distinguishing the samples they are to analyze. Tis is usually accomplished through balloting, or by consensus, where panelists in groups smell, taste, or otherwise sensorially interact with the samples a few at a time, and generate sensory terms for each, which are collected by a panel leader. Te panel leader will then make references for each term, which the panelists will test and either accept or tell the panel leader how or why the reference doesn’t match the sensory attribute. Tis is repeated over several sessions, with the panel agreeing to eliminate redundant or unimportant attributes, until a refned list of attributes and agreed-upon references for each have been generated.
During DA, panelists typically rate attributes by selecting a point along an unstructured line scale with labeled ends (e.g. “very low” to “very high” or “not present” to “most intense”); this has an advantage over rating with integers (for example 0-9 or 1-10) because it results in truly continuous data that can be analyzed with a variety of statistical techniques. One goal of DA is to determine which attributes are most important for describing how samples in a set difer from one another. For this purpose, a univariate Analysis of Variance (ANOVA) on each descriptor will identify those descriptors which difer signifcantly in intensity between the samples (by comparing variance that can be identifed as coming from samples to variance which derives from random error). Since the sensory attributes of foods, beverages, and other similar products tend to be complex, DA ofen results in data on many attributes whose explication and analysis
12 call for the use of multivariate statistical techniques. Whereas univariate methods can describe the relationship between one dependent variable and (sometimes) several independent variables, multivariate methods can simultaneously examine multiple dependent variables (e.g. sensory characteristics) and the relationships among them.
Flavor and Multivariate Statistics
Principal Component Analysis
Principal Component Analysis (PCA) is a method for simplifying a dataset with minimal information loss, wherein multiple dependent variables (e.g. sensory descriptors) each describe multiple samples or objects. PCA identifes the main interrelationships among samples, among dependent variables, and among samples and dependent variables. Formalized in 1933, it is possibly the oldest multivariate data analysis technique, and is used in one form or another in most scientifc disciplines (Abdi and Williams 2010). In the context of favor research, PCA is usually performed on sensory data, but it can also be used to examine latent relationships among other variables, such as volatile profles (Wang et al 2009).
Descriptive analysis performed on a set of X samples with N descriptors results in a dataset that can be thought of as an N-dimensional cloud of data—if every descriptor is an axis, and the intensity of each descriptor has been rated in each samples, then each of these ratings can be thought of as describing a distance along this axis, and each product8 has a position defned by its coordinates on each of the N axes. In a descriptive analysis experiment with ratings on thirty sensory qualities (which is fairly typical), this representation of the sensory qualities of each sample will be a thirty-dimensional “cloud” with each sample as a point with a defned position on each of thirty axes, each of which are orthogonal to the others.
A Principal Component Analysis (PCA) is a dimensionality-reducing statistical technique
8 Either each product as rated by one panelist during one replicate, or the mean value of all of these ratings, of which there is (# of panelists)*(# of replicates) for each product 13 that identifes latent correlations (“Principal Components”) between axes in a multivariate dataset, and uses these to reduce the dimensionality of the dataset while preserving as much variance as possible. Tis can aid interpretation of the descriptive analysis data “cloud” mentioned above by spatially describing relationships between many products and descriptors in two or three dimensions instead of thirty. Te graphical output of a principal component analysis on a sensory dataset is a space where the relative positions of samples and descriptors conveys information about their interrelationships. Tis is usually presented as a two-dimensional biplot (though higher dimensions can be plotted and are ofen useful to examine), showing the positions of scores (samples) and loadings (descriptors) in relation to the frst two principal components.
Loadings are typically plotted as vectors, and the greater the magnitude of a vector, the more variance there is in the dataset for that descriptor. A loading with a very long vector can be interpreted as having a greater importance for determining diferences in the dataset. Te smaller the angle of a loading vector to a principal component, the more that principal component refects that particular descriptor. Between multiple loadings, the closer the angle between their loadings is to zero, the more highly correlated they are to each other; a 90-degree angle between loadings means those descriptors are uncorrelated in relation to the plotted principal components, and an angle close to 180 degrees refects a negative correlation between descriptors.
Te positions of samples in the biplot refects how close or diferent each sample is from a sensory ‘average’ score, by its overall distance from the center, and how much of its sensory qualities are explained by each PC, i.e., by its x- and y- coordinates. Samples with scores plot that are close to each other are more similar, overall, than samples with scores plot that are far apart from each other. In addition, samples that plot closer to a particular loading vector are more highly correlated to that sensory quality than samples plotting away from that vector. In this way, a PCA allows for interpretations of latent correlations or similarities between samples and their descriptors overall, as well as between descriptors, between samples, and between specifc descriptors and samples in a holistic, multivariate context. Te validity of this interpretation
14 depends on the amount of variance explained by the retained PCs, and if a fairly low amount of variance is explained by the frst two PCs, or if the third or higher PCs explain a similar amount of variance as the frst or second PCs, it may be necessary to retain, replot, or at least address diferences in interpretation introduced by these PCs (Heymann and Lawless 2010).
Partial Least Squares Regression
Partial Least Squares (PLS, also called Projection onto Latent Spaces) is a form of data analysis that allows for the simultaneous modeling and comparison of relationships among multiple independent and multiple dependent variables. PLS borrows features from both PCA and multiple linear regression with a goal of predicting as much of the dependent dataset as possible given a dataset of independent variables, based on their common structure (Abdi 2010).
As a variance-based form of structural equation modeling, PLS focuses on capturing as much variance in multiple dependent variables that can be explained by variance in multiple independent variables (Heinlein and Kaplan, 2004). PLS has been shown to work robustly on data with high multicolinearity of dependent variables (Cassel et al 1999). Both of these make PLS a useful analytical tool for studying the relationships between (dependent) sensory characteristics and the (independent) instrumentally-measurable chemical and physical properties that provide the stimuli for those sensory characteristics. PLS has been used to predict sensory characteristics of Chardonnay and aged red wines from their volatile profles (Noble and Ebeler 2002, Lee and Noble 2006, Aznar et al 2003), predict the geographic origin of wines from selected chemical parameters (Capron et al 2007), predict the sensory qualities of cooked salmon from the sensory qualities of raw salmon (Rodbotten et al 2009), and to diferentiate between “stuck” or sluggish wine fermentations and fermentations proceeding normally using headspace volatiles (Malherbe et al 2009). In a favor-chemical study, the goal is to understand the molecular drivers of favor, so chemical diferences not correlated to sensory diferences are not useful for analysis. Multicolinearity in sensory data could come from certain diferent favor attributes arising in parallel from the same process, and from individual compounds contributing to more than one aroma quality. 15 It may be useful to distinguish between two approaches to using chemical data in a multivariate analysis. In a chemometric approach to analyzing, say, GC-MS data, one would seek to retain as much chemical variance as possible in distinguishing among samples. PLS maximizes the variance explained in the dependent (Y) rather than the independent (X) dataset, so with sensory (Y) and chemical (X) data, less variance in the chemical dataset may be retained than in a simple PCA of the chemical dataset, but the variance that is retained explains more of the variance in the sensory dataset.9
Multivariate Statistical Techniques on Non-Continuous or Nontraditional Data
In some cases, useful data can be collected from sensory experiments that don’t follow a typical descriptive analysis procedure, requiring the use of multivariate statistical analysis techniques beyond the PCA and the PLS. Verbal methods (Valentin et al 2012) such as check- all-that-apply (CATA) and free choice (FCP) profles have each panelist describe samples by the presence or absence (rather than intensity) of attributes, which are either from a master list (CATA) or are panelist-generated (FCP). Using citation frequency as a measure of the importance of sensory descriptors rather than variations in intensity necessitates analytical methods which deal with categorical rather than continuous data, such as a correspondence analysis. Similarity- based sensory methods such as the sorting task (Chollet et al 2011, Lelievre et al 2008) ask panelists to group samples based on their similarity to each other and don’t necessarily use a descriptive component. Methods such as DISTATIS (Abdi et al 2005, 2007) can then be used to compare latent similarities in individual distance matrices to generate a consensus map that can be interpreted much like a PCA plot of scores. Rapid profling data from FCP, CATA, Sorting, Projective Mapping and other types of sensory analysis can be co-analyzed with descriptive or instrumental data using a multiple factor analysis (MFA) (Chollet et al 2011, Abdi 2003), allowing the latent similarities in several data sets to be visualized and interpreted simultaneously. Specifc details are described more fully in pertinent following chapters.
9 So, the ultimate question is, are we trying to model chemical variance or use chemistry to explain sensory variance? In the case of this dissertation it is generally the latter. 16 Applied Flavor Chemistry: Science, Food, Culture
Flavor and chemistry of wine and other beverages
Studies on the relationship between chemistry, especially volatile profles, and sensory perception of plants and plant products are ofen undertaken or published under the guise or purview of the favor and fragrance industries, with some intersection with plant sciences, entomology, and other felds. Agricultural and food products that have undergone some kind of human-directed processing and that have economic importance are also common subjects for sensory and/or favor chemical analysis; for example cheddar cheeses, sherry vinegars, wines from diverse varietals and regions, aged buckwheat vinegar (Aili et al 2011), huitlacoche (Lizarraga- guerra et al 1997), qu (rice wine) (Mo et al 2010), molded-surface ripened cheeses (Martin et al 2001), and others. Tese favor chemistry studies of foods, in characterizing an economically or culturally important product in terms of its sensory and chemical qualities, provide data in a form that can be communicated objectively10 and publicly11; that can be used for quality and process optimization; and that enables further generation of knowledge through comparison to other analyses on similar or related products.
Sensory and volatile analyses have been applied extensively to a wide range of beverages, including wine, beer, whisk(e)y—Scotch, Bourbon, etc.—tequila, rum, mezcal, gin, and others. Despite the long history of alcoholic beverages as part of human culture, many of these products have at best been investigated only preliminarily. Artisanal and traditional distilled beverages, and alcoholic beverages incorporating plants, such as gin, bitters, amaro, chartreuse, or absinthe, have long histories, diverse formulations, highly developed cultural signifcances, and complex favors (Tonutti et al 2010), but their favor chemistries have not been examined in-depth.
Wine is probably the most well-studied of any alcoholic beverage, and has been analyzed, extensively for all of the outlined purposes. Early wine favor chemistry studies (in the 19th
10 In the sense that described sensory qualities are tied to real references, and chemical profles are instrumentally measurable qualities 11 In the sense that published data, even if paywalled, is not secret 17 to mid-20th century) (Ebeler and Torngate 2009) focused on characterizing major chemical components—alcohol, tannin, organic acids, sugars—as well as categorizing and attempting to prevent common chemical and microbial defects (Polaskova et al 2008). As analytical methods have become more sensitive and sophisticated, and computing power has increased, emphasis has shifed from defects to characterizing the fundamentals of wine favor chemistry and perception (Hjelmeland et al 2013, Hein et al 2009, Heymann et al 2013, King et al 2013), and the favor chemistry of high-quality wines. Tis includes identifying the many hundreds to thousands of volatile and non-volatile compounds present in wines that afect their favors, tracing the relationship between these compounds and grape, microbial, or oak metabolism and chemical or physical interactions, and characterizing their sensory interactions which result in perceived favor.
Sensory and favor-chemical studies of wine have been followed by the development of research that has explicitly addressed wine not just as a complex assemblage of intersecting chemical, biological, and sensory qualities, but as a cultural entity. Considering wine in isolation from all other factors ignores the fact that perception of favor during consumption is, in everyday life, an experience mediated by culture, and that culturally, wine is frequently consumed as an accompaniment to food. Terefore, a scientifc understanding of the sensory qualities of wine is incomplete until it includes an understanding of how these sensory qualities, and changes in these qualities, are related to the various culturally-important ways and contexts in which wine is experienced by humans.
One example of this is the sensory analysis of wine and food pairing. Tis practice is subject to extensive attention in the popular press, is the primary task of sommeliers, and is a defning component of wine economics, however, until recently it received very little attention from sensory descriptive analysis research. Preference for one of six diferent wines paired with a chicken dish and a wild boar dish was evaluated in a 1997 Swedish study involving 220 subjects (Nilsson et al 1997), and the frst quantitative study focusing on perceptual changes induced by
18 food-wine interactions focused on Chardonnay wines and Hollandaise sauces. Wine and cheese pairing has been studied with white wine and blue mold cheese (Nygren et al 2002), red wine and eight diferent cheeses of diverse styles (Madrigal-Galan and Heymann 2006), and a mix of red and white wines with diferent cheeses (Harrington and Hammond 2006). A 2010 study of specifc pairings of ultra-premium wines ($25-$80 bottle price) and artisanal cheeses, with and without additional food accompaniments (such as chutney or nuts) had panelists rate “match level” of favor intensity and body, but did not look at specifc favor attributes (Harrington et al 2010).
Food, is a signifcant intersection between physical qualities, sensation, and culture. An academic understanding of this continues to develop, as does a parallel appreciation by the crafspeople who produce these foods. Besides wine pairings, sensory and hedonic qualities of other food combinations have received attention, including pork and vegetable accompaniments (Aaskyn et al 2010), olive oil with other foods (Cerretani et al 2007), salmon and sauces (Paulsen et al 2012, 2013), chocolate and beverages (Donadini et al 2012), and beer and cheese (Donadini et al 2013). An analysis of beef stock production with wine as an ingredient found that the use of chemically and sensorially diferent wines correlated to chemically and sensorially diferent stocks and reductions, though how diferences in wines led to these diferences in stocks was not clear (Snitkjaer et al 2011). Tis latter study intersects wine chemistry with the nascent academic discipline of “molecular gastronomy”, discussed further below.
Cuisine Research and Development
Beyond wine, explorations of food as an intersection of physical and chemical properties, human perception, and culture, have occurred in a number of guises and forms. Food science and technology is ofen defned by a research focus benefting the interests and interrogating the issues facing the food processing industry (see, for example, the homepages of the food science departments at UC Davis or Perdue University). Interest in the science behind cuisine or “the art 19 of cookery” has developed parallel to these concerns. In 1794 Sir Benjamin Tompson, Count Rumford wrote (Tompson, 1794):
Te advantage that would result from an application of the late brilliant discoveries in philosophical chemistry and other branches of natural philosophy and mechanics to the improvement of the art of cookery are so evident that I cannot help fattering myself that we shall soon see some enlightened and liberal-minded person of the profession to take up the matter in earnest and give it a thoroughly scientifc investigation. In what art or science could improvements be made that would more powerfully contribute to increase the comforts and enjoyments of mankind?
