BARRANGOU, LISA. Sensory texture and fundamental rheology of agar and agarose

Transcription

1 Abstract BARRANGOU, LISA. Sensory texture and fundamental rheology of agar and agarose gels. (Under the direction of Professor E. Allen Foegeding and Professor Christopher R. Daubert). Texture properties of foods are an important component of food quality perception and acceptability. In order to design specific textures with predictable sensory attributes, a molecular understanding of food structures and their corresponding texture is necessary. Fundamental rheological methods are valuable tools for investigating structural mechanisms because they are based on physical and chemical theory, and when combined with descriptive sensory analysis, structure-function relationships can be established. The overall objective of this research was to utilize model food systems to further elucidate how physical properties of foods relate with the dynamic sensory perception of texture. Agar and agarose gels were used as model food gel systems. Initially, rheological profiles of agarose gels were developed, including linear, non-linear and fracture properties. Gel properties were examined under conditions of varying agarose concentration, solvent conditions, and strain rate. Increasing concentrations of agarose produced an increasingly stronger, more brittle network, while increasing concentrations of glycerol produced an increasingly stronger, more deformable network. All fracture properties and non-linear behaviors increased with increasing strain rate in a similar manner, suggesting a general mechanism responsible for strain rate effects that is similar

2 for non-linear and fracture behavior. Additionally, a new model was proposed to reliably describe and quantify non-linear behavior. Descriptive analysis was used to quantify the perceived hand texture characteristics of agarose gels, and results were compared with fundamental rheological profiles to determine if structure-function relationships could be established. Sensory small-strain and fracture causing forces were capable of differentiating the gels equally as well, indicating that relative gel strength was perceived similarly with non-destructive and fracture causing deformations. Surprisingly, hand force terms correlated more highly with fracture modulus (fractures stress / fracture strain) values (r 0.98, p 0.001) than fracture stress values (r = , p 0.05), suggesting sensory perception of force includes a coupling of stress and strain. Additionally, sensory deformation perceived at fracture correlated highly with fracture strain values (r = 0.98, p 0.001). Small-strain rheological tests could not distinguish gels as sensitively as fracture properties, indicating that fracture properties relate to sensory texture better than small-strain rheological properties. Descriptive sensory analysis and fundamental large-strain rheological methods were also used to characterize texture characteristics of agar gels. Gels were differentiated in the same manner by sensory texture analysis in the mouth and rheological properties (p 0.05), and significant correlations between sensory and rheological properties were reported. First bite and chew-down sensory terms highly correlated with each other and with fracture properties. Specifically, the first bite sensory term of force required to cause fracture correlated well with the chew-down sensory term chewiness (r > 0.99, p ), and both of these sensory terms highly correlated with

3 the fundamental rheological property, fracture modulus (r > 0.94, p 0.05). The first bite sensory term deformability perceived at fracture highly correlated with fracture strain values (r = 0.88, p 0.05), while the chew down property of particle breakdown negatively correlated with fracture stress values (r = -0.97, p 0.05). Additionally, the sensory properties that contribute to perception of strain-hardening were determined, and were found to correlate with non-linear rheological behavior, which is an important first step in understanding how non-linearity influences sensory perception of texture. These findings clearly demonstrate that fundamental large-strain rheological properties give valuable information toward the understanding of sensory perception of physical properties of foods.

4 SENSORY TEXTURE AND FUNDAMENTAL RHEOLOGY OF AGAR AND AGAROSE GELS by LISA BARRANGOU A dissertation submitted to the Graduate Faculty of North Carolina State University in partial fulfillment of the requirements for the Degree of Doctor of Philosophy FOOD SCIENCE Raleigh 2005 APPROVED BY: Dr. E. Allen Foegeding Chairman of Advisory Committee Dr. Christopher R. Daubert Co-chairman of Advisory Committee Dr. MaryAnne Drake Dr. Den Truong

5 Biography Lisa Barrangou was born on August 23, 1976 in Inglewood, California. She was raised in California and graduated from Fairfield High School in She attended Cornell University (Ithaca, NY) between 1994 and 1998 and obtained a Bachelor of Science degree in Nutrition, Food and Agriculture. In June 1998 she began her Master of Science program in the Department of Food Science at North Carolina State University (Raleigh, NC), under the direction of Dr. Roger McFeeters in the USDA Vegetable Fermentation Laboratory. Upon graduating in December 2000, she began working as a Scientist at Pepsi-Cola Company (Valhalla, NY) in the area of Product Development. In July 2002, she returned to North Carolina State University (Raleigh, NC) to begin her Ph.D. program in the Department of Food Science, under the direction of Dr. E. Allen Foegeding. She married Rodolphe Barrangou December 12, 2002 on the beautiful island of St. Lucia. Lisa, Rodolphe, Buddha and Zola (their beloved dogs) will be moving to Madison, Wisconsin in February 2005 to continue their professional careers as scientists in the food industry. ii