Tis early proposal of turning science towards the study of food as a product of enjoyment did not take root much further until one hundred and seventy-fve years later, when in a presentation to the Royal Society entitled “Te Physicist in the Kitchen”, the physicist Nicholas Kurti echoed this concept, remarking “I think it is a sad refection on our civilization that while we can and do measure the temperature in the atmosphere of Venus we do not know what goes on inside our soufés” (McGee 1999). Te year afer the publication of his seminal On Food and Cooking (McGee 1984, 2004), food science writer Harold McGee argued for “considering cookery as a discipline with a scientifc base” as “the perspective ofered by science enriches the experiences of cooking and eating” (McGee 1986). McGee and Kurti, along with the cookbook writer Elizabeth Cawdry Tomas and physical chemist Herve Tis, went on to organize a series of seminars at the Ettore Majorana Centre for Scientifc Culture in Erice, Sicily on “Molecular and Physical Gastronomy,” later renamed “Molecular Gastronomy,” starting in 1992 (Arnold 2006). Te latter term was also used in several books by Tis (Tis and DeBevoise 2013, Tis 2013) and became the colloquial umbrella name for research focusing on cooking or cuisine (Mielby et al 2010, Barham et al 2010, Risbo et al 2013, Snitkjaer 2010, Snitkjaer 2011). Academic “molecular gastronomy”
20 has examined the chemical and sensory aspects of beef stock cooking (both with (Snitkjaer 2011) and without (Snitkjaer 2010) the addition of red wine), as well as the efects of expectation and surprise in diners’ perceptions of during an ultra-high-end, avant-garde meal (Mielby 2010).
Te evolution of molecular gastronomy as a research discipline has run parallel to the development of modernist and experimental approaches to cooking and ideas about food at infuential restaurants. In a study of how priming by waitstaf afected diner’s perceptions of an haute cuisine meal molecular gastronomy is referred to as “the scientifc study of deliciousness” as well as a “culinary trend”, highlighting some controversy about the diferences between popular food trends or movements and their relationships to science (Mielby et al. 2010). Tis was highlighted in a 2006 “Statement on the new cookery” signed by McGee along with chefs Ferran Adria, Heston Blumenthal, Tomas Keller, and others- all of whom had gained prominence by way of the San Pellegrino World’s 50 Best Restaurants list and signifcant media attention as part of the experimental culinary avant-garde. Te Statement argued that “the term ‘molecular gastronomy’ does not describe our cooking, or any style of cooking” (Adria et al 2006). Te “New Cookery” that the statement references, also called “Modernist Cuisine”, is widely agreed to have its roots in Adria’s restaurant, elBulli, beginning in the late 1980’s (Myhrvold 2011). Chefs identifed or labeled as “modernist” today are heavily represented among what are widely considered to be the best restaurants in the world (Edelstein 2013, Smith 2013). While they have extremely diverse styles, generally speaking “Modernist Cuisine” describes a culinary modernism, which rejects culinary rules in favor of creatively re-defning cuisine and cooking; experimentation to develop new techniques, technology and ideas; new uses for ingredients and entirely new ingredients; and the perfection of recipes by way of borrowing from, collaborating with, and performing scientifc research to better understand food functionality and enable development, the informed rejection of tradition, and new culinary possibilities (Myhrvold et al 2011, Blumenthal 2006, Keller 2008, Atala 2013, Aduriz 2012, Smith 2013, Kamozawa and Talbot 2010). Te same statement by Adria et al that criticized the use of the term “molecular gastronomy” to describe cuisine went on to say that “a spirit of collaboration and sharing is 21 essential to true progress […] to explore the full potential of food and cooking, we collaborate with scientists, from food chemists to psychologists” (Adria et al 2006).
R&D activity has played an increasingly important role at the restaurants of these and similarly avant garde chefs, both internally and via collaboration with scientists working in industry and academia, with innovation becoming a marker and driver of quality in high-end restaurants (Tan 2013). Blumenthal’s restaurant Te Fat Duck has had a dedicated experimental kitchen (separate from the service kitchen which prepares food for diners) since 2004 (Jonny Lake, Fat Duck head of R&D, personal communication, Mar 26 2014) , and elBulli opened a physically separate R&D facility called El Taller in 1997. El Taller was active for six months of each year (with the restaurant operating for the other six) until the restaurant closed, permanently, in 2011, to become a dedicated research foundation devoted to food (Andrews 2011). Te Basque restaurant Mugaritz, headed by chef Andoni Luiz Aduriz, has a formal relationship and full R&D kitchen within the food science research center Azti-Tecnalia in Bilbao, Spain (Aduriz 2012). Similarly, the Danish chef Rene Redzepi, of restaurant Noma in Copenhagen, founded the Nordic Food Lab in 2008 (Mouritsen 2013) and also maintains a dedicated test kitchen (Redzepi 2013), and the Momofuku group of restaurants, centered in New York City, established a culinary lab12 in 2010.
Te introductory notes of papers on favor chemistry frequently make mention of the economic and/or cultural importance of the foodstuf being analyzed. Besides having an important place within the restaurant world, experimental haute cuisine, with an emphasis on empiricism and R&D, is increasingly an important part of the public consciousness. Chef Rene Redzepi has appeared on the cover of the international edition of Time magazine twice in the last two years (March 26, 2012 and November 18, 2013). On the latter, he appeared with chefs David
12 It should be noted for clarity that these facilities aren’t strictly laboratories in the academic sense of the term, functioning more like development kitchens with a heavy research component. However, many of them disseminate research fndings through talks, lectures, and peer-reviewed publications, and many have actively collaborated with scientists and other aca- demics, as addressed below. 22 Chang and Alex Atala, who have both published peer-reviewed work on applications of botany, microbiology, and chemistry to cuisine, as have Aduriz (Arboleya et al 2008) and Blumenthal (Oruna-Concha et al 2007). In addition to formal papers, many of these chefs have been active in, and received considerable attention for, engaging in research independent of their restaurants (Redzepi’s Nordic Food Lab and Alex Atala’s ATA group), sharing of results via Internet platforms (Williams 2011) and other media (Blumenthal 2002), and public dissemination working alongside academics (Chang et al 2011, 2012, Chang 2011, 2012).
Research and development in contemporary experimental cuisine has had a focus on culinary uses for new, native, very old, or underutilized plant and animal ingredients; including seaweeds, insects, and foraged and indigenous plants unique to a region, such as Scandinavia or the Amazon (Blumenthal 2013, Mouritsen et al 2012, Williams and Hermansen 2012, Redzepi 2010, Redzepi 2013, Nilsson 2012, Atala 2012, 2013). Novel techniques and new applications for existing techniques are also of great interest; from adapting food-industry hydrocolloids for fne-dining applications like gels, fuid gels, foams/espumas, and “spherifcation” (Myhrvold et al 2011, Achatz and Kokonas 2008, Adria et al 2006), to interest in craf cocktails and other alcoholic beverages (Adria et al 2006, Conigliaro 2012), to current interest in traditional chemical processes such as nixtamalization and microbial processing by lactic acid bacteria, yeasts, acetic bacteria, and molds and fungi such as Aspergillus oryzae (Chang et al 2014, Chang 2010, Katz 2012, Nilsson 2012, Felder et al 2012).
Te development and implementation of many of these processes and products for the kitchen has frequently been directed by the goals of creating, capturing, and controlling favor. A theoretical, experimental, or technical approach to favor explicitly has also been the focus of a number of recent cookbooks (Kunz 2008, Page 2008, Segnit 2012, Afel and Patterson 2004). Some observations and theories on favor originating in the culinary world have become the basis for peer-reviewed work, refecting the potential for further collaboration on questions of favor. In one such example, the observation by a chef13 of more intense umami favor in the pulp and seeds
13 Heston Blumenthal of the Fat Duck 23 of tomatoes than the fesh led to a collaboration with favor chemists at the University of Reading that measured higher levels, on average, of the umami-tasting compounds glutamic acid, aspartic acid, and several 5’-ribonucleotides in the pulp than the fesh of several diferent tomato varieties (Oruna-Concha et al 2007). In another case, questions about why ingredients pair together well in terms of favor led to a creative theory, developed by the same chef working with favorists, which stated that new favor pairings of ingredients not commonly used together in one dish could be identifed by comparing their volatile profles, and that shared supra-threshold volatiles would allow for a harmonious pairing (Blumenthal 2002, 2008). Some investigation of the mechanisms underlying paired favors by a sensory science team found that the rated pleasantness of binary odor mixtures is not solely determined by the rated pleasantness of individual components, but did not measure overlapping compounds (Møller et al 2011).
Other work on favor combinations has been more methodologically fawed. A data- mining experiment of recipes collected from three internet repositories concluded that North American recipes tended to pair ingredients that shared favor compounds, while East Asian recipes did not tend to pair ingredients that share favor compounds (Ahn et al 2011). However, the authors do not describe any experimental design or controls on bias for collecting the 56,000 recipes in the dataset, and state that the largeness of the dataset allows for factors relating to artistic creativity, ingredients (such as egg) included for structural rather than favor properties, and favors arising from cooking method rather than measurable in raw ingredients to be systematically fltered out, though they do not describe a method for this fltering. Te authors describe the most “authentic” ingredients in North American cuisine (“authenticity” defned as used more ofen in North American recipes in the dataset than other region’s recipes) to be milk, butter, vanilla, egg, molasses, and wheat, which would seem to suggest that North American cuisine is defned almost entirely by dairy and desserts.
Issues with false positives and with conclusions drawn from analysis of large datasets without taking into account current knowledge from small, well-designed experiments in the feld
24 have been noted in other disciplines such as evolutionary biology, genomics, and medicine (Graur et al 2013, Alberts 2012, PLoS Medicine Editors 2005); Ahn et al may be doing something similar in drawing conclusions about the prevalent recipes and ingredients that make up the essence of a culture’s cuisine, favor chemistry, and cooking, without citing references from the felds of food cultural anthropology, favor chemistry, or gastronomy in any manner more in-depth than passing. Given that as of April 2014 it has been cited 30 times, and an analysis by Altmetric puts it in the top 5% of all articles in terms of social and traditional media attention, the high interest in favor and chemistry as it relates to cooking and cuisine is an opening, even a demand, for more rigorously designed favor chemistry studies of gastronomically interesting products and ideas.
Snitkjaer et al. (2010 and 2011) provide one example of how this can be done: both papers are singular in their combined use of established experimental methods and explicit gastronomic context. Beyond this this, the application of sensory and chemical profling, shown to provide valuable insight for so many other food products, has been applied infrequently to these new and resurgent products and processes. Several high-profle calls (Dufresne 2012, Barber 2012, Myhrvold et al 2011), for greater chef-scientist involvement reveal a latent demand for science that addresses cuisine and culinary questions explicitly, and research characterizing the favor chemistry of ingredients and culinary processes builds on both this demand and the nascent body of scientifc work addressing food in a culinary context, under the heading of “molecular gastronomy” or otherwise.
With the above as context, the research covered in this dissertation is particularly interested in characterizing the favor chemistry of products of cultural and gastronomic interest, particularly those in current use that have not received much scientifc attention. In particular, I will focus on aromatic cocktail bitters and uses of acetic fermentation to produce novel vinegars.
Bitters, highly concentrated alcoholic extractions of favorful plant materials, have become popular as an ingredient for delivering favors in craf cocktails (Sandham 2012, Soole 2013), with many well-respected cocktail bars stocking dozens or more diferent styles for specifc
25 favor profles (Parsons 2011). Bitters were in widespread use in the 19th and early 20th centuries (Wondrich 2007, Haigh 2009) and have recently experienced a resurgence, with homemade and house-made recipes becoming internationally distributed brands (Regan 2003, Parsons 2011). Teir use refects a growing interest in application for favor extraction in a culinary environment (Conigliaro 2013, Liu 2013). Despite this interest, there is essentially no data available on their favor or favor chemistry.
Transformation of alcoholic liquids into those containing acetic acid (vinegars) by aerobic bacteria has been a process used in various cuisines for thousands of years (Mazza et al 2009). More recently, the process of acetic “fermentation”14 has become a popular tool in the kitchens of a number of high-end and experimental restaurants (Porcelli 2013, Redzepi 2013, Nilsson 2012, Reade 2012) for preserving seasonal favors, transforming the favor of ingredients not widely made into vinegar (such as mushroom juice or spruce needles), and creating culinary sources of acidity in locations where citrus fruits are not grown. Te favor chemistry of traditional vinegars, such sherry vinegar, Balsamic vinegar, and Shanxi buckwheat vinegar, have been characterized (Callejon et al 2008, Cirlini et al 2011, Aili et al 2010), but the products of this new interest in making novel vinegars in the restaurant world have not until now been studied. In collaboration with the research and development staf at the Nordic Food Lab, novel vinegars produced using rapid, forced-air acetifcation methods and passive acetifcation are analyzed.
Methods designed to address the complex relationship between chemical makeup and
perceived favor are also used and developed, including multivariate statistical techniques and in- instrument gas chromatography methods for recomposition olfactometry experiments.
14 Biochemically speaking, not technically fermentation, as transformation of ethyl alcohol into acetic acid is an aerobic process; however, from this point forward I will use the common convention of describing this process as acetic fermentation, without the use of quotation marks 26 References Aaslyng, M., & Frøst, M. (2010). Te efect of the combination of salty, bitter, and sour accompaniment on the favor and juiciness of pork patties. Journal of Sensory Studies, 25(4), 536–548. doi:10.1111/j.1745-459X.2010.00285.x Abdi, H. (2003). Multivariate Analysis. In M. Lewis-Beck, A. Bryman, & T. Futing (Eds.), Encyclopedia of Social Sciences Research Methods (pp. 1–4). Tousand Oaks, CA: Sage. Abdi, H., Toole, A. J. O., Valentin, D., & Edelman, B. (2005). DISTATIS : Te Analysis of Multiple Distance Matrices. In C. Schmid, S. Soatto, & C. Tomasi (Eds.), Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (p. 42). Los Alamitos, CA: IEEE Computer Society. Abdi, H., Valentin, D., Chollet, S., & Chrea, C. (2007). Analyzing assessors and products in sorting tasks: DISTATIS, theory and applications. Food Quality and Preference, 18(4), 627– 640. doi:10.1016/j.foodqual.2006.09.003 Abdi, H., & Williams, L. J. (2010). Principal Component Analysis. Wiley Interdisciplinary Reviews: Computational Statistics, 2(4), 433–459. Abend, L. (2012, March 26). Locavore Hero. Time Magazine, p. Cover. Aceña, L., Vera, L., Guasch, J., Busto, O., & Mestres, M. (2010). Comparative study of two extraction techniques to obtain representative aroma extracts for being analysed by gas chromatography-olfactometry: application to roasted pistachio aroma. Journal of chromatography. A, 1217(49), 7781–7787. doi:10.1016/j.chroma.2010.10.030 Achatz, G., & Kokonas, N. (2008). Alinea. Ten Speed Press. Retrieved from http://books.google. com/books?id=y59tuJs4v5QC Acree, T. E., & Barnard, J. (1984). A Procedure for the Sensory Analysis of Gas Chromatographic Efuents. Food Chemistry, 14, 273–286. Adria, F, Soler, J., & Adria, A. (2005). El Bulli: 1998-2002. New York, NY: HarperCollins. Retrieved from http://books.google.com/books?id=ofvwGwAACAAJ Adria, F, Soler, J., & Adria, A. (2006a). El Bulli 2003-2004. New York, NY: HarperCollins. Retrieved from http://books.google.com/books?id=MXarnQEACAAJ Adria, F, Soler, J., & Adria, A. (2006b). El Bulli 1994-1997. New York, NY: HarperCollins. Retrieved from http://books.google.com/books?id=6PCrdGUUyz4C Adrià, F., Soler, J., & Adrià, A. (2006). El bulli: 2005. RBA Libros, Barcelona. Retrieved from http://books.google.com/books?id=5ihfAAAACAAJ Adria, F., Blumenthal, H., Keller, T., & McGee, H. (2006, December 9). Statement on the “new cookery.” Te Observer. Retrieved from http://www.theguardian.com/uk/2006/dec/10/ foodanddrink.obsfoodmonthly Aduriz, A. L. (2012). Mugaritz: A Natural Science of Cooking. London: Phaidon Press. Retrieved from http://books.google.com/books?id=Uvf-ygAACAAJ
27 Afoakwa, E. O., Paterson, A., Fowler, M., & Ryan, A. (2008). Flavor formation and character in cocoa and chocolate: a critical review. Critical reviews in food science and nutrition, 48(9), 840–57. doi:10.1080/10408390701719272 Afel, M., & Patterson, D. (2004). Aroma. New York, NY: Artisan. Retrieved from http://books. google.com/books?id=msEMk17upAMC Ahn, Y.-Y., Ahnert, S. E., Bagrow, J. P., & Barabási, A.-L. (2011). Flavor network and the principles of food pairing. Scientifc reports, 1, 196. doi:10.1038/srep00196 Aili, W., Huanlu, S., & Zaigui, L. (2012). Key aroma compounds in Shanxi aged tartary buckwheat vinegar and changes during its thermal processing. Flavour and Fragrance Journal, (January 2011), 47–53. doi:10.1002/f.2079 Alberts, B. (2012). Te End of “ Small Science ”? Science, 337(September), 1583. Amerine, M. A., Pangborn, R. M., Roessler, E. B., & others. (1965). Principles of sensory evaluation of food. Principles of sensory evaluation of food. New York, NY: Academic Press, Inc.,. An, M., Hai, T., & Hatfeld, P. (2001). On-site feld sampling and analysis of fragrance from living lavender (Lavandula angustifolia L.) fowers by solid-phase microextraction coupled to gas chromatography and ion-trap mass spectrometry. Journal of chromatography. A, 917(1-2), 245–50. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/11403476 Andrews, C. (2011). Ferran: Te Inside Story of El Bulli and the Man Who Reinvented Food. New York: Penguin Group US. Retrieved from http://books.google.com/books?id=k- KAn8krJyUC Arboleya, J.-C., Olabarrieta, I., Luis-Aduriz, A., Lasa, D., Vergara, J., Sanmartín, E., … Martínez de Marañón, I. (2008). From the Chef’s Mind to the Dish: How Scientifc Approaches Facilitate the Creative Process. Food Biophysics, 3(2), 261–268. doi:10.1007/s11483-008- 9078-3 Arnold, D. (2006, June). What About Tis. Food Arts, 31–33. Atala, A. (2013). D.O.M.: Rediscovering Brazilian Ingredients. London: Phaidon Press. Retrieved from http://books.google.com/books?id=vNlWmQEACAAJ Atala, Alex. (2012). A new ingredient: Te introduction of priprioca in gastronomy. International Journal of Gastronomy and Food Science, 1(1), 61–63. doi:10.1016/j.ijgfs.2011.11.001 Authors, T. Pl. M. (2005). Why bigger is not yet better: the problems with huge datasets. PLoS medicine, 2(2), e55. doi:10.1371/journal.pmed.0020055 Auvray, M., & Spence, C. (2008). Te multisensory perception of favor. Consciousness and cognition, 17(3), 1016–31. doi:10.1016/j.concog.2007.06.005 Axel, R. (2005). Scents and sensibility: a molecular logic of olfactory perception (Nobel lecture). Angewandte Chemie (International ed. in English), 44(38), 6110–27. doi:10.1002/ anie.200501726 Aznar, M., López, R., Cacho, J., & Ferreira, V. (2003). Prediction of aged red wine aroma properties from aroma chemical composition. Partial least squares regression models. Journal of agricultural and food chemistry, 51(9), 2700–7. doi:10.1021/jf026115z 28 Barber, D. (2012). Te Taste of Wheat. Mad Symposium, July 2 2012, Copenhagen. Retrieved October 04, 2014, from http://www.madfood.co/dan-barber-2/ Barham, P., Skibsted, L. H., Bredie, W. L. P., Frøst, M. B., Møller, P., Risbo, J., … Mortensen, L. M. (2010). Molecular gastronomy: a new emerging scientifc discipline. Chemical reviews, 110(4), 2313–65. doi:10.1021/cr900105w Barkat, S., Le Berre, E., Coureaud, G., Sicard, G., & Tomas-Danguin, T. (2011). Perceptual Blending in Odor Mixtures Depends on the Nature of Odorants and Human Olfactory Expertise. Chemical senses, (1994), 159–166. doi:10.1093/chemse/bjr086 Bentley, R. (2006). Te nose as a stereochemist. Enantiomers and odor. Chemical reviews, 106(9), 4099–4112. doi:10.1021/cr050049t Berglund, B., & Olsson, M. J. (1993). Odor-intensity interaction in binary and ternary mixtures. Perception & psychophysics, 53(5), 475–82. Retrieved from http://www.ncbi.nlm.nih.gov/ pubmed/8332416 Bicchi, C., Iori, C., Rubiolo, P., & Sandra, P. (2002). Headspace Sorptive Extraction ( HSSE ), Stir Bar Sorptive Extraction ( SBSE ), and Solid Phase Microextraction ( SPME ) Applied to the Analysis of Roasted Arabica Cofee and Cofee Brew. Journal of agricultural and food chemistry, 50(3), 449–459. Blumenthal, H. (2006). Heston Blumenthal: In Search of Perfection: Reinventing Kitchen Classics. New York, NY: Bloomsbury USA. Retrieved from http://books.google.com/ books?id=GcdmngEACAAJ Blumenthal, H. (2013). Historic Heston. New York, NY: Bloomsbury USA. Retrieved from http:// books.google.com/books?id=bQqcAQAAQBAJ Blumenthal, H, Lane, T., & Sewell, A. (2010). Heston’s Fantastical Feasts. New York, NY: Bloomsbury USA. Retrieved from http://books.google.com/books?id=xQ2IRAAACAAJ Blumenthal, Heston. (2002, May 3). Weird but wonderful. Te Guardian. London. Retrieved from http://www.guardian.co.uk/lifeandstyle/2002/may/04/foodanddrink.shopping Bushdid, C., Magnasco, M. O., Vosshall, L. B., & Keller, a. (2014). Humans Can Discriminate More than 1 Trillion Olfactory Stimuli. Science, 343(6177), 1370–1372. doi:10.1126/ science.1249168 Cairncross, S. E., & Sjostrom, L. B. (1950). Flavor profles: a new approach to favor problems. In M. Gakula (Ed.), Descriptive Sensory Analysis in Practice. Trumbull, CT: Food & Nutrition Press. Caldeira, M., Rodrigues, F., Perestrelo, R., Marques, J. C., & Câmara, J. S. (2007). Comparison of two extraction methods for evaluation of volatile constituents patterns in commercial whiskeys Elucidation of the main odour-active compounds. Talanta, 74(1), 78–90. doi:10.1016/j.talanta.2007.05.029 Callejón, R. M., Morales, M. L., Ferreira, A. C. S., & Troncoso, A. M. (2008). Defning the typical aroma of sherry vinegar: sensory and chemical approach. Journal of agricultural and food chemistry, 56(17), 8086–8095. doi:10.1021/jf800903n
29 Capron, X., Smeyersverbeke, J., & Massart, D. (2007). Multivariate determination of the geographical origin of wines from four diferent countries . Food Chemistry, 101(4), 1585– 1597. doi:10.1016/j.foodchem.2006.04.019 Cerretani, L., Biasini, G., Bonoli-Carbognin, M., & Bendini, A. (2007). Harmony of virgin olive oil and food pairing: a methodological proposal. Journal of Sensory Studies, 22(4), 403–416. doi:10.1111/j.1745-459X.2007.00115.x Chang, D., & Meehan, P. (2010). Momofuku. New York, NY: Crown Publishing Group. Retrieved from http://books.google.com/books?id=ovwudzKRRKQC Chang, D., Meehan, P., & Ying, C. (2014). Lucky Peach Issue 10: Street Food Issue. New York, NY: Lucky Peach. Retrieved from http://books.google.com/books?id=pWk3nwEACAAJ Chollet, S., Lelièvre, M., Abdi, H., & Valentin, D. (2011). Sort and beer: Everything you wanted to know about the sorting task but did not dare to ask. Food Quality and Preference, 22(6), 507–520. doi:10.1016/j.foodqual.2011.02.004 Cirlini, M., Caligiani, a., Palla, L., & Palla, G. (2011). HS-SPME/GC–MS and chemometrics for the classifcation of Balsamic Vinegars of Modena of diferent maturation and ageing. Food Chemistry, 124(4), 1678–1683. doi:10.1016/j.foodchem.2010.07.065 Civille, G. V., & Lyon, B. G. (1996). Aroma and favor lexicon for sensory evaluation: terms, defnitions, references, and examples. West Conshohocken, PA: ASTM. Conigliaro, T. (2013). Te Cocktail Lab: Unraveling the Mysteries of Flavor and Aroma in Drink, With Recipes. Ten Speed Press. Retrieved from http://books.google.com/ books?id=vF3RYlosi80C Conigliaro, Tony. (2012). Q&A: the science of cocktails. Flavour, 1(1), 19. doi:10.1186/2044-7248- 1-19 De Mello Castanho Amboni, R. D., da Silva Junkes, B., Yunes, R. a, & Heinzen, V. E. (2000). Quantitative structure-odor relationships of aliphatic esters using topological indices. Journal of agricultural and food chemistry, 48(8), 3517–21. Retrieved from http://www.ncbi. nlm.nih.gov/pubmed/10956142 Domrachev, D. V., Karpova, E. V., Goroshkevich, S. N., & Tkachev, a. V. (2012). Comparative analysis of volatiles from needles of fve-needle pines of northern and eastern Eurasia. Russian Journal of Bioorganic Chemistry, 38(7), 780–789. doi:10.1134/S1068162012070059 Donadini, G., Fumi, M. D., & Lambri, M. (2012). Te hedonic response to chocolate and beverage pairing: A preliminary study. Food Research International, 48(2), 703–711. doi:http://dx.doi. org/10.1016/j.foodres.2012.06.009 Donadini, G., Fumi, M. D., & Lambri, M. (2013). A preliminary study investigating consumer preference for cheese and beer pairings. Food Quality and Preference, 30(2), 217–228. doi:http://dx.doi.org/10.1016/j.foodqual.2013.05.012 Dufresne, W. (2012). An Appetite for Knowledge. Mad Symposium, July 2 2012, Copenhagen. Retrieved October 04, 2014, from http://www.madfood.co/wylie-dufresne/
30 Ebeler, S. E., & Torngate, J. H. (2009). Wine chemistry and favor: looking into the crystal glass. Journal of agricultural and food chemistry, 57(18), 8098–8108. doi:10.1021/jf9000555 Edelstein, S. (2013). Food Science: An Ecological Approach. Sudbury, MA: Jones & Bartlett Learning. Retrieved from http://books.google.com/books?id=aDI5Bmn7TRQC Edwards, P. a., Anker, L. S., & Jurs, P. C. (1991). Quantitative structure-property relationship studies of the odor threshold of odor active compounds. Chemical Senses, 16(5), 447–465. doi:10.1093/chemse/16.5.447 Fan, W., Xu, Y., & Han, Y. (2011). Quantifcation of Volatile Compounds in Chinese Ciders by Stir Bar Sorptive Extraction (SBSE) and Gas Chromatography-Mass Spectrometry (GC-MS). Journal of the Institute of Brewing, 117(1), 61–66. doi:10.1002/j.2050-0416.2011.tb00444.x Felder, D., Burns, D., & Chang, D. (2012). Defning microbial terroir: Te use of native fungi for the study of traditional fermentative processes. International Journal of Gastronomy and Food Science, 1(1), 64–69. doi:10.1016/j.ijgfs.2011.11.003 Ferraces-Casais, P., Lage-Yusty, M. A., Rodríguez-Bernaldo de Quirós, A., & López-Hernández, J. (2013). Rapid identifcation of volatile compounds in fresh seaweed. Talanta, 115, 798–800. doi:10.1016/j.talanta.2013.06.049 Floriano, W. B., Vaidehi, N., Goddard, W. a, Singer, M. S., & Shepherd, G. M. (2000). Molecular mechanisms underlying diferential odor responses of a mouse olfactory receptor. Proceedings of the National Academy of Sciences of the United States of America, 97(20), 10712–6. Retrieved from http://www.pubmedcentral.nih.gov/articlerender. fcgi?artid=27088&tool=pmcentrez&rendertype=abstract Giboreau, A., Dacremont, C., Egorof, C., Guerrand, S., Urdapilleta, I., Candel, D., & Dubois, D. (2007). Defning sensory descriptors: Towards writing guidelines based on terminology. Food quality and preference, 18(2), 265–274. Graur, D., Zheng, Y., Price, N., Azevedo, R. B. R., Zufall, R. a, & Elhaik, E. (2013). On the immortality of television sets: “function” in the human genome according to the evolution- free gospel of ENCODE. Genome biology and evolution, 1–43. doi:10.1093/gbe/evt028 Grosch, W. (2001). Evaluation of the key odorants of foods by dilution experiments, aroma models and omission. Chemical senses, 26, 533–545. Retrieved from http://www.ncbi.nlm. nih.gov/pubmed/11418500 Grosch, Werner. (1993). Detection of potent odorants in foods by aroma extract dilution analysis. Trends in Food Science & Technology, 4, 68–73. Haenlein, M., & Kaplan, A. M. (2004). A Beginner’s Guide to Partial Least Squares Analysis. Understanding Statistics, 3(4), 283–297. doi:10.1207/s15328031us0304_4 Haigh, T. (2009). Vintage Spirits and Forgotten Cocktails: From the Alamagoozlum to the Zombie 100 Rediscovered Recipes and the Stories Behind Tem. Beverly, MA: Quarry Books. Retrieved from http://books.google.com/books?id=sCR7wWhM7IQC