6 Acknowledgements I would first like to thank Dr. E. Allen Foegeding, for inviting me to work in his laboratory in pursuit of my Ph.D. I also extend my gratitude to Dr. Christopher R. Daubert, Dr. MaryAnne Drake, and Dr. Den Truong for serving on my graduate advisory committee. Thank you for allowing me to work independently, and for pushing me to be the scientist that I have become. I would also like to thank all of my friends and coworkers from the Foegeding Lab: Paige, Jack, Jessica, Dany, Mandy and Matt. Thanks for making the lab a wonderful and entertaining work environment. Finally, a very special thank you to my husband, Rodolphe, for knowing my potential before I had the foresight to see it, for encouraging me to come back to school for this degree, and for making my life more enjoyable and happy than I ever imagined it could be. iii

7 Table of contents LIST OF TABLES. VII LIST OF FIGURES. VIII LIST OF SYMBOLS AND ABBREVIATIONS. XII CHAPTER 1 LITERATURE REVIEW INTRODUCTION THE FOUNDATION OF TEXTURE RESEARCH Classification of texture characteristics Development of standard rating scales Texture Profile Method Instrumental texture profile analysis SENSORY ANALYSIS OF TEXTURE INSTRUMENTAL ANALYSIS OF TEXTURE Imitative and empirical tests Fundamental rheological tests RELATING FUNDAMENTAL RHEOLOGICAL PROPERTIES TO SENSORY TEXTURE PHYSIOLOGICAL STUDIES MODEL SYSTEMS CONCLUSIONS REFERENCES. 29 CHAPTER 2 TEXTURAL PROPERTIES OF AGAROSE GELS. I. RHEOLOGICAL AND FRACTURE PROPERTIES ABSTRACT. 34 iv

8 2.2 INTRODUCTION MATERIALS AND METHODS Materials Gel preparation Small-strain rheology Large-strain rheology Modeling of large-strain rheological data Statistical analyses RESULTS AND DISCUSSSION Small-strain rheology Fracture properties Large-strain modeling Strain-hardening behavior CONCLUSIONS REFERENCES. 61 CHAPTER 3 TEXTURAL PROPERTIES OF AGAROSE GELS. II. RELATIONSHIPS BETWEEN FRACTURE PROPERTIES, SMALL-STRAIN RHEOLOGY, AND SENSORY TEXTURE ABSTRACT INTRODUCTION MATERIALS AND METHODS Materials Gel preparation Sensory analysis Rheological analysis Statistical analyses RESULTS AND DISCUSSION Sensory analysis Fracture properties and relationships with sensory texture Small-strain rheology and relationhips with sensory texture CONCLUSIONS. 97 v

9 3.6 REFERENCES. 102 CHAPTER 4 SENSORY TEXTURE RELATED TO FRACTURE PROPERTIES AND LARGE-STRAIN RHEOLOGY OF AGAR GELS ABSTRACT INTRODUCTION MATERIALS AND METHODS Materials Gel preparation Descriptive sensory analysis Rheological analysis Statistical analyses RESULTS AND DISCUSSION Sensory analysis Rheological properties Sensory attributes related to rheological properties CONCLUSIONS REFERENCES. 137 vi

10 List of tables CHAPTER 2 1. MODELING PARAMETERS OBTAINED FROM TORSION CURVE FITTING. 64 CHAPTER 3 1. SENSORY HAND TEXTURE TERMS AND DEFINITIONS PEARSON CORRELATION COEFFICIENTS (r-values) BETWEEN SENSORY AND RHEOLOGICAL TERMS STORAGE MODULI (G ) COLLECTED AT Hz AND 1 Hz. 101 CHAPTER 4 1. SENSORY TEXTURE TERMINOLOGY AND SCALE USAGE PEARSON CORRELATION COEFFICIENTS (r-values) BETWEEN SENSORY AND RHEOLOGICAL TERMS. 136 vii

11 List of figures CHAPTER 2 1A. STORAGE MODULI (G ) OBTAINED FROM FREQUENCY SWEEPS OF GELS WITH VARYING CONCENTRATIONS OF AGAROSE (A, % W/W). 65 1B. STORAGE MODULI (G ) OBTAINED FROM FREQUENCY SWEEPS OF GELS WITH VARYING CONCENTRATIONS OF GLYCEROL (G, % W/W). 66 2A. PHASE ANGLES (δ) OBTAINED FROM FREQUENCY SWEEPS OF GELS WITH VARYING CONCENTRATIONS OF AGAROSE. 67 2B. PHASE ANGLES (δ) OBTAINED FROM FREQUENCY SWEEPS OF GELS WITH VARYING CONCENTRATIONS OF GLYCEROL. 68 3A. FRACTURE STRESS (σ f ) VALUES OF AGAROSE GELS AT DIFFERENT STRAIN RATES. 69 3B. FRACTURE STRAIN (γ f ) VALUES OF AGAROSE GELS AT DIFFERENT STRAIN RATES. 70 3C. FRACTURE MODULUS (G f ) VALUES OF AGAROSE GELS AT DIFFERENT STRAIN RATES MODELING OF AGAROSE GEL LARGE-STRAIN DATA WITH THE BST- EQUATION AND A POLYNOMIAL EQUATION. 72 5A. PREDICTED LARGE-STRAIN DATA FOR GELS WITH VARYING CONCENTRATIONS OF AGAROSE AT A STRAIN RATE OF 0.17 s B. PREDICTED LARGE-STRAIN DATA FOR GELS WITH VARYING CONCENTRATIONS OF AGAROSE AT A STRAIN RATE OF s viii