31 Harrington, R. J., & Hammond, R. (2006). Te Direct Efects of Wine and Cheese Characteristics on Perceived Match. Journal of Foodservice Business Research, 8(4), 37–54. doi:10.1300/ J369v08n04 Harrington, R. J., McCarthy, M., & Gozzi, M. (2010). Perceived Match of Wine and Cheese and the Impact of Additional Food Elements: A Preliminary Study. Journal of Foodservice Business Research, 13(4), 311–330. doi:10.1080/15378020.2010.524541 Hein, K., Ebeler, S. E., & Heymann, H. (2009). Perception of Fruity and Vegetative Aromas in Red Wine. Journal of Sensory Studies, 24(3), 441–455. doi:10.1111/j.1745-459X.2009.00220.x Heymann, H., Hopfer, H., & Bershaw, D. (2013). An Exploration of the Perception of Minerality in White Wines by Projective Mapping and Descriptive Analysis. Journal of Sensory Studies, 29(1), 1–13. doi:10.1111/joss.12076 Hjelmeland, A. K., King, E. S., Ebeler, S. E., & Heymann, H. (2012). Characterizing the Chemical and Sensory Profles of United States Cabernet Sauvignon Wines and Blends. American Journal of Enology and Viticulture, 64(2), 169–179. doi:10.5344/ajev.2012.12107 Hong, S., & Corey, E. J. (2006). Enantioselective syntheses of georgyone, arborone, and structural relatives. Relevance to the molecular-level understanding of olfaction. Journal of the American Chemical Society, 128(4), 1346–52. doi:10.1021/ja057483x Jinks, A., & Laing, D. G. (1999). A limit in the processing of components in odour mixtures. Perception, 28(3), 395–404. doi:10.1068/p2898 Kaack, K., Christensen, L. P., Hughes, M., & Eder, R. (2006). Relationship between sensory quality and volatile compounds of elderfower (Sambucus nigra L.) extracts. European Food Research and Technology, 223(1), 57–70. doi:10.1007/s00217-005-0122-y Kamozawa, A., & Talbot, H. A. (2010). Ideas in Food: Great Recipes and Why Tey Work. New York, NY: Crown Publishing Group. Retrieved from http://books.google.com/ books?id=GKgyvKxI4DcC Kanavouras, A., Kiritsakis, A., & Hernandez, R. J. (2005). Comparative study on volatile analysis of extra virgin olive oil by dynamic headspace and solid phase micro-extraction. Food Chemistry, 90, 69–79. doi:10.1016/j.foodchem.2004.03.025 Katz, S. E., & Pollan, M. (2012). Te Art of Fermentation: An In-Depth Exploration of Essential Concepts and Processes from around the World. White River Junction, VT: Chelsea Green Publishing. Retrieved from http://books.google.com/books?id=-zmLa205d0QC King, E. S., Dunn, R. L., & Heymann, H. (2013). Te infuence of alcohol on the sensory perception of red wines. Food Quality and Preference, 28(1), 235–243. doi:10.1016/j. foodqual.2012.08.013 Laing, D. G., & Francis, G. W. (1989). Te capacity of humans to identify odors in mixtures. Physiology & behavior, 46, 809–814. Retrieved from http://www.ncbi.nlm.nih.gov/ pubmed/2628992
32 Laing, D. G., Legha, P. K., Jinks, a L., & Hutchinson, I. (2003). Relationship between molecular structure, concentration and odor qualities of oxygenated aliphatic molecules. Chemical senses, 28(1), 57–69. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/12502524 Lawless, H. T., & Heymann, H. (2010a). Data Relationships and Multivariate Applications. In Sensory Evaluation of Food (pp. 433–449). Berlin: Springer Science+Business Media. doi:10.1007/978-1-4419-6488-5 Lawless, H. T., & Heymann, H. (2010b). Descriptive Analysis. In Sensory Evaluation of Food (pp. 227–257). Berlin: Springer Science+Business Media. doi:10.1007/978-1-4419-6488-5 Le Berre, E, Béno, N., Ishii, A., Chabanet, C., Etiévant, P., & Tomas-Danguin, T. (2008). Just noticeable diferences in component concentrations modify the odor quality of a blending mixture. Chemical senses, 33(4), 389–395. doi:10.1093/chemse/bjn006 Le Berre, Elodie, Tomas-Danguin, T., Béno, N., Coureaud, G., Etiévant, P., & Prescott, J. (2008). Perceptual processing strategy and exposure infuence the perception of odor mixtures. Chemical senses, 33(2), 193–199. doi:10.1093/chemse/bjm080 Lee, S, & Ahn, B. (2009). Comparison of volatile components in fermented soybean pastes using simultaneous distillation and extraction (SDE) with sensory characterisation. Food Chemistry, 114(2), 600–609. doi:10.1016/j.foodchem.2008.09.091 Lee, Seung-joo, & Noble, A. C. (2006). Use of Partial Least Squares Regression and Multidimensional Scaling on Aroma Models of California Chardonnay Wines. American Journal of Enology and Viticulture, 57(3), 363–370. Lelièvre, M., Chollet, S., Abdi, H., & Valentin, D. (2008). What is the validity of the sorting task for describing beers? A study using trained and untrained assessors. Food Quality and Preference, 19(8), 697–703. doi:10.1016/j.foodqual.2008.05.001 Liu, K. (2013). Craf Cocktails at Home: Ofeat Techniques, Contemporary Crowd-Pleasers, and Classics Hacked with Science. Seattle, WA: Amazon Digitla Services. Retrieved from http:// books.google.com/books?id=OOsX44gYoo8C Livermore, a, & Laing, D. G. (1996). Infuence of training and experience on the perception of multicomponent odor mixtures. Journal of experimental psychology. Human perception and performance, 22(2), 267–77. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/8934843 Livermore, A., & Laing, D. G. (1998). Te infuence of chemical complexity on the perception of multicomponent odor mixtures. Perception & Psychophysics, 60(4), 650–661. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/9628996 Lizarraga-Guerra, R., Guth, H., & Lopez, M. G. (1997). Identifcation of the Most Potent Odorants in Huitlacoche ( Ustilago maydis ) and Austern Pilzen ( Pleurotus sp .) by Aroma Extract Dilution Analysis and Static Head-Space Samples. Journal of agricultural and food chemistry, 45, 1329–1332. Lopez, R., Aznar, M., Cacho, J., & Ferreira, V. (2002). Determination of minor and trace volatile compounds in wine by solid-phase extraction and gas chromatography with mass spectrometric detection. Journal of Chromatography A, 966, 167–177.
33 Madrigal-galan, B., & Heymann, H. (2006). Sensory Efects of Consuming Cheese Prior to Evaluating Red Wine Flavor. American Journal of Enology and Viticulture, 1, 12–22. Malherbe, S., Watts, V., Nieuwoudt, H. H., Bauer, F. F., & du Toit, M. (2009). Analysis of volatile profles of fermenting grape must by headspace solid-phase dynamic extraction coupled with gas chromatography-mass spectrometry (HS-SPDE GC-MS): novel application to investigate problem fermentations. Journal of agricultural and food chemistry, 57(12), 5161– 6. doi:10.1021/jf900532v Malnic, B., Hirono, J., Sato, T., & Buck, L. B. (1999). Combinatorial receptor codes for odors. Cell, 96(5), 713–23. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/10089886 Martin, N., Berger, C., Du, C. Le, & Spinnler, H. E. (2001). Aroma Compound Production in Cheese Curd by Coculturing with Selected Yeast and Bacteria, 2125–2135. Mazza, S., & Murooka, Y. (2009). Vinegars Trough the Ages. In L. Solieri & P. Giudici (Eds.), Vinegars of the World (pp. 18–39). Springer Milan. McGee, H. (1999). Taking stock of new favours. Nature, 400(6739), 17–8. doi:10.1038/21775 Mielby, L. H., & Frøst, M. B. (2010). Expectations and surprise in a molecular gastronomic meal. Food Quality and Preference, 21(2), 213–224. doi:10.1016/j.foodqual.2009.09.005 Mo, X., Xu, Y., & Fan, W. (2010). Characterization of aroma compounds in Chinese rice wine Qu by solvent-assisted favor evaporation and headspace solid-phase microextraction. Journal of agricultural and food chemistry, 58(4), 2462–9. doi:10.1021/jf903631w Møller, P., Hartvig, D., & Bredie, W. (2011). Determinants of the pleasantness of odor mixtures. Chemical Senses, 36, E71–E72. doi:10.1093/chemse/bjq126 Mouritsen, O G, Johansen, M., & Mouritsen, J. D. (2013). Seaweeds: Edible, Available, and Sustainable. Chicago, IL: University of Chicago Press. Retrieved from http://books.google. com/books?id=sLDb13RjmHcC Mouritsen, Ole G, Williams, L., Bjerregaard, R., & Duelund, L. (2012). Seaweeds for umami favour in the New Nordic Cuisine. Flavour, 1(1), 4. doi:10.1186/2044-7248-1-4 Murray, J. M., & Delahunty, C. M. (2000). Selection of standards to reference terms in a cheddar- type cheese favor language. Journal of Sensory Studies, 15(2), 179–199. Myhrvold, N. (2011). Te Art in Gastronomy: A Modernist Perspective. Gastronomica, 11(1), 13–23. Nilsson, M. (2012). Fäviken. London: Phaidon Press. Retrieved from http://books.google.com/ books?id=JvKxuAAACAAJ Noble, A. C., & Ebeler, S. E. (2002). Use of multivariate statistics in understanding wine favor. Food Reviews International, 18(1), 1–20. doi:10.1081/FRI-120003414 Nygren, I T, Gustafsson, I., & Johansson, L. (2002). Editorial review Perceived favour changes in white wine afer tasting blue mould cheese. Food Service Technology, 2, 163–171.
34 Nygren, I.T., Gustafsson, I., Haglund, Å., Johansson, L., & Noble, a C. (n.d.). Flavor changes produced by wine and food interactions: Chardonnay wine and hollandaise sauce. Journal of Sensory Studies, 16(2001), 461–470. Nygren, Ingemar T, Gustafsson, I., & Johansson, L. (2003). Perceived favour changes in blue mould cheese afer tasting. Food Service Technology, 143–150. Oruna-Concha, M.-J., Methven, L., Blumenthal, H., Young, C., & Mottram, D. S. (2007). Diferences in glutamic acid and 5’-ribonucleotide contents between fesh and pulp of tomatoes and the relationship with umami taste. Journal of agricultural and food chemistry, 55(14), 5776–5780. doi:10.1021/jf070791p Page, K., & Dornenburg, A. (2008). Te Flavor Bible: Te Essential Guide to Culinary Creativity, Based on the Wisdom of America’s Most Imaginative Chefs. New York, NY: Little, Brown and Co. Parsons, B. T., & Anderson, E. (2011). Bitters: A Spirited History of a Classic Cure-All, with Cocktails, Recipes, and Formulas. Berkeley, CA: Ten Speed Press. Retrieved from http:// books.google.com/books?id=OPlRDjfGnloC Patton, S., & Josephson, D. (1957). A method for determining signifcance of volatile favor compounds in foods. Journal of Food Science, 22(3), 316–318. Paulsen, M. T., Næs, T., Ueland, Ø., Rukke, E.-O., & Hersleth, M. (2013). Preference mapping of salmon–sauce combinations: Te infuence of temporal properties. Food Quality and Preference, 27(2), 120–127. doi:http://dx.doi.org/10.1016/j.foodqual.2012.09.010 Paulsen, M. T., Ueland, Ø., Nilsen, A. N., Öström, Å., & Hersleth, M. (2012). Sensory perception of salmon and culinary sauces – An interdisciplinary approach. Food Quality and Preference, 23(2), 99–109. doi:http://dx.doi.org/10.1016/j.foodqual.2011.09.004 Pineau, B., Barbe, J.-C., Van Leeuwen, C., & Dubourdieu, D. (2007). Which impact for beta- damascenone on red wines aroma? Journal of agricultural and food chemistry, 55(10), 4103– 4108. doi:10.1021/jf070120r Pineau, B., Barbe, J.-C., Van Leeuwen, C., & Dubourdieu, D. (2009). Examples of perceptive interactions involved in specifc “red-” and “black-berry” aromas in red wines. Journal of agricultural and food chemistry, 57(9), 3702–8. doi:10.1021/jf803325v Poinot, P., Grua-Priol, J., Arvisenet, G., Rannou, C., Semenou, M., Le Bail, A., & Prost, C. (2007). Optimisation of HS-SPME to study representativeness of partially baked bread odorant extracts. Food Research International, 40(9), 1170–1184. doi:10.1016/j.foodres.2007.06.011 Polásková, P., Herszage, J., & Ebeler, S. E. (2008). Wine favor: chemistry in a glass. Chemical Society reviews, 37(11), 2478–89. doi:10.1039/b714455p Porcelli, A. (2013). Cook it Raw. London: Phaidon Press. Retrieved from http://books.google.com/ books?id=iaFmngEACAAJ Powers, J. J. (1988). Current practices and application of descriptive methods. In J. Piggott (Ed.), Sensory analysis of foods (pp. 187–266). London: Elsevier Applied Science.
35 Preston, L. D., Block, D. E., Heymann, H., Soleas, G., Noble, A. C., & Ebeler, S. E. (2008). Defning Vegetal Aromas in Cabernet Sauvignon Using Sensory and Chemical Evaluations. American Journal of Enology and Viticulture, 59(2), 137–145. Pryde, J., Conner, J., Jack, F., Lancaster, M., Meek, L., Paterson, R. (2011). Sensory and Chemical Analysis of “Shackleton’s” Mackinlay Scotch Whisky. Journal of the Institute of Brewing, 117(2), 156–165. Redzepi, R. (2010). Noma: Time and Place in Nordic Cuisine. London: Phaidon Press. Retrieved from http://books.google.com/books?id=z3jyRQAACAAJ Redzepi, R. (2013). A Work in Progress. London: Phaidon Press. Retrieved from http://books. google.com/books?id=njlOnwEACAAJ Regan, G. (2003). Te Joy of Mixology. New York, NY: Crown Publishing Group. Retrieved from http://books.google.com/books?id=7TlhtrpXa-MC Ribeiro, J. S., Augusto, F., Salva, T. J. G., Tomaziello, R. A., & Ferreira, M. M. C. (2009). Prediction of sensory properties of Brazilian Arabica roasted cofees by headspace solid phase microextraction-gas chromatography and partial least squares. Analytica chimica acta, 634(2), 172–9. doi:10.1016/j.aca.2008.12.028 Rødbotten, M., Lea, P., & Ueland, Ø. (2009). Quality of raw salmon fllet as a predictor of cooked salmon quality. Food Quality and Preference, 20(1), 13–23. doi:10.1016/j. foodqual.2008.06.004 Rossiter, K. J. (1996). Structure − Odor Relationships. Chemical Reviews, 96(8), 3201–3240. Ryan, D., Prenzler, P. D., Saliba, A. J., & Scollary, G. R. (2008). Te signifcance of low impact odorants in global odour perception. Trends in Food Science & Technology, 19(7), 383–389. doi:10.1016/j.tifs.2008.01.007 Sanchez, J. V. S. G. (2011). Comparison of Descriptive Analysis and Projective Mapping Techniques in the Aroma Evaluation of the Distilled Spirits , Gin and Tequila. Master’s Tesis, University of California, Davis. 87 pp. Sandham, T. (2012). World’s Best Cocktails: 500 Signature Drinks from the World’s Best Bars and Bartenders. Minneapolis, MN: Fair Winds Press. Retrieved from http://books.google.com/ books?id=bQQNAAAAQBAJ Sanz, G., Tomas-Danguin, T., Hamdani, E. H., Le Poupon, C., Briand, L., Pernollet, J.-C., … Tromelin, A. (2008). Relationships between molecular structure and perceived odor quality of ligands for a human olfactory receptor. Chemical senses, 33(7), 639–53. doi:10.1093/ chemse/bjn032 Schoeller, M. (2013, November 18). Gods of Food. Time Magazine, p. Cover. Segnit, N. (2010). Te Flavor Tesaurus: A Compendium of Pairings, Recipes and Ideas for the Creative Cook. New York: Bloomsbury. Small, D. M., Bender, G., Veldhuizen, M. G., Rudenga, K., Nachtigal, D., & Felsted, J. (2007). Te role of the human orbitofrontal cortex in taste and favor processing. Annals of the New York Academy of Sciences, 1121, 136–51. doi:10.1196/annals.1401.002 36 Smith, A., & Kraig, B. (2013). Te Oxford Encyclopedia of Food and Drink in America. New York: Oxford University Press. Retrieved from http://books.google.com/ books?id=DOJMAgAAQBAJ Snitkjær, P., Frøst, M. B., Skibsted, L. H., & Risbo, J. (2010). Flavour development during beef stock reduction. Food Chemistry, 122(3), 645–655. doi:10.1016/j.foodchem.2010.03.025 Snitkjær, P., Risbo, J., Skibsted, L. H., Ebeler, S., Heymann, H., Harmon, K., & Frøst, M. B. (2011). Beef stock reduction with red wine – Efects of preparation method and wine characteristics. Food Chemistry, 126(1), 183–196. doi:10.1016/j.foodchem.2010.10.096 Soole, S., & Caudle, N. (2013). Cocktail Culture: Recipes & Techniques from Behind the Bar. Victoria, BC: TouchWood Editions. Retrieved from http://books.google.com/ books?id=g7H51fXdqV8C Stephenson, T. (2013). Te Curious Bartender. London: Ryland Peters & Small. Retrieved from http://books.google.com/books?id=jbq-lwEACAAJ Stone, H., Sidel, J., Oliver, S., Woolsey, A., & Singleton, R. C. (1974). Sensory evaluation by quantitative descriptive analysis. In M. Gacula (Ed.), Descriptive Sensory Analysis in Practice (pp. 23–34). Trumbull, CT: Food & Nutrition Press. Szczesniak, A. S., Brandt, M., & Friedman, H. (1963). Development of standard rating scales for mechanical parameters of texture and correlation between the objective and the sensory methods of texture evaluation. Journal of Food Science, 28(4), 397–403. Tan, V. (2013, January). Intentional Ambiguity. PhD Dissertation, Harvard University, Cambridge, MA. 163 pp. Tis, H, & DeBevoise, M. (2013). Molecular Gastronomy: Exploring the Science of Flavor. New York, NY: Columbia University Press. Retrieved from http://books.google.com/ books?id=QVXKZ1-z-eIC Tis, Hervé. (2013). Building a Meal: From Molecular Gastronomy to Culinary Constructivism. New York, NY: Columbia University Press. Retrieved from http://books.google.com/ books?id=qRc2AAAAQBAJ Tompson, B. (1802). On the Construction of Kitchen Fireplaces and Kitchen Utensils: Together with Remarks and Observations Relating to the Various Processes of Cookery; and Proposals for Improving that Most Useful Art. London: T. Cadell and W. Davies. Retrieved from http:// books.google.com/books?id=DylhMgEACAAJ Tonutti, I., & Liddle, P. (2010). Aromatic plants in alcoholic beverages. A review. Flavour and Fragrance Journal, 25(5), 341–350. doi:10.1002/f.2001 Valentin, D., Chollet, S., Lelièvre, M., & Abdi, H. (2012). Quick and dirty but still pretty good: a review of new descriptive methods in food science. International Journal of Food Science & Technology, 47(8), 1563–1578. doi:10.1111/j.1365-2621.2012.03022.x Verhagen, J. V, & Engelen, L. (2006). Te neurocognitive bases of human multimodal food perception: sensory integration. Neuroscience and biobehavioral reviews, 30(5), 613–50. doi:10.1016/j.neubiorev.2005.11.003
37 Wang, Y., Yang, C., Li, S., Yang, L., Wang, Y., Zhao, J., & Jiang, Q. (2009). Volatile characteristics of 50 peaches and nectarines evaluated by HP–SPME with GC–MS. Food Chemistry, 116(1), 356–364. doi:10.1016/j.foodchem.2009.02.004 Wardencki, W., Michulec, M., & Curylo, J. (2004). A review of theoretical and practical aspects of solid-phase microextraction in food analysis. International Journal of Food Science and Technology, 39(7), 703–717. doi:10.1111/j.1365-2621.2004.00839.x Williams, L. (2011). Fish Sauce. Nordic Food Lab Blog. Retrieved March 11, 2013, from http:// nordicfoodlab.org/blog/2011/10/fsh-sauce Williams, L., & Hermansen, M. E. (2012). Delineating the Edible and Inedible. Mad Symposium, July 2 2012, Copenhagen. Retrieved October 04, 2014, from http://www.madfood.co/ nordic-food-lab-2/ Willner, B., Granvogl, M., & Schieberle, P. (2013). Characterization of the Key Aroma Compounds in Bartlett Pear Brandies by Means of the Sensomics Concept. Journal of agricultural and food chemistry, 61, 9583–9593. doi:10.1021/jf403024t Wilson, D. A., & Stevenson, R. J. (2006). Learning to smell: olfactory perception from neurobiology to behavior. Baltimore, MD: Johns Hopkins University Press. Wondrich, D. (2007). Imbibe!: From Absinthe Cocktail to Whiskey Smash, a Salute in Stories and Drinks to “Professor” Jerry Tomas, Pioneer of the American Bar. New York, NY: Perigee Books. Retrieved from http://books.google.com/books?id=IV3s-NUYnfEC Wong, J. W., Webster, M. G., Halverson, C. A., Hengel, M. J., Ngim, K. K., & Ebeler, S. E. (2003). Multiresidue Pesticide Analysis in Wines by Solid-Phase Extraction and Capillary Gas Chromatography − Mass Spectrometric Detection with Selective Ion Monitoring. Journal of Agricultural and Food Chemistry, 51(5), 1148–1161. Woods, V. (1995). Efect of Geographical Origin and Extraction Method on the Sensory Characteristics of Vanilla Essences. M.S. Tesis, University of Missouri, Columbia, MO. 77 pp.
38 Chapter 2: Perceptual Characterization and Analysis of Aroma Mixtures using In-Instrument Gas Chromatography Recombination- Olfactometry1
Abstract Tis paper describes the design of a new instrumental technique, Gas Chromatography Recomposition-Olfactometry (GC-R), that adapts the reconstitution technique used in favor chemistry studies by extracting volatiles from a sample by headspace solid-phase microextraction (SPME), separating the extract on a capillary GC column, and recombining individual com- pounds selectively as they elute of of the column into a mixture for sensory analysis (Figure 2.1). Using the chromatogram of a mixture as a map, the GC-R instrument allows the operator to “cut apart" and recombine the components of the mixture at will, selecting compounds, peaks, or sections based on retention time to include or exclude in a reconstitution for sensory analysis. Se- lective recombination is accomplished with the installation of a Deans Switch directly in-line with the column, which directs compounds either to waste or to a cryotrap at the operator's discretion. Tis enables the creation of, for example, aroma reconstitutions incorporating all of the volatiles in a sample, including instrumentally undetectable compounds as well those present at concen- trations below sensory thresholds, thus correcting for the “reconstitution discrepancy" sometimes noted in favor chemistry studies. Using only fowering lavender (Lavandula angustifola ‘Hidcote Blue’) as a source for volatiles, we used the instrument to build mixtures of subsets of lavender volatiles in-instrument and characterized their aroma qualities with a sensory panel. We showed evidence of additive, masking, and synergistic efects in these mixtures and of “lavender' aroma character as an emergent property of specifc mixtures. Tis was accomplished without the need 1 Note: Tis was published as Johnson, A. J., Hirson, G. D., & Ebeler, S. E. (2012). Perceptual Char- acterization and Analysis of Aroma Mixtures Using Gas Chromatography Recomposition-Olfactometry. (E. M. C. Skoulakis, Ed.)PLoS ONE, 7(8), e42693. doi:10.1371/journal.pone.0042693 and appears here as the accepted text (including abstract)and fgures. 39 for chemical standards, reductive aroma models, or calculation of Odor Activity Values, and is broadly applicable to any aroma or favor.
Introduction Aroma plays a dominant role in the multisensory perception of favor. It is itself a construct perceived in response to stimulation of the olfactory system by volatile chemicals and mixtures thereof, with mixtures being commonly encountered in everyday life in the form of food, wine, plants, perfume, etc. While our understanding of the neurobiological and psychological mechanisms that translate volatiles into aroma perceptions has advanced signifcantly in recent years (Buck 2004, Axel 2004), analytical approaches for characterizing the perception of these aroma mixtures are still limited. Te relationship between chemical composition of a mixture of volatiles and its perceived aroma or favor is complex and difcult to predict on the basis of chemical data or simple sensory data alone. Analytical chemistry approaches for characterizing aromas or favors typically rely on separation-based chromatographic methods that quantify the aroma strength of individual compounds in a mixture, refected as either the concentration present in the mixture divided by a measured sensory threshold concentration (Odor Activity Value, OAV) (Patton and Josephson 1957, Guadagni et al. 1966) or the number of N-fold dilutions required to suppress detectability of a compound when analyzed by gas chromatography with a human subject acting as an olfactory detector (GC-Olfactometry or GC-O; CHARM; or Aroma Extract Dilution Analysis) (Acree and Barnard 1984, Grosch 1993, Grosch 2001). Reconstitution and omission experiments evaluate the role of specifc compounds in the perceived aroma of a mixture, whereby a blend of compounds hypothesized to be detectable in a food, beverage, or other sample by OAV is mixed from chemical standards, and compared to similar mixtures prepared by omitting one of these compounds at a time (Grosch 2001). If a diference is detectable in the “whole” mix versus a
“whole-minus-one-compound” mix, that particular compound is considered important to the aroma of the sample.
40 Figure 2.1 Conceptual schematic of the In-Instrument Gas Chromatograph Recombination Olfactometer (GRO) instrument. Volatiles are extracted onto a solid phase (via solid-phase microextraction or SPME) from the headspace of a food, beverage, or other sample, in this case, lavender fowers, and initially they are separated conventionally on an analytical capillary GC column. In-line with the GC column, a pneumatic Deans Switch followed by a cold trap allows the experimenter to build a mixture of these separated volatiles that is held until the cryotrap is rapidly heated, releasing the mixture for a subject to smell at the olfactory port and evaluate.
Knowledge from other disciplines studying aroma, such as sensory psychophysics, cognitive psychology, and molecular neurobiology, suggests limitations of these methodologies. Chromatographic techniques only assess the aroma quality of individual compounds, rather than mixtures of compounds. However, the aroma of a mixture is frequently perceptually distinct from that of its individual components (Wilson and Stevenson 2006, Laing and Francis 1989) and may have qualities not found in any of these components (Le Berre et al. 2008). Te mixing-dependent nature of aroma quality is evidenced by the relative lack of aroma impact compounds, or those compounds that are singularly responsible for the overall aroma impression of a food or beverage. On the other hand, omission experiments rely on an assumption that all sensorially important compounds have been correctly identifed and quantifed and that any compound occurring at a concentration below its putative sensory threshold is not important to the overall aroma. Recently published results suggest that this is not the case (Pineau et al. 2007). Despite having identical concentration profles of supra-threshold odorants, the aroma of a reconstitution sometimes still smells diferent from the original mixture (Steinhaus et al. 2009) a phenomenon referred to as “reconstitution discrepancy” (Bult et al. 2001). Some recent omission experiments have included sub-threshold components in the reconstitution (Bult et al. 2001), but this is not a universal practice, and can greatly complicate and enlarge the experimental design.
41 Figure 2.2 Schematic of (a) standard GC-MS; (b) GC-MS with splitter at end of column for olfactometry; and (c) In-instrument Gas chromatograph- Recombination Olfactometer or GRO with Deans switch, splitter, cryogenic trap and olfactory port. Abbreviations: i-inlet; c-column; d-detector; o-oven; olf-olfactometry port; sp-splitter; sw-Deans switch 1; w-waste; cr-cryogenic trap; and cb-switch 2 on control box.
We propose here a novel platform for the analytical characterization of aroma and favor perception that incorporates and merges aspects of the previously described techniques and knowledge from other related disciplines. We describe a series of non-reductive, in-instrument recombination and omission experiments using a Gas Chromatograph modifed with a switch and then a cold trap in-line between the capillary column and the chemical and olfactory detectors to characterize the aroma of lavender (Lavandula angustifola ‘Hidcote Blue’). Te volatile chemical composition of lavender, a potently aromatic herb with numerous culinary, cosmetic, and fragrance uses, has previously been characterized (An et al. 2001), but there are no lavender impact compounds currently identifed. Tis suggests that “lavender” aroma character arises from the perception of a mixture of volatiles rather than a single molecule, making this an ideal mixture for evaluation of perceptual interactions using our in-instrument gas chromatography recombination-olfactometry (GRO) approach.
Materials and Methods Instrument: Te GRO Gas Chromatograph is shown in Figure 2.2C. An Agilent model 6890 gas chromatograph/5972 mass spectral detector (GC-MSD) was modifed with the addition of a Deans switch apparatus (Agilent Technologies, Santa Clara, CA), an auxiliary pressure controller (EPC, Agilent) to control fow through the Deans switch, a splitter (Gerstel), a cryotrap (Micro Cryo- 42 Table 2.1 Experimental GC-O conditions and aroma descriptors for mixtures of volatiles from the lavender chromatograms.
Experimental Abbreviation Chromatogram Top Descriptors Condition Sections Included in Mixture Whole W 0-40 minutes Floral, citrus, dried lavender, fresh lavender, mint, wood Omission 1 O1 16-40 minutes Citrus, fresh lavender, dusty, foral, grassy/green, mint, pine, rotten Omission 2 O2 0-16+25-40 Citrus, haylike, foral, pine, root beer minutes Omission 3 O3 0-25 minutes Citrus, grassy/green, mint, wood, soapy Perceptual P1 0-11 minutes grassy/green, wood Interaction 1 Perceptual P2 11-16 minutes Floral, wood Interaction 2 Perceptual P3 16-20.5 minutes Citrus, foral, soapy Interaction 3 Perceptual P4 20.5-25 minutes Dusty, rotten, wet dirt Interaction 4 Perceptual P5 25-32 minutes Black pepper, haylike, citrus, foral, grassy/green Interaction 5 Perceptual P6 32-40 minutes Citrus, smoke Interaction 6 Lavender Reference Not separated; Citrus, foral, fresh lavender, mint, wood, hay, dried Flowers lavender, grassy/green Reference whole lavender fowers trap and model 971 controller, Scientifc Instrument Services, Ringoes NJ) and an olfactometry port (ODP-2, Gerstel, Linthicum, MD). Deactivated fused silica was used for all transfer lines. Te transfer line from the Deans switch to the splitter was 4 m. Te dimensions of the transfer line from the splitter to the MSD was 1 m x 0.15 mm; the dimensions of the transfer line from the splitter to the olfactory port was 1 m x 0.25 mm resulting in a 1.86:1 split ratio between the olfactory port and MSD.