12 5C. PREDICTED LARGE-STRAIN DATA FOR GELS WITH VARYING CONCENTRATIONS OF AGAROSE AT A STRAIN RATES OF s A. PREDICTED LARGE-STRAIN DATA FOR AGAROSE GELS WITH VARYING CONCENTRATIONS OF GLYCEROL AT A STRAIN RATE OF 0.17 s B. PREDICTED LARGE-STRAIN DATA FOR AGAROSE GELS WITH VARYING CONCENTRATIONS OF GLYCEROL AT A STRAIN RATE OF s C. PREDICTED LARGE-STRAIN DATA FOR AGAROSE GELS WITH VARYING CONCENTRATIONS OF GLYCEROL AT A STRAIN RATE OF s COMPARISON OF PREDICTED VALUES OF SHEAR MODULI (G BST OR G POLY ) WITH VALUES OBTAINED FROM RAW DATA (G INITIAL ) AT A STRAIN RATE OF s SUBSTITUTION OF FRACTURE PROPERTIES INTO THE POLYNOMIAL EQUATION TO VERIFY THEIR PROPOSED RELATIONSHIP WITH THE ESTIMATED PARAMETER, k. 80 9A. EFFECTS OF VARYING AGAROSE CONCENTRATION ON STRAIN RATE DEPENDENCE OF NON-LINEARITY (k). 81 9B. EFFECTS OF VARYING GLYCEROL CONCENTRATION ON STRAIN RATE DEPENDENCE OF NON-LINEARITY (k). 82 CHAPTER 3 1A. HAND SMALL-STRAIN FORCE (HSF) SCORES FOR AGAROSE GELS (A = % W/W AGAROSE; G = % W/W GLYCEROL) B. HAND FRACTURE FORCE (HFF) SCORES FOR AGAROSE GELS C. HAND FRACTURE DEFORMATION (HFD) SCORES FOR AGAROSE GELS. 106 ix

13 2A. FRACTURE MODULUS (G f ) VALUES OF AGAROSE GELS B. LINEAR REGRESSION OF G f VALUES WITH HAND FRACTURE FORCE (HFF) AND HAND SMALL-STRAIN FORCE (HSF) SCORES A. FRACTURE STRAIN (γ f ) VALUES OF AGAROSE GELS B. LINEAR REGRESSION OF γ f VALUES WITH HAND FRACTURE DEFORMATION (HFD) SCORES FRACTURE STRESS (σ f ) VALUES OF AGAROSE GELS A. TEXTURE MAPS GENERATED FROM HAND TEXTURE TERMS (HFD = HAND FRACTURE DEFORMATION; HFF = HAND FRACTURE FORCE). _ 112 5B. TEXTURE MAPS GENERATED FROM FRACTURE PROPERTIES (γ f = FRACTURE STRAIN; G f = FRACTURE MODULUS). 113 CHAPTER 4 1A. SENSORY SMALL-STRAIN FORCE AND FRACTURE FORCE SCORES FOR AGAR GELS (A = % W/W AGAR; G = % W/W GLYCEROL) B. SENSORY DEFORMABILITY SCORES FOR AGAR GELS C. SENSORY PARTICLE BREAKDOWN SCORES FOR AGAR GELS D. SENSORY CHEWINESS SCORES FOR AGAR GELS A. FRACTURE STRESS (σ f ) VALUES OF AGAR GELS B. FRACTURE STRAIN (γ f ) VALUES OF AGAR GELS. 145 x

14 2C. SHEAR MODULI (INITIAL MODULUS (G i ) AND FRACTURE MODULUS (G f )) VALUES OF AGAR GELS D. STRAIN-HARDENING (k) VALUES OF AGAR GELS A. LINEAR REGRESSION OF SENSORY SMALL-STRAIN FORCE (SSF) AND FRACTURE FORCE (FF) SCORES WITH INITIAL MODULUS (G i ) AND FRACTURE MODULUS (G f ) VALUES, RESPECTIVELY B. LINEAR REGRESSION OF SENSORY DEFORMABILITY (DEF) SCORES WITH FRACTURE STRAIN (γ f ) VALUES C. LINEAR REGRESSION OF SENSORY PARTICLE BREAKDOWN (PB) SCORES WITH FRACTURE STRESS (σ f ) VALUES D. LINEAR REGRESSION OF SENSORY CHEWINESS (CHEW) SCORES WITH FRACTURE MODULUS (G f ) VALUES E. LINEAR REGRESSION OF A PROPOSED SENSORY CALCULATION FOR STRAIN-HARDENING WITH THE RHEOLOGICAL STRAIN-HARDENING CONSTANT (k) A. TEXTURE MAPS FOR AGAR GELS GENERATED FROM FRACTURE PROPERTY VALUES (γ f = FRACTURE STRAIN; G f = FRACTURE MODULUS) B. TEXTURE MAPS FOR AGAR GELS GENERATED FROM SENSORY ATTRIBUTE SCORES. 154 xi

15 List of symbols and abbreviations G Storage Modulus G Loss Modulus G* Complex modulus δ Phase angle σ Shear stress γ Shear strain γ& Shear strain-rate G Shear Modulus σ f γ f G f G initial G poly G BST k n hsf hss hff hfd ssf ff def pb chew Fracture stress Fracture strain Fracture modulus Shear modulus in the initial, linear portion of the curve Shear modulus estimated from the polynomial equation Shear modulus estimated from the BST-equation Estimated non-linear parameter constant from the polynomial equation Estimated non-linear parameter constant from the BST-equation Hand small-strain force Hand springiness Hand fracture force Hand fracture deformation Small-strain force Fracture force Deformability Particle Breakdown Chewiness xii

16 CHAPTER I - Literature review 1

17 1.1 Introduction Texture properties of foods are an important component of food quality perception and acceptability (Bourne, 1978). The International Organization for Standardization has defined texture as all the mechanical, geometrical and surface attributes of a product perceptible by means of mechanical, tactile and, where appropriate, visual and auditory receptors (ISO, 1992). Texture evaluation of food products is a complex and dynamic process because physical properties of foods change continuously throughout the sensory experience. Texture assessment begins from the initial sight of a food product, and continues through touch, initial ingestion, mastication and swallowing. Food product developers and manufacturers are challenged to formulate specific textures and mouthfeels, largely due to a limited understanding of the relationships between texture perception and food structures. In order to have a complete understanding of food texture, a multi-disciplinary approach must be taken, combining research from sensory studies, physiological studies, and physical and chemical properties of foods (Wilkinson et al., 2000). 1.2 The foundation for texture research Classification of texture characteristics Alina S. Szczesniak, while working at General Foods Corporation, was the pioneer in texture research who recognized the need to develop a system of texture nomenclature that could serve as a bridge between popular nomenclature and fundamental rheological 2

18 principles, and be applicable to a variety of food products. Szczesniak (1963) defined and classified texture characteristics into three main classes: (1) mechanical characteristics, which are determined by the reaction of the food to stress; (2) geometrical characteristics, which refer to the arrangement of constituents of the food; and (3) other characteristics, which refer mainly to moisture and fat content of foods and other mouthfeel factors. Szczesniak (1963) further classified mechanical characteristics into five primary parameters and three secondary parameters. The primary parameters include: (1) Hardness, defined as the force necessary to attain a given deformation, (2) Cohesiveness, defined as the strength of the internal bonds making up the body of the product, (3) Viscosity, defined as the rate of flow per unit force, (4) Elasticity, defined as the rate at which a deformed material goes back to its undeformed condition after the deforming force is removed, and (5) Adhesiveness, defined as the work necessary to overcome the attractive forces between the surface of other materials with which the food comes into contact (e.g. tongue, teeth, palate, etc.). The three secondary parameters include: (1) Brittleness, defined as the force with which the material fractures, and is related to the primary parameters of hardness and cohesiveness, (2) Chewiness, defined as the energy required to masticate a solid food product to a state ready for swallowing, and is related to the primary parameters of hardness, cohesiveness, and elasticity, and (3) Gumminess, defined as the energy required to disintegrate a semisolid food product to a state ready for swallowing, and is related to the primary parameters of hardness and cohesiveness. Geometrical characteristics were divided into two general groups of qualities: (1) those related to size and shape of particles, and (2) those related to shape and orientation. Civille & Szczesniak (1973) later changed the terms Elasticity to 3

19 Springiness, and Brittleness to Fracturability (definitions of original terms remained unchanged), and also changed the definition of cohesiveness to extent to which a material can be deformed before it ruptures. Szczesniak s work (1963) was the basis for three additional manuscripts which were published concurrently by scientists at General Foods Corporation. The classification of textural terms gave rise to a profiling method of texture description, the Texture Profile Method, which was applied to both sensory (Brandt et al., 1963) and instrumental measurements (Friedman et al., 1963), and also served as a basis for the development of quantitative rating scales (Szczesniak et al., 1963). The classification of texture terminology was intended to be used with both sensory and instrumental measurements of texture because it was thought that if the two methods used the same terminology and definitions, then correlations between sensory and instrumental measurement would be easily facilitated (Szczesniak, 1975a) Development of standard rating scales Standard rating scales of hardness, brittleness (fracturability), chewiness, gumminess and adhesiveness were established for quantitative sensory evaluation of food texture (Szczesniak et al., 1963). Scales for elasticity (springiness) and cohesiveness were not developed because these attributes were difficult to perceive in the mouth, and sensitivity of the judges was not acute enough to deem the establishment of broad scales. The scales developed were believed to cover the entire intensity range found in food products, and were designed to have the flexibility of expansion at any desired point to allow for 4