Sampling and Chromatographic Conditions Lavender (Lavandula angustifola ‘Hidcote Blue’) fowers ( 0.50 g) were weighed and placed in a 20 mL amber glass headspace vial and sealed with a crimp cap with a PTFE-faced silicone septum (Supelco, St. Louis, MO). A Solid Phase Microextraction fber (2 cm length, 50/30 um divinylbenzene/carboxen/polydimethylsiloxane coating, Supelco) was used for extraction. 43 Te fber was exposed to the headspace of the vial for 30 minutes at room temperature, then withdrawn and immediately desorbed in the GC inlet. Chromatographic conditions were adapted from (An et al. 2001). Separation was performed using a 30m x 25mm i.d. x 0.25 um flm thickness DB-5MS column (J&W, Folsom, CA). Inlet was maintained at 240 °C in splitless mode. Helium was used as the carrier gas and was held at constant pressure at 15.5 psi. Te auxiliary pressure controller was maintained at 3.4 psi. Te SPME assembly was introduced manually into the inlet and allowed to desorb for a total of 10 minutes. Te oven was held at 60 °C for 3 minutes, then ramped to 150 °C at a rate of 3 °C /min, then ramped to 325 at a rate of 30 °C /min and held for 1 min for a total runtime of 40 minutes. Te olfactory port transfer line was maintained at 100 °C and the MSD transfer line was maintained at 260 °C. Afer a 0.5 min solvent delay, the mass spectrometer scanned from 50-230 amu. With the Deans switch set in the “of” position, the fow is directed to the splitter, MSD, cold trap, and ODP. When set to the “on” position, the fow is directed to waste. Te switch is programmed in the “runtime” tab of the Enhanced Chemstation Sofware (Hewlett Packard, version B.01.00) to direct the fow over the course of the runtime as desired by the operator.
Sensory Conditions Based on retention time, the Deans Switch sends specifc packets of volatiles to the cryotrap. Here we used one of 10 programs (W, O1-O3, P1-P6; see Figure 2.3, Table 2.1) where at the conclusion of the separation run, the cold trap was heated and the mixture was snifed and described by a sensory panelist. Te W condition, analogous to a full aroma reconstitute, contains all the volatiles of lavender, with conditions O1-O3 and P1-P6 omitting groups of these volatiles for descriptive comparison to the aroma of the W sample and to lavender fowers. Tree panelists (Females, ages 28-45 with previous sensory experience) smelled each of the 10 mixtures in triplicate and generated terms to describe the perceived odor. Before smelling each mixture, each panelist frst smelled and described a standard of lavender fowers, picked at the same time as the fowers used for SPME sampling, and also rated how well the sample mixture represented the aroma of the standard on a scale of 0-10. 44 Data Analysis Te terms used to describe the 10 mixtures were tabulated by frequency of use. Te descriptors used most ofen for each mixture, in a mixture-by-descriptor data matrix, was analyzed with a correspondence analysis to identify latent trends in similarity and diference in the multidimensional set. A three-way Analysis of Variance (ANOVA) with all two-way interactions was performed with rated representativeness of each mixture compared to a fresh lavender standard as the response factor and panelist, mixture, and replicate as main efects. A Tukey’s Honest Signifcant Diference multiple comparisons test (HSD) was performed on the representativeness ratings. Te
R statistical computing package was used for all statistical analyses (http://www.r-project.org/).
Results and Discussion We modifed a GC-MS to allow for the in-instrument preparation of volatile mixtures containing precise sections from a chromatogram, up to and including the entire volatile fraction and allowing for aroma characterization of the aroma of one or a few of the volatiles in a complex
Figure 2.3 Top aroma descriptors for mixtures of sections of the lavender chromatogram by cut time and chromatogram composition. Abbreviations correspond to Experimental Conditions described in Table 1. As chemical complexity and number of components per mixture approaches the makeup of the whole chromatogram (W) mixture, there is evidence of perceptual additivity as increasing cross- utilization of terms from simpler mixtures, masking as reduced use of dominant terms for simpler (P1- P6) mixtures, and synergistic efects as new complex or composite terms like “fresh lavender” become important.
45 mixture (Figure 2.1). Compounds were introduced into the inlet of the modifed GC-MS and separated on the analytical column. At the end of the column, the fow of carrier gas and analytes encountered a frst switch, a commercially available Deans switch, that was set to direct the fow either towards the splitter or towards waste (here waste was vented to the oven). Te splitter subsequently split the fow to both a mass spectrometer (MS) detector and to an olfactory port. Along the transfer line to the olfactory port was a trap controlled by a second switch at the control box; the switch allowed the trap to be cooled with liquid carbon dioxide or heated so that the eluant was either held within the trap (i.e., cryotrapped) or released to the olfactory port. By programming the switches to cryotrap or exclude selected peaks or peak regions (Table 2.1) two types of experiments were performed. In perceptual interaction experiments, all of the chromatogram except for a small section of peaks was cut away, and the section of interest was assessed at the olfactory port as a mixture. In omission experiments, small groups of peaks (or individual peaks) were cut away and the rest of the compounds in the chromatogram were smelled as a mixture. See supporting fgures S2.1 and S2.2 for examples of these chromatograms.
Figure 2.4 Correspondence Analysis of (A) lavender volatile mixtures; and (B) lavender volatile mixture descriptors. Abbreviations for mixtures correspond to those in Table 1. Terms generated by the panelists to describe the perceived odor of from each Experimental Condition described in Table 1 were tabulated by frequency of use and used for the Correspondence Analysis. 30.57% of variance explained by dimension 1 (x), 22.84% of variance explained by dimension 2 (y).
46 Using our new approach, ten aroma mixtures (Table 2.1, Figure 2.3) were created in-instrument directly from the headspace-extracted volatiles of fowering lavender. “Fresh Lavender” and “Dried Lavender” were both predominant descriptors for the “Whole Volatile” recombination mixture W. Of the more chemically complex omission mixtures O1-O3, only O1, which incorporated the section of volatiles eluting from 16-40 min of the lavender chromatogram and omitted volatiles eluting between 0-16 min, was described as having “fresh lavender” properties. O1 overlapped with O2 from 25-40 min and with O3 from 16-25 minutes and incorporated the perceptual mixtures P3-P6, however, none of these other omission or perceptual mixtures had fresh or dried lavender among their commonly used descriptors. Tis suggests that there are two subsets of compounds, the frst eluting between 16-25 min and the other eluting between 25-40 min, that are each necessary for the perception of “lavender character” but are not alone sufcient for inducing this perception without some mixing with compounds in the other elution group. Tese results also suggest that “lavender character” is an emergent perceptual property arising from the mixing of these volatiles or some subset thereof. We performed a Correspondence Analysis on the descriptors-by-mixtures data matrix to compare dimensionally-reduced latent trends in the sensory profles of the mixtures to the diferences evident in top descriptors for each mixture (Figure 2.4). Correspondence Analysis separates dissimilar categories in space; mixtures and sensory descriptors spaced closely together share more similarities than those spaced further apart. Tis plot shows that, generally, removing more volatiles results in greater dissimilarity between a given mixture and the all-volatiles- included mixture W. Te relatively tight clustering of W and omission mixtures O1-O3 in the Correspondence Analysis refects the sensory similarity of these mixtures; perceptual mixtures P2 and P3 also cluster nearby, refecting some of the overlapping characteristics of these mixtures (Figure 2.4). Te location of mixture W in the center of the main cluster in the Correspondence
Analysis, suggests its aroma was perceived, in part, as a sensory average of some of the less- complex mixtures. However, a truly averaged perceptual character would be in the center of
47 the plot; the fact that mixture W is ofset from the geometric center implies that the mixing- dependent interactive efects of the lavender volatiles perceived in mixture W play a noticeable role in afecting its overall aroma character. Mixture W shares many similar descriptors (Table 2.1) with O1-O3 and P2 and P3, but all of these except O1 lack a dominant lavender character. Mixtures P1 and P5 are close to the central cluster but are approximately equi-distant in space from mixture W. Tis refects some of the similarities in the descriptors that P1 and P5 share with mixture W, but also refects the domination of the aromas of these mixtures by either a unique character (“black pepper”) in the case of P5, or the relative simplicity of the aroma in the case of P1 (Figure 2.4a). Te comparative distancing of mixtures P4 and P6 from the other mixtures refects the relative uniqueness of their aroma descriptors. Locations of descriptors suggest that along the frst (x) dimension of Figure 2.4b, there is a distinction between fresher, more “sweet” and fower-associated terms on the right side and earthier, heavier aroma terms on the lef. Borrowing more qualitative terms from the tradition of perfumery (which at its essence is the craf of observing and optimizing the perceptual efects of mixing volatiles), we observe a rough Figure 2.5 Te rated representativeness of the aroma progression, from lef to right along the of samples W, O1-O3, and P1-P6 as compared by panelists to the aroma of whole fowering lavender. x-axis, of base, middle, and top-note (Afel Letters a, b, c refer to the mixture’s Signifcant Diference from each other- if two samples do 2004) related terms. Along the second (y) not share a letter, they are signifcantly diferent. Samples P1, P5, and P6 are signifcantly less dimension the separation is dominated by the representative of the aroma of fowering lavender than sample W, which incorporates all the volatiles marked diference of P4 and P6 from each in fowering lavender. other and from the rest of the mixtures, and correspondingly by their unique descriptors “wet dirt” and “smoky” in fgure 4b. Generally, the terms on the other arm of the y-dimension tend to be shared by multiple mixtures, or refect more composite aroma characteristics.
48 While sample P1 appears to be the closest to the central or average sample in this set, it is clearly separated from the cluster centered around mixture W along the third (z) dimension (Figure S3 supporting information). Te third dimension also further separates mixture P5 from the central W-associated cluster and increases the distinction between “grassy/green”- “woody” descriptors on one side and “dried lavender”-“black pepper” descriptors on the other. Importantly, the Correspondence Analysis, while unable to describe absolute diferences, provides valuable information not only on the sources of variation in the complex sensory data but also on the interrelationships of the mixtures and their sensory properties. Te method used to create an extract of volatile compounds can alter the perceived aroma of that extract and failure to obtain a representative sample can lead to unreliable conclusions about the composition of the aroma active components (Abbott et al. 1993, Etievant et al. 1993, van Ruth et al. 2004, Plutowska and Wardencki 2008, Aceña et al. 2010). While many extraction methods have been employed in order to produce an aroma extract (Plutowska et al. 2008, Aceña et al 2010, San-Juan et al. 2010), the creation of a representative aroma can be very difcult for complex matrices (Aceña et al. 2010, Pérez-Silva et al 2006), and the sensory representativeness of this extract is not always evaluated. Here, the aroma of the SPME extracts of lavender corresponded closely to the original product (Table 2.1). Similar representative aroma samples have been obtained using SPME to sample “baked potato” aroma (Poinot et al. 2007). Importantly, the GRO approach provides a rapid, easy, and efective tool to assess the representativeness of an extract regardless of the extraction method employed, such as in cases where SPME coatings may not be able to produce an appropriate extract (Ferreira et al. 2002). Since the SPME extraction produced an aroma mixture representative of lavender, it was possible to perform omission and interaction experiments based on a starting point nearly identical to the intact lavender sample, eliminating “reconstitution discrepancy” (Bult et al. 2001). Comparing the aroma of the GRO mixtures in this study to the aroma of whole lavender fowers, panelists found that mixtures P1, P5, and P6 were signifcantly less representative (Figure 2.5) of the aroma of the whole fowers than mixtures W, O1-O3 and P2-P4. Tese samples also tended to have either fewer commonly used descriptors or descriptors not found for other mixtures (such as “black pepper” for P5 and “smoke” for P6; Table 2.1). 49 In this experimental design, mixtures of compounds were omitted to assess the resulting aroma. Cut times were chosen to include chemically similar compounds in the same mixture, for example, monoterpene acetate esters in mixture P5 and sesquiterpenes in mixture P6. However, the omitted compounds/fractions in a theoretical GRO experiment need not be contiguous. It is possible, for example, to remove every other chromatographic peak, to remove only the 3rd and 17th peak, etc. while trapping and evaluating the remaining components. Te apparatus could additionally be used to perform single omission experiments, where compounds are omitted one at a time to screen for potential impact odorants, or perceptual interaction experiments where only 2 or 3 peaks are included in the mixture. Te fexibility in the compounds that can be removed and assessed is only limited by the rapid switching time of the Deans switch. By using a Mass Spectrometric detector, compounds in the sample can be identifed (Table S2.4, Supporting Information) however, an obvious advantage of performing an omission experiment in this manner is that the compounds need not be identifable or available to perform the experiment. Reconstitution experiments ofen require the experimenter to perform lengthy and labor-intensive syntheses to prepare a component for the reconstitution model (Steinhaus et al. 2009) only to fnd that the component can be omitted with no change in the overall aroma of the solution. Furthermore, there is always some fraction of the total compounds identifed that are not included in the reconstitution because they are deemed to have a concentration too low to have an efect on the overall aroma. However, compounds with low odor activity values ofen still have a considerable efect on the overall aroma of the mixture (Pineau et al. 2007, Escudero et al. 2004, Ryan et al. 2008). With this instrument there is no simplifed reconstitute - the omission experiment is performed on the entire sample. While compounds with low OAVs may be important to the aroma of the mixtures, the opposite case can also occur, and the sensitivity of the human nose is frequently orders of magnitude greater than an instrumental detector. As a result, the nose may detect an aroma where there is no peak on a chromatogram (Etievant et al. 1993). Particularly as compared to reconstitution studies, this is another distinct advantage of the GRO approach since even compounds not detected by the detector (MS, FID) will be included in the aroma sample as it is
50 assessed by a subject at the olfactometry port. Traditionally, full separation of volatile compounds on the chromatographic column is necessary in order to meaningfully describe the aroma character of the eluant by GC-O since it simplifes the recognition task for the assessor (San-Juan et al. 2010). However, it is more ofen the case that a complex mixture of aroma compounds is responsible for the overall aroma of a food or beverage. In addition, a mixture of two or more odorants can frequently lead to an aroma that is not similar to any of its individual components (Le Berre et al. 2008a, b). Using an GRO technique, any of these interactions can readily be investigated; and all that is necessary to characterize any type of aroma interaction is a sample of the food, beverage, fower, etc. of interest. Compounds detectable by GC-O but not GC-MS, compounds below putative aroma thresholds, compounds at levels that cannot be quantifed, and compounds not commercially available or easily synthesized can all be perceptually analyzed if they are found in one or more aromatic samples available to the researcher.