20 greater precision in a narrower range. Each point on each attribute scale is represented by a reference food sample, which makes it possible to assign a numerical rating to an unknown product by comparing it to a known reference (Szczesniak et al., 1963). Further development and modification of standard rating scales was later implemented by Munoz (1986), and included new reference scales for wetness, adhesiveness to lips, roughness, self-adhesiveness, springiness, cohesiveness of mass, moisture absorption, adhesiveness to teeth, and manual adhesiveness Texture profile method The classification of texture characteristics (Szczesniak, 1963) combined with the developed standard rating scales (Szczesniak et al., 1963) led to the development of a comprehensive sensory method for evaluating texture of food products, known as the Texture Profile Method (Brandt et al., 1963). The Texture Profile Method (also known as sensory Texture Profile Analysis, or TPA) was modeled after the Flavor Profile Method established by A.D. Little Company (Cairncross & Sjostrom, 1950), and was designed to evaluate the entire texture of a food product, from first bite through complete mastication. A texture profile was defined as the organoleptic analysis of the texture complex of a food in terms of its mechanical, geometrical, fat, and moisture characteristics, the degree of each present, and the order in which they appear from first bite through complete mastication (Brandt et al., 1963). Since texture follows a definite pattern regarding the order in which characteristics are perceived, the order of appearance principle (Cairncross & Sjostrom, 1950) was applied to texture profiling. The ordered phases 5

21 include: (1) First-bite, or initial phase, which includes mechanical characteristics of hardness, brittleness (fracturability) and viscosity, and any geometrical characteristics initially observed; (2) Masticatory phase, which includes the mechanical characteristics of gumminess, chewiness and adhesiveness, and any geometrical characteristics observed during chewing (attributes described in the masticatory phase are often referred to as chew-down terms); and (3) Residual phase, which includes changes made during mastication, such as rate of breakdown, type of breakdown, mouth-coating, and moisture absorption. The Texture Profile Method involves training a panel to use the texture terminology described by Szczesniak (1963) to descriptively analyze and quantify mechanical and geometrical characteristics of foods utilizing the standard rating scales developed by Szczesniak et al. (1963) Instrumental texture profile analysis An instrumental texture profiling method was also developed to complement the sensory Texture Profile Method. Instrumental TPA was first conducted with the General Foods Texturometer (Friedman et al., 1963). The Texturometer was designed to imitate the action of a jaw during the first bites of a food product. The instrument used a small flat cylindrical plunger to perform a two-cycle compression of a bite-size piece of food (usually 1.2 cm 3 ) to 25% of its original height. Force and displacement (time) data were generated, and analysis of the force-time curve led to the extraction of seven textural parameters for solid or semi-solid foods, five of which were measured, and two of which were calculated from the measured parameters: (1) Hardness, measured from the force- 6

22 time curve as the height of the first compression peak (chew); (2) Cohesiveness, measured as a ratio of area under the second compression peak (A 2 ) and the area under the first compression peak (A 1 ); (3) Elasticity (springiness), measured as C-B, where B is the distance (time) from the initial sample contact to the contact on the second chew and C is the same measurement made on a completely inelastic standard material such as clay; (4) Adhesiveness, measured as the area of the negative peak beneath the baseline of the force-time curve (A 3 ); (5) Brittleness (fracturability), measured as the height of the first significant break in the multi-peak shape of the first chew; (6) Chewiness, calculated as the product of hardness, cohesiveness and elasticity; and (7) Gumminess, calculated as the product of hardness and cohesiveness. For temperature sensitive foods it was later recommended that tests be performed at several temperature levels (Szczesniak, 1975b). Good correlations were initially found between instrumental and sensory evaluation of texture for the attributes hardness, brittleness, chewiness, gumminess, viscosity, and adhesiveness (Szczesniak et al., 1963). This instrumental TPA method was later adapted to the Instron Universal Testing Machine (Bourne, 1968), and is now commonly found in software form on any uniaxial compression/tension-based texture instrument. 1.3 Sensory analysis of texture The work discussed above is the foundation upon which texture research has since been conducted. Since then, many variations on the original Texture Profile Method (Brandt et al., 1963) and its training procedures (Civille & Szczesniak, 1973) have been developed for the evaluation of sensory perception of textural characteristics of food 7

23 products. The Texture Profile Method is one of many types of descriptive sensory analysis methods currently available to sensory scientists. Descriptive sensory analyses are aimed at profiling a product on all of its perceived sensory characteristics. Descriptive analyses are used to investigate not only texture, but also flavor, aroma, appearance, aftertaste and sound properties of products. Unlike affective tests, which involve untrained consumers and their perceptions of acceptability and texture concepts, descriptive sensory analyses involve the use of trained panelists whose responses are treated as instrumental data. Descriptive sensory tests involve the detection (discrimination) and description of both the qualitative and quantitative sensory components of consumer products (Meilgaard et al., 1999). Qualitative aspects include product attributes (e.g. hardness, cohesiveness, etc.), while quantitative aspects include the intensity/strength of the attributes. There are several different methods of descriptive analysis available, including the Flavor Profile Method, Texture Profile Method, Quantitative Descriptive Analysis (QDA), the Spectrum method, Quantitative Flavor Profiling, Free-choice profiling, and generic descriptive analysis (Murray et al., 2001). The specific methods reflect various philosophies and approaches. Generic descriptive analysis can combine different approaches from a variety of methods, and is frequently used during practical applications in order to meet specific project objectives (Murray et al., 2001). Critical aspects of descriptive texture analysis include panelist selection, scales and scale usage, and training. The first step is identifying panelists. Ideally, panelists should have previous training experience or at least an orientation to the process. A reasonable level of sensory acuity is also desired. The most important part of panelist selection is 8

24 their commitment and motivation to perform (Murray et al., 2001). The final panel selected should usually include a minimum of 10 individuals for sufficient statistical power. One of the most important aspects of training a panel is using a good texture language, or lexicon. An appropriate texture lexicon will include clear and precise definitions, non-redundant terminology, standardization procedures, standard order of term evaluation, references for each term, and scale anchors (Drake & Civille, 2003). Previously published lexicons may be used, but introduction of new terminology may be relevant, depending on the specific project. If a lexicon has not been identified prior to training, training should begin by having panelists evaluate a wide sample set in order to develop the appropriate texture language. Attributes and their definitions should be agreed upon by all panelists, and appropriate references should be identified. Once texture terminology and definitions are agreed upon, the panel must be trained to use a common frame of reference to define the product attributes and their intensities with regards to the specific products being investigated (Munoz & Civille, 1998). During training it should be reinforced that panelists are rating products in the context of all those which they have been exposed to during training sessions, not in the context of what they have personally experienced (Murray et al., 2001). A quantitative frame of reference represents the boundaries or limits that a panelist uses when rating the intensities of perceived attributes. During training, the quantitative frame of reference is set by selecting the perception that corresponds to the highest intensity on the chosen scale, and establishing that point as the upper limit. This intensity becomes the frame of reference for all other intensity evaluations (Munoz & Civille, 1998). Depending on how 9

25 the highest intensity of the frame of reference is chosen, a panel will rate products using either universal, product or attribute specific scaling techniques. Universal scaling is based on attribute intensities being rated on an absolute and universal basis. All intensity boundaries are established considering all products and intensities to define the highest intensity point on the scale. Since all product categories are considered, the selected highest intensity represents the absolute highest intensity. This point of reference is used by all panelists as the guideline to rate all attribute intensities for all products (Munoz & Civille, 1998). For example, a rating of 5 for sweetness in cheese would have the same intensity as a rating of 5 for sweetness in yogurt, candy, fruit, etc. Also, a 5 in sweetness would have the same relative intensity as a 5 in sour, or a 5 in any other attribute. Additionally, if a product had a 4 in salty and a 7 in sour, the product would be deemed more sour than salty. Product specific scaling, on the other hand, is based on attribute intensities being rated only within the product category being studied (Munoz & Civille, 1998). For example, a rating of 5 for sweetness in yogurt could not be compared to a rating of 5 for sweetness in any other product, because each product has its own unique scale. However, a 5 in sweetness would still be the same intensity as a 5 in sour or any other attribute measure for a specific product. Finally, attribute specific scaling is based on each attribute being rated independently from each other within a product, and therefore each attribute has its own scale and intensity references (Munoz & Civille, 1998). With attribute specific scaling a 5 in sweetness in yogurt could not be compared to sweetness in any other product, nor 10

26 could it be compared to an intensity of 5 for any other attribute tested within the same product. There are, of course, advantages and disadvantages to using each scaling approach. The primary advantages of each are as follows: Universal scaling is the most comprehensive approach, as it establishes a system for attribute intensities within a product and across products to be inter-compared. Product specific scaling is the best scaling method for evaluation of a single product category. Product specific scaling is less time intensive and easier to train that universal scaling. Attribute scaling is the closest method to what consumers follow in rating attributes. This method can be useful when there are only a few attributes to evaluate and the panel needs to focus on each attribute independently (Munoz & Civille, 1998). In addition to the various scaling techniques, there are three primary types of quantitative scales commonly used in descriptive analysis (Meilgaard et al., 1999): (1) Catergory scales, where words or numbers are constructed to maintain equal intervals between categories, and signify increasing sensation intensities. The 9-point hedonic scale is a common example of a category scale; (2) Line scales, which utilize an anchored 15 cm line where panelists rate their evaluation with a hash mark (no numbers are present); (3) Magnitude Estimation (ME) scales, which use numbers generated in direct proportion to the intensity of a stimuli relative to a reference. For example, if a sample is judged to be twice as hard as the given reference, and the reference was given a value of 10, then the sample would be given a value of 20. Clearly, the selection of scaling technique and scale usage should be carefully considered when proceeding with panel training. 11

27 Once panelists have been trained on the concepts, including attribute terminology, definitions and procedures, and they have learned how to utilize the selected scales, panelists must continually work on consistency and reproducibility of sample evaluation. A minimum of hours of training along with an experienced panel leader is usually required for adequate training in preparation for formal evaluations (Foegeding et al., 2003). 1.4 Instrumental analysis of texture Since 1963, numerous studies on a wide range of foods have used instrumental TPA alone or in combination with sensory analysis to evaluate textural properties. One of the main reasons for the development of instrumental measurement of textural properties is to alleviate the need for sensory analysis. Sensory perception of textural properties of foods, however, is a dynamic process which is difficult to mimic by instrumental means. Although instrumental measures of texture cannot completely imitate oral motion, rates of force application, or the effects of temperature and saliva (Jack et al., 1993), accurate measurement of physical properties of foods and determination of their relationships with the dynamic perception of texture can lead to a better understanding of structure-function relationships Imitative and empirical tests Imitative tests such as the popular instrumental TPA (Szczesniak et al., 1963; Bourne, 1978) and the relatively new Bi-cyclical Instrument for Texture Evaluation (BITE 12