Conclusions Te perception of aroma and favor has ofen been approached as a problem of many individual parts, with chemistry, neurobiology, sensory science, psychology, and other disciplines focused on answering questions about some aspect of the relationship between stimulus (a fower, a glass of wine, a plate of food), response (perceived favor, liking or disliking, intake and satiety), or the pathway between the two (genetics, receptor binding, transduction, translation to cortical neurons). Tis has yielded a great deal of information about those individual parts, but not a well- developed understanding of how they work together for complex, everyday stimuli and activities like eating and drinking. Te need for a holistic approach to address this has been identifed previously (Shepherd 2006), i.e., a praxis which would bring together knowledge and research techniques from these diverse, ofen isolated, but orthogonally-related scientifc felds, and would include expertise or information from applied, non-analytical felds with a well-developed shared intuition about the nature of aroma and favor in practice, such as cuisine and perfumery. While the described approach of in-instrument gas chromatography recombination-olfactometry has its
51 roots in a traditional coupling of analytical chemistry and sensory science, it is highly informed by this multidisciplinary understanding of aroma and favor and allows for the analysis of previously uncharacterized emergent perceptual properties of complex mixture interaction efects in everyday smell and favor situations.
References
Abbott, N., Etihvant, P., Langlois, D., Lesschaeve, I., Issanchou, S., Recherche, L. De, … Cedex, D. (1993). Evaluation of the Representativeness of the Odor of Beer Extracts Prior to Analysis by GC Eluate Snifng. Journal of agricultural and food chemistry, 41, 777–780. Aceña, L., Vera, L., Guasch, J., Busto, O., & Mestres, M. (2010). Comparative study of two extraction techniques to obtain representative aroma extracts for being analysed by gas chromatography-olfactometry: application to roasted pistachio aroma. Journal of chromatography. A, 1217(49), 7781–7787. doi:10.1016/j.chroma.2010.10.030 Acree, T. E., & Barnard, J. (1984). A Procedure for the Sensory Analysis of Gas Chromatographic Efuents. Food Chemistry, 14, 273–286. Afel, M. (2004). Essence and Alchemy: A Natural History of Perfume. New York: Gibbs Smith. An, M., Hai, T., & Hatfeld, P. (2001). On-site feld sampling and analysis of fragrance from living lavender (Lavandula angustifolia L.) fowers by solid-phase microextraction coupled to gas chromatography and ion-trap mass spectrometry. Journal of chromatography. A, 917(1-2), 245–50. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/11403476 Axel, R. (2004). Scents and Sensibility: A Molecular Logic. In Les Prix Nobel. Te Nobel Prizes 2004 (pp. 234–256). Buck, L. B. (2004). Unraveling the sense of smell. In Les Prix Nobel. Te Nobel Prizes 2004 (pp. 267–283). Bult, J. H., Schiferstein, H. N., Roozen, J. P., Voragen, A. G., & Kroeze, J. H. (2001). Te infuence of olfactory concept on the probability of detecting sub- and peri-threshold components in a mixture of odorants. Chemical senses, 26(5), 459–469. Retrieved from http://www.ncbi.nlm. nih.gov/pubmed/11418491 Escudero, A., Gogorza, B., Melus, M. A., Ortin, N., Cacho, J., & Ferreira, V. (2004). Characterization of the aroma of a wine from maccabeo. Key role played by compounds with low odor activity values. Journal of agricultural and food chemistry, 52(11), 3516–24. doi:10.1021/jf035341l Etiévant, P. X., Moio, L., Guichard, E., Langlois, D., Leschaeve, I., & Schlich, P. (1993). Aroma extract dilution analysis (AEDA) and the representativeness of the odour of food extracts. In H. Maarse & D. G. Van Der Heij (Eds.), Trends in favour research, volume 35 of 52 Developments in food science (pp. 179–190). New York: Elsevier. Ferreira, V., Lopez, R., & Aznar, M. (2002). Olfactometry and aroma extract dilution analysis of wines. In J. F. Jackson & H. F. Linskens (Eds.), Analysis of Taste and Aroma Volume 21 (pp. 89–122). Heidelberg: Springer. Grosch, W. (2001). Evaluation of the key odorants of foods by dilution experiments, aroma models and omission. Chemical senses, 26, 533–545. Retrieved from http://www.ncbi.nlm. nih.gov/pubmed/11418500 Grosch, Werner. (1993). Detection of potent odorants in foods by aroma extract dilution analysis. Trends in Food Science & Technology, 4, 68–73. Guadagni, D. G., Buttery, R. G., & Harris, J. (1966). Odour intensities of hop oil components. J. Sci. Food Agric., 17(1), 142–144. Laing, D. G., & Francis, G. W. (1989). Te capacity of humans to identify odors in mixtures. Physiology & behavior, 46, 809–814. Retrieved from http://www.ncbi.nlm.nih.gov/ pubmed/2628992 Le Berre, E, Béno, N., Ishii, A., Chabanet, C., Etiévant, P., & Tomas-Danguin, T. (2008). Just noticeable diferences in component concentrations modify the odor quality of a blending mixture. Chemical senses, 33(4), 389–395. doi:10.1093/chemse/bjn006 Le Berre, Elodie, Tomas-Danguin, T., Béno, N., Coureaud, G., Etiévant, P., & Prescott, J. (2008). Perceptual processing strategy and exposure infuence the perception of odor mixtures. Chemical senses, 33(2), 193–199. doi:10.1093/chemse/bjm080 Patton, S., & Josephson, D. (1957). A method for determining signifcance of volatile favor compounds in foods. Journal of Food Science, 22(3), 316–318. Pérez-Silva, a., Odoux, E., Brat, P., Ribeyre, F., Rodriguez-Jimenes, G., Robles-Olvera, V., … Günata, Z. (2006). GC–MS and GC–olfactometry analysis of aroma compounds in a representative organic aroma extract from cured vanilla (Vanilla planifolia G. Jackson) beans. Food Chemistry, 99(4), 728–735. doi:10.1016/j.foodchem.2005.08.050 Pineau, B., Barbe, J.-C., Van Leeuwen, C., & Dubourdieu, D. (2007). Which impact for beta- damascenone on red wines aroma? Journal of agricultural and food chemistry, 55(10), 4103– 4108. doi:10.1021/jf070120r Plutowska, B., & Wardencki, W. (2008). Application of gas chromatography–olfactometry (GC– O) in analysis and quality assessment of alcoholic beverages – A review. Food Chemistry, 107(1), 449–463. doi:10.1016/j.foodchem.2007.08.058 Poinot, P., Grua-Priol, J., Arvisenet, G., Rannou, C., Semenou, M., Le Bail, A., & Prost, C. (2007). Optimisation of HS-SPME to study representativeness of partially baked bread odorant extracts. Food Research International, 40(9), 1170–1184. doi:10.1016/j.foodres.2007.06.011 Ryan, D., Prenzler, P. D., Saliba, A. J., & Scollary, G. R. (2008). Te signifcance of low impact odorants in global odour perception. Trends in Food Science & Technology, 19(7), 383–389. doi:10.1016/j.tifs.2008.01.007
53 San-Juan, F., Pet’ka, J., Cacho, J., Ferreira, V., & Escudero, A. (2010). Producing headspace extracts for the gas chromatography–olfactometric evaluation of wine aroma. Food Chemistry, 123(1), 188–195. doi:10.1016/j.foodchem.2010.03.129 Shepherd, G. M. (2006). Smell images and the favour system in the human brain. Nature, 444(7117), 316–21. doi:10.1038/nature05405 Steinhaus, M., Sinuco, D., Polster, J., Osorio, C., & Schieberle, P. (2009). Characterization of the key aroma compounds in pink guava (Psidium guajava L.) by means of aroma re- engineering experiments and omission tests. Journal of agricultural and food chemistry, 57(7), 2882–2888. doi:10.1021/jf803728n Van Ruth, S. M., Geary, M. D., Buhr, K., & Delahunty, C. M. (2004). Representative sampling of volatile favor compounds: Te model mouth combines with gas chromatography and direct mass spectrometry. In K. D. Deibler & J. Delwiche (Eds.), Handbook of Flavor Characterization: Sensory Analysis, Chemistry, and Physiology (pp. 303–311). New York: CRC Press. Wilson, D. A., & Stevenson, R. J. (2006). Learning to smell: olfactory perception from neurobiology to behavior. Baltimore, MD: Johns Hopkins University Press.
Acknowledgments
We thank Jim McCurry and Agilent Technologies for the auxiliary pressure controller and John Torngate for helpful discussions in early stages of this project. We thank the sensory judges who participated in this study.
54 Supplementary Information:
Supplementary Figure S2.1 Te chromatogram of mixture O2. Compounds eluting between 16 and 25 minutes were vented to waste by the Deans Switch and were consequently excluded from the smelled mixture and not sent to the mass spectrometer.
Supplementary Figure S2.2 Te chromatogram of mixture P5. Compounds eluting between 0 and 25 minutes and 32 and 40 minutes were vented to waste by the Deans Switch and were consequently excluded from the smelled mixture and not sent to the mass spectrometer.
55 Supplementary Figure S2.3 Alternate views of correspondence analysis (fgure 4) incorporating the frst 3 dimensions of variation. 30.57% of variance explained by dimension 1 (x), 22.84% of variance explained by dimension 2 (y), 14.03% of variance explained by dimension 3 (z).
56 Supplementary Table S2.4 Tentative identifcation of lavender volatile compounds. Volatiles were identifed by matching their mass spectra to the NIST 05 Mass Spectral Library (National Institute of Standards and Technology, Gaithersberg, MD) and to chemical standards, as noted. Te table is divided by cut time for perceptual mixtures P1-P6. Retention Time Retention Identif- Perceptual Mixture Compound CAS Number Index cation Group 7.23 3-hexen-1-ol 544-12-7 893 standard P1 7.64 1-hexanol 111-27-3 900 standard P1 9.82 alpha-thujene 2867-05-2 940 library P1 10.15 alpha-pinene 80-56-8 946 standard P1 10.79 camphene 79-92-5 958 standard P1 11.83 sabinene 3387-41-5 978 library P2 12.00 beta-pinene 127-91-3 982 standard P2 12.45 3-octanone 106-68-3 990 standard P2 12.57 myrcene 123-35-3 993 standard P2 12.72 not identifed 996 P2 13.15 alpha-phellandrene 99-83-2 1004 library P2 13.50 hexyl acetate 142-92-7 1011 standard P2 13.23 not identifed 1006 P2 13.65 3-carene 13466-78-9 1014 standard P2 14.52 not identifed1 1032 P2 14.67 eucalyptol 470-82-6 1035 standard P2 15.41 trans-beta-ocimene 3779-61-1 1050 library P2 15.79 cis-beta-ocimene 3338-55-4 1058 library P2 16.06 gamma-terpinene 99-85-4 1064 library P3 16.44 beta-terpineol 138-87-4 1072 library P3 17.34 terpinolene 586-62-9 1091 standard P3 18.90 linalool 78-70-6 1125 standard P3 18.62 octen-1-ol acetate 32717-31-0 1119 standard P3 19.04 3-octylacetate 4864-61-3 1128 library P3 19.64 allo-ocimene 7216-56-0 1142 library P3 20.21 lavandulol 507-70-0 1155 library P3 20.43 camphor 464-49-3 1160 standard P3 22.08 terpinen-4-ol 562-74-3 1199 standard P4 22.66 hexyl butyrate 2639-63-6 1212 library P4 23.20 alpha-terpineol 98-55-5 1225 standard P4 24.34 bornyl formate 7492-41-3 1253 library P4 26.53 linalyl acetate 115-95-7 1309 library P5 27.28 isobornyl acetate 125-12-2 1328 library P5 27.42 lavandulyl acetate 25905-14-0 1332 library P5 27.71 geranyl acetate 105-87-3 1340 library P5 30.63 neryl acetate 141-12-8 1420 library P5 31.67 Not identifed 2 1449 P5 57 Supplementary Table S2.4 continued 32.85 alpha-bergamotene 17699-05-7 1484 library P6 33.65 alpha-santalene 512-61-8 1507 library P6 33.78 beta-caryophyllene 87-44-5 1511 standard P6 34.48 beta-farnesene 77129-48-7 1532 standard P6 34.99 alpha-bisabolene 17627-44-0 1547 library P6 35.16 germacrene D 37839-63-7 1553 library P6 35.73 alpha-amorphene 483-75-0 1570 library P6 1possible terpene 2possible acetate ester
58 Chapter 3: Volatile and Sensory Profling of Cocktail Bitters
Introduction Extracting plant matter into alcohol is an ancient process, see, for example, the Hippocratic wine of the Greeks (Tonutti et al. 2010). As distilled liquor became more widely available, it was put to use in making plant extractions, mostly for medicinal purposes. Stoughton’s Great Cordial Elixir, a distilled-alcohol based herbal extraction bittered with gentian root became available commercially in 1690. While this was a patent medicine marketed for its medicinal properties it is the closest ancestor of what we today know as bitters. While the Elixir could be taken straight, it was ofen diluted into wine to make “instant” Purl-royal, a popular drink resembling vermouth, and was also ofen subsequently mixed with straight or burnt brandy (brandy with sugar added and reduced in alcohol by igniting it). Adding Stoughton’s Elixir to a dram of brandy yielded a “bitter draught” that was to be administered medicinally; recreational mixing soon followed (Wondrich 2007). Two brands of bitters on the market today, Angostura and Peychaud’s, have been in production since the nineteenth century; the Angostura company was founded in 1824 and Peychaud’s in 1837 (Parsons 2012). Angostura is ofen considered a “default” bitters in a cocktail recipe when a specifc type is not called for, and Peychaud’s bitters have a famously anise-heavy aroma and bright red color, and are used in the so-called signature cocktail of New Orleans, the Sazerac (Simmons 2011). Several named types of bitters are called for in Jerry Tomas’ Te Bon Vivant’s Companion or How to Mix Drinks, the frst book on cocktails (Grimes 2002), frst published in 1862. Tese named bitters include Boker’s (or Bogart’s), Stoughton’s, and “Jerry Tomas’ Own Decanter Bitters”, with recipes provided for the bitters themselves. While the commercial production of many popular nineteenth-century bitters, such as Boker’s, Stoughton’s, and Abbot’s, ceased during or relatively soon afer Prohibition (Parsons 2012), Boker’s and Jerry Tomas’ Own Decanter Bitters are both currently commercially available, having been re- 59 created from historical recipes (Table 1). Boker’s, Angostura, and Decanter bitters are commonly informally categorized as “aromatic” styles, with spice favors such as cinnamon, cloves, and cardamom (Clarke 2010). Peychaud’s and other brands inspired by it are ofen noted as either a subtype of aromatic bitters or a separate style, due to anise favors that are more intense than in other aromatic bitters (Clarke 2010, Parsons 2012, Bovis 2012, Sandham 2012). A third important “classic” style is citrus bitters, especially orange bitters (Clarke 2010, Parsons 2012); Regan’s Orange Bitters (Regan 2002), a popular contemporary commercial brand (Parsons 2012), were based on a recipe from another historical manual, Charles H Baker’s Te Gentleman’s Companion: Being an exotic drinking book; or, Around the world with jigger, beaker and fask, published in 1939. Baker also called for the use of celery bitters in the recipe for a cocktail called the Fourth Regiment; celery bitters is another style that had died out commercially until relatively recently (Baker 1939). Along with commercial reintroduction of defunct historical styles of bitters since the early 2000’s, there has been an introduction of many new styles and types of bitters, driven in part by a rise in bartenders developing their own bitters in-house (Sandham 2012, Parsons 2012, Clarke 2011). While these “new bitters” have a range of ingredients and favors with nearly indefnable boundaries, a number of recently invented styles have gained prominence. “Xocolotl Mole Bitters”, produced by the company Bittermen’s starting in 2007 based on the chocolate, chile, and cinnamon favor profle of Mole Poblano, has been widely praised in print and used widely at craf cocktail bars (Parsons 2012, Clarek 2012). A similar “new classic” trend is more heavily-spiced bitters intended for Tiki drinks, which ofen use spice-heavy ingredients such as falernum (and almond and clove syrup) and pimento dram (an allspice liqueur) in conjunction with robust Jamaican or Agricole rums. With these somewhat loose categories as a guide, at least two commercially available examples of each style (aromatic, citrus, New Orleans-style, tiki, mole, celery) were included in the samples for this study (Table 3.2).