28 master) (Meullenet et al., 1997) attempt to simulate the mechanical motions of biting or chewing, and are popular because they generate multiple texture parameters that correlate well with sensory texture properties. Similarly, empirical (non-imitative) instrumental tests such as puncture, shear, and extrusion-type tests have been found from practical experience to be correlated with textural quality (Bourne, 1978). Empirical tests provide single-point information, and include tests such as the shear-press, gelometers, viscometers, tenderometers, etc. (Friedman et al., 1963). Specifically, sensory perception and imitative instrumental determination of hardness have repeatedly been demonstrated to correlate very well (Montejano et al., 1985; Munoz et al., 1986; Meullenet et al., 1997; Meullenet et al., 1998; Breuil & Meullenet, 2001). Lower degrees of correlation, however, have typically been obtained for other terms such as cohesiveness, springiness, and chewiness (Lyon et al., 1980; Munoz et al., 1986; Casiraghi et al., 1989; Meullenet et al., 1997; Meullenet et al., 1998). Although imitative tests have been successful at generating instrumental parameters that correlate with sensory terms, and they can, to some extent, provide an instrumental alternative to sensory analysis, they are empirical in nature and provide no fundamental information about structure-function relationships with regards to texture Fundamental rheological tests Fundamental rheological tests, on the other hand, have the advantage of being linked to theories that explain molecular mechanisms. Fundamental rheological tests are based on physical and chemical theory, allowing the chemical properties of molecules to be 13

29 linked with bulk rheological properties of the system. Rheological tests determine how materials respond to force and deformation within a particular time frame, and this information is used to calculated standard rheological properties. Rheological properties are characteristic of specific materials and are therefore theoretically independent of the testing instrument. The goal of measuring fundamental rheological properties of foods is to measure physical properties of the system, not to mimic the human sensory process (Foegeding et al., 2003). Nevertheless, accurate measurement of physical properties of foods and determination of their relationships with the dynamic perception of texture can lead to a better understanding of structure-function relationships. Rheology is the study of deformation and flow of materials (Steffe, 1996). Rheological properties are based on relationships between stress and strain. Stress (σ) is a function of force applied to a material per unit area: Force σ = Area The direction of the force with respect to the surface area impacted will determine the type of stress experienced. Application of a force directly perpendicular to a surface will result in a normal stress, while application of a force parallel to a surface will result in a shear stress (Daubert & Foegeding, 1998). The direction of the applied stress with respect to the material surface will also determine the type of strain (normal or shear) experienced by the sample. Normal strain (ε c ) is calculated as the relative deformation of a sample and is dimensionless. Normal strain is represented as: ε = c L L i 14

30 where L i is the initial sample length, and L is the change in length that occurs upon deformation. Shear strain is represented as: γ = tan 1 L h where h is the sample height. Shear strain-rate (γ& ) is used to quantify strain during flow of a liquid, and is the degree of deformation with respect to time. Compressive and tensile modes of rheological analysis are common normal tests, while rotational rheometry, including small-strain oscillatory tests or torsion, are common shear tests. Elastic and viscous extremes determine the spectrum of mechanical properties of materials. Ideal elastic (solid) materials follow Hooke s Law, where stress and strain are linearly related. The constants of proportionality, used to equate stress with strain, are called moduli. Under normal stress conditions, the proportionality constant is called the elastic modulus (E): σ E = ε c while under shear stress conditions, the proportionality constant is called the shear modulus (G): σ G = γ Ideal viscous (fluid) materials follow Newton s law, where stress and strain-rate are linearly related. For Newtonian fluids, the viscosity function is constant and is called the Newtonian viscosity (µ): σ µ = & γ 15

31 Viscoelastic materials lie between the extreme boundaries of Hookean solids and Newtonian fluids, displaying properties of both solid-like and fluid-like behavior. Viscoelastic properties are dependent upon the rate at which a force is applied, therefore the relative timeframe used to record a measurement must always be taken into account. Rheological measurements may be characterized into three regions for viscoelastic solid foods (Foegeding et al., 2003): (1) the linear viscoelastic region (LVR), where the relationship between stress and strain is proportionate; (2) the non-linear viscoelastic region (NLVR), where the relationship between stress and strain deviates from linearity (stress either increases or decreases non-linearly with strain); and (3) the fracture region, where sample failure occurs. Small-strain rheological tests are typically used to probe the linear region, while large-strain rheological tests are typically used to probe the nonlinear and fracture regions Small-strain rheology Small-strain rheological tests are designed to non-destructively probe the linear region of a material. These tests are only valid within the linear region, and therefore Hooke s Law is always obeyed. Typical instruments used to conduct small-strain rheology include controlled stress or controlled strain rheometers. In a controlled stress rheometer, stress is applied and the resulting strain is measured, while in a controlled strain rheometer, strain is applied and the resulting stress is measured. Oscillation tests are commonly used to investigate network properties of materials. During oscillation, the response of a material to sinusoidally varying stresses or strains is recorded. A sinusoidal oscillation has an amplitude (maximum level of stress or strain) 16