Table 3.1 shows ingredients commonly used in making historical and contemporary bitters. Specifc historical recipes for Boker’s, Stoughton’s, Orange, and Jerry Tomas’ Own
60 Table 3.1: Ingredients used in historical and contemporary recipes for bitters, listed by taxonomic name and literature source
Component Source1 Taxonomic name Component Source1 Taxonomic name orange T,W,O,J,B Citrus sinensis black walnut leaf P Juglans nigra Elettaria cardamom T,P,W,O,B cardamomum burdock P Arctium lappa quassia T,P,W,O,B Quassia amara chirayata P Swertia chirayita Syzygium clove T,P,W,O,J aromaticum dandelion P Taraxacum ofcinale gentian T,P,W,O,S Gentiana spp. devil’s club P Oplopanax horridus calamus T,P,W,B Acorus calamus fennel P Foeniculum vulgare allspice T,P,W,J Pimenta dioica fringe tree P Chionanthus spp Coriandrum coriander T,P,O sativum horehound P Marrubium vulgare Cinnamomum cinnamon P,W,J verum hyssop P Hyssopus ofcinalis raisins T,W,J Vitis vinifera juniper P Juniperus spp snake root T,W,J several species lavender P Lavandula angustifolia Cinnamomum cassia P,W,S verum lemongrass P Cymbopogon citratus Matricaria chamomile T,W,S chamomilla licorice P Glycyrrhiza glabra catechu W,B Senegalia catechu milk thistle P Silybum marianum caraway P,O Carum carvi peppermint P Mentha x piperita cinchona P,O Cinchona spp rose hip P Rosa spp lemon T,W,J Citrus limon sarsparilla P Smilax regelii bitter orange W,S Citrus aurantium sassafras P Sassafras spp calumba W,S Jateorhiza palmata schizandra P Schisandra chinensis angelica T,P Angelica sylvestris star anise P Illicium verum Aframomom grains of paradise T,P melegueta wild cherry wood P Prunus avium hops T,P Humulus lupulus aloe T Aloe spp nutmeg T,P Myristica fragrans centaurium T Centaurium erythraea orris T,P Iris spp galanga root T Alpinia spp Artemisia wormwood T,P absinthium ginger T Zingiber ofcinale anise T,P Pimpinella anisum myrrh T Commiphora myrrha arnica P Arnica spp polypody T Polypodium spp barberry P Berberis vulgaris safron T Crocus sativus birch leaf P Betula spp 1 T=Tomas 1862; P=Parsons et al. 2011; W=Wondrich, 2007; O=Orange bitters (Regan 2003); J=Jerry Tomas’ Own Decanter Bitters, (Tomas 1862); S=Stoughton’s bitters (Wondrich 2007); B=Boker’s bitters, (Wondrich 2007).
61 Decanter Bitters are included; most contemporary makers of bitters do not publicize their formulas, so ingredients listed in contemporary recipes are included. In some cases there is quite a bit of overlap; for example, many aromatic bitters recipes include citrus peel, and the included orange bitters recipe uses several spices. Te ultimate favor profle of any of these is therefore likely more dependent on proportions of ingredients than use of specifc ingredients. Te earliest written description of the cocktail comes from an 1806 edition of the Hudson, New York newspaper Te Balance and Columbian Repository: “Cock Tail, then, is a stimulating liquor, composed of spirits of any kind, sugar, water and bitters” (Sampson et al. 1806). While spirits had been imbibed sweetened and diluted with water, in a drink called the “sling”, for some time before, this was the frst time that the cocktail, as a separate drink and including bitters as an ingredient, was codifed in writing (Wondrich 2007). Two conclusions can be made from this: the cocktail is almost as old as America itself, and bitters are its defning component, setting it apart from other spirit-based drinks. In cocktail-making (or mixology, which curiously enough is not a recent neologism but rather a term introduced facetiously in the Knickerbocker Magazine in 1856 [Wondrich 2007]) bitters are usually used at a maximum concentration of 1-2% of the total volume of the drink. Bitters are probably most familiar in drinks like the Old-Fashioned (Whiskey, sugar, and bitters) and the Manhattan (Whiskey, sweet vermouth, and bitters), though they were considered an essential component in the Martini until the middle of the twentieth century, and in current practice are added to a wide variety of similarly spirit-forward drinks as well as citrus-lightened sours. Indeed, while cocktail bitters are formulated to be concentrated and bitter to the point that they are considered non-potable unless heavily diluted, there is a growing corpus of drinks recipes that makes use of, for example, two ounces of angostura bitters as their most abundant ingredient (Baker 1939). But setting aside these extreme cases, bitters are more generally used to add aroma complexity to a cocktail, to complement and contrast the favors already present in the component liquors, and, by selecting diferent styles, to subtly alter the favor of the same base cocktail without changing its essence.
62 Table 3.2 Bitters samples used in the study, with historical sources and precedents, and style noted.
Name Brand Type Abbreviation Code
Boker’s Bittersa Dr. Adam Elmegirab Aromatic BOKERS A1
Angostura Bittersb Angostura Aromatic ANGOSTURA A2
Jerry Tomas’ Own Bitter Truth Aromatic JTDECANTER A3 Decanter Bittersc
WHISKEY BARREL- Whiskey Barrel-Aged Bitters Fee Brothers Aromatic A4 AGED Regan’s Orange Bitters Bufalo Trace Citrus REGAN’S ORANGE C1 Number 6d
Hopped Grapefruit Bitters Bittermen’s Citrus HOP-GRAPEFRUIT C2
Grapefruit Bitters Scrappy’s Citrus SCRAPPY GRAPEFRUIT C3
Orange Bitters Scrappy’s Citrus SCRAPPY ORANGE C4
Xocolatl Mole Bitters Bittermen’s Mole XOCOLOTL MOLE M1
Mole Bitters Bitter Truth Mole BT-MOLE M2
‘Elamakule Tiki Bitters Bittermen’s Tiki ELAMAKULE-TIKI T1
Jamaica Bitters Bittercube Tiki JAMAICA T2
Creole Bitters Bitter Truth New Orleans BT-CREOLE NO1
Peychaud’s Bittersb Peychaud New Orleans PEYCHAUD NO2
Orchard St Celery Bittermen’s Celery BMCELERY C1
Celery Bitters Scrappy’s Celery SCRAPPYCELERY C2
a=based on historical recipe for now-defunct Boker’s brand (Parsons et al. 2012). b=19th-century brand (Parsons et al. 2012). c=based on historical recipe from Te Bon Vivants Companion or How to Mix Drinks (Wondrich 2007). d=based on historical recipe from Te Gentleman’s Companion: Being an Exotic Drinking Book or Around the World with Jigger, Beaker and Flask (Baker 1939, Regan 2003).
63 Despite the long timeline of bitters, and the cultural role they have played both historically and in their current resurgence, they have not been the subject of any published study of their chemical composition or sensory properties. Terefore, the objectives of this study are to describe, map, and analyze the favor chemistry of the most common styles of bitters currently available (16 commercial bitters refecting six common categories [see Table 3.2] are used, as it would be unfeasible to capture every available product) using volatile profling via Gas Chromatography-Mass Spectrometry, sensory descriptive analysis with trained panelists, and multivariate statistical analysis to reveal product-descriptor and sensory-chemical correlations.
Materials & Methods Samples 16 bitters (Table 3.2) were purchased from Astor Wines & Spirits (New York, NY), Cask (San Francisco, CA), Amor y Amargo (New York, NY), and Union Square Liquors (New York, NY). Chemical Analysis A 200 μL aliquot of bitters was pipetted into 10 mL of water in 20 mL amber glass headspace vials (Agilent Technologies, Santa Clara, CA) capped with magnetic, PTFE-lined silicone septa headspace caps. 2-undecanone was used as an internal standard at 50 μg/L (Sigma-Aldrich, St Louis, MO). A conditioned, 2-cm long PDMS-DVB-Carboxen SPME fber (Supelco, Bellefonte, PA) was introduced into the headspace of the vial for 40 minutes at 25ºC with rotational shaking at 250 RPM. A Gerstel MPS2 autosampler performed the extraction and the injection (Gerstel, Linthicium, MD). Te fber was removed from the headspace of the vial and immediately introduced into the inlet of an Agilent model 6890 GC-single quadrupole-MS (Agilent Technologies) with a DB-WAX column (30 meters long, 0.25 mm ID, 0.25 μm flm thickness) (J&W Scientifc, Folsom, CA). Te inlet was held at 250ºC with a 10:1 split. Te carrier gas was helium, with a 1 mL/minute constant fow rate. Te starting oven temperature was 40ºC, held for 3 minutes, followed by a 2ºC/minute ramp until 180ºCwas reached, then the ramp was
64 increased to 30ºC/minute until 250ºC was reached, and held for 3 minutes. Te total runtime was 47 minutes. Te mass spectrometer had a 1.5-minute solvent delay and was run in scan mode with Electron Impact Ionization at 70eV, from m/z 40 to m/z 300. Te samples were analyzed in triplicate. Kovats retention indices were calculated using C8-C20 hydrocarbon mixture (Sigma- Aldrich) retention times for identical instrumental conditions. Peak identifcations were made by matching the background-subtracted average mass spectrum across half peak height for each peak to the NIST 05 mass spectral database, followed by verifcation by retention index and pure standards where available. Following identifcation, GC peaks were manually integrated and converted into headspace concentration in μg/L 2-undecanone equivalents by dividing the analyte peak area by the 2-undecanone peak area.
Sensory Analysis A descriptive analysis procedure was used to profle the sensory characteristics of the bitters. A group of 14 panelists, 10 Male, 4 Female, ages 21-35 were recruited from a pool of students and postdoctoral scholars in the department of Viticulture and Enology at the University of California, Davis. Over four training sessions, the panelists met in groups, smelled the bitters blind, and generated, discussed, and pooled descriptors by consensus until a fnal list of 30 terms was agreed upon. Samples were presented as 400 uL bitters in 20 mL deionized water in opaque black wineglasses. In the frst training session, four of the bitters were smelled and discussed; in the second, third, and fourth sessions, six bitters were smelled and discussed, so that each bitters was smelled at least once during the training. References (Table 3.3) were made for each descriptor, and these were smelled and refned over the second, third, and fourth sessions. Over two additional sessions, the descriptors and references were fxed and the panelists analyzed the intensity of each descriptor for each bitters in a training exercise in sensory booths. Te descriptive analysis proper was performed in triplicate, with each panelist smelling each reference, then rating the intensity of the aroma of each reference in each bitters over six sessions on an unstructured 9-cm line scale from “low intensity” to “high intensity.” Te panelists were
65 Table 3.3 Sensory terms and references used in the descriptive analysis on bitters.
Descriptor Reference aroma intensity overall intensity cardamom 4 crushed green cardamom pods grapefruit 2 cm * 8cm strip fresh grapefruit peel, oils manually expressed into glass frst molasses 10 mL molasses chocolate 10g shaved dark chocolate (Valhrona) celery seed 2g celery seeds, crushed cola 20 mL cola soapy 1 g unscented ivory soap root beer 20 mL root beer (Virgil’s) orange candy 5 orange jelly beans, halved (Jelly Belly) green 2 g each fresh cilantro leaf, fennel, and cucumber, brunoised tea 2 g black English Breakast tea leaves (Peet’s) brown sugar 5 g brown sugar (C&H) lime peel 1 cm * 5 cm strip fresh lime peel, oils expressed manually into glass frst black pepper 6 black peppercorns, lightly crushed alfalfa hay 1g alfalfa juniper 3 dried juniper berries, crushed mint 2 peppermint and 2 spearmint leaves (fresh) ginger 2cm * 2cm * 3mm piece of fresh ginger, minced orange peel 2 cm * 8cm strip fresh orange peel, oils manually expressed into glass frst earthy 5 g freshly dug soil with 2 mL water dried fruit 10 golden raisins, 2 dried apricots, 2 dried cherries (Trader Joe’s) anise 1 star anise pod cinnamon 2 g cinnamon powder wood 2g medium toasted oak chips (Evoak) clove 3 cloves nutmeg 1 g shaved nutmeg pod caraway 1 g caraway seeds, lightly crushed vanilla 2 cm length vanilla pod + 500 uL vanilla extract (Nieman-Massey) chile 1 g dried ancho chile, chopped
66 presented with eight samples per session, in lidded, opaque, black wineglasses under red light with random 3-digit codes as labels in a Williams Latin Square presentation design. Descriptive analysis was performed using FIZZ (Biosystemes, Couternon, France).
Statistical Analysis Te sensory data was subjected to a 3-way analysis of variance (ANOVA) with 2-way interactions for all 30 descriptors in the R statistical package. Te main efects were product, judge, and replicate. For descriptors with a signifcant Judge*Product interaction, a pseudo- mixed model (with Mean Square of Judge*Product replacing Mean Square of Error in the F-value calculation for Product efect) was used. Products were considered signifcantly diferent in a given aroma when p<0.05. For signifcant descriptors, mean values for each descriptor-product pair were calculated over all judges and replicates and a principal component analysis (PCA) was performed on the mean data in R. Mean ratings for each descriptor and normalized mean relative headspace concentrations of each volatile in each sample were analyzed with Partial Least Squares Regression (PLS) in Unscrambler, with the sensory data being set as the dependent variable to the independent-variable headspace volatile dataset. For purposes of data interpretation, aroma descriptors referenced for individual compounds were taken from the Perfavory and Flavornet websites (Luebke 2014, Acree and Arn 2004).
Results & Discussion
Sensory Analysis Te panel agreed upon 30 aroma descriptors for the bitters, listed in table 3.3 with their references. Of these, all except nutmeg were signifcantly diferent among samples by the ANOVA (p<0.05). Many of the descriptors, such as clove, cinnamon, and celery seed refected ingredients commonly used in bitters. Others, such as cola, soapy, root beer, and earthy, refected non- ingredient aromas that may arise from perceptual blending of the mixtures.
67