32 and a duration, or frequency (time to complete one oscillatory cycle). For a controlled strain rheometer, the strain input function is: γ = γ sin ( ω ) 0 t where γ 0 is the strain amplitude, ω is the frequency, and t is time. The resulting stress is: σ = σ 0 sin( ωt δ ) where σ 0 is the stress amplitude, and δ is the phase angle (described below). The sample being measured during oscillation either stores energy or dissipates energy in the form of heat. An ideally elastic material will store all energy, while an ideally viscous material will dissipate and lose all energy. The quantity of stored energy is referred to as the elastic component, G, and the quantity of energy lost is referred to as the viscous component, G : G = G *cos( δ ) G = G *sin( δ ) where G* is the complex modulus, defined as the ratio of stress and strain amplitudes: σ 0 G * = γ 0 and δ is the phase angle, defined as the ratio of energy lost (G ) to energy stored (G ): G tan( δ ) = G Phase angle (δ) is a measure of the relative degree of viscoelasticity of a material. An ideally elastic material will have a phase angle of 0, while an ideally viscous material will have a phase angle of 90, and a viscoelastic material will have a phase angle somewhere in between. A material with a larger G component will behave more 17

33 elastically (solid-like), while a material with a larger G component will behave more viscous (fluid-like). It is important to remember that viscoelastic properties are dependent upon the rate at which a force is applied, therefore the relative timeframe used to record a measurement must always be taken into account. Silly putty is a common example displaying the concept of time dependence: Silly putty displays more viscous, fluid-like properties during slow stretching of the material, but it displays more elastic, solid-like properties when it is quickly bounced on the ground. Since most foods are viscoelastic materials, this concept of time dependence is specifically important when trying to identify relationships between instrumental measurements and sensory measurements of texture. Clearly, similar rates of force application should be applied for both methods in order to accurately compare data sets Large-strain rheology Large-strain rheological methods typically probe non-linear and fracture regions. Materials undergoing large-strain testing are deformed to an extent to which the material is damaged or fractured. Large-strain rheological methods are, therefore, particularly relevant to investigating food texture because all foods are subjected to large deformations in the mouth during processing (Borwankar, 1992). Large-strain rheological tests typically measure the forces and deformations associated with first bite during consumption, which represents 2 10% of the total normal mastication time (Bourne, 1975). 18

34 Compression, extension or torsion methods are examples of large-strain tests. Although uniaxial compression is the most popular large-strain rheological test used, one of the major limitations of this method is the potential for friction occurrence between the sample and instrumental plate. Friction could cause variations in sample deformation, or contribute directly to the stress measurement, consequently making the assumption of an expanding cylinder invalid (Gwartney et al., 2002). Another problem is that required strains to cause fracture can be so large that it becomes impossible to fracture the material (Hamann, 1983). These limitations can be avoided by using torsional methods. For torsion testing, samples are ground into a capstan shape with a known center diameter, and then twisted to fracture by holding one end of the specimen stable while rotating the opposite end. Torque and angular displacement values are used to calculate corresponding stress and strain values using the equations described by Diehl et al. (1979). For a capstan-shaped specimen the following equations apply: Shear stress (σ) at the minimum cross-section is calculated as: σ = M Qc where M is torque (Nm); Q is a shape factor constant for the mid-section of the specimen, equal to 8.35 x 10 6 m -3 ; and c is a constant defined as: 2K = π r Q c 3 min where K is a shape factor constant for the boundary of the minimum cross-section of the specimen, equal to 1.08; and r min is the radius of the smallest cross-section, equal to 5 mm. Shear strain (γ) at the minimum cross-section is calculated as: 19

35 2π ( rpm) γ = ct 60 Q Q + U where t is time (s); rpm is the rotational speed; and U is a shape factor constant for the ends of the specimen, equal to 1.09 x 10 6 m -3. The corrected shear strain (γ corr ) is a necessary correction for strains > 0.8, and is calculated as: 2 2 γ γ ln γ γ corr = Shear moduli (G) are calculated as: σ G = γ Strain rates (γ& ) at the minimum cross-section are calculated as: 2π ( rpm) γ& = c 60 The advantages of using torsion over extension or compression methods were described by Hamann (1991) as follows: (1) Torsion produces pure shear, which is a stress condition that does not change the sample volume even if the material is compressible; (2) The sample shape does not change during testing; (3) The calculated shear stress and strain values are true up to large twist angles; (4) There is no restriction on fracture criterion; the material can fail in shear, tension, compression, or a combination; (5) Maximum shear, tension, and compression stresses all have the same magnitude but act in different directions, making it easy to determine the type of failure due to the fracture plane; and (6) No friction is created between the specimen and the test fixture. 20

36 Large strain methods are valuable for providing meaningful fracture properties of materials, such as fracture stresses (σ f ), strains (γ f ) and moduli (G f ), but quantifiable information about the non-linear region is still relatively limited and ill-understood. Phenomenological models have been applied to large-strain behavior in attempts to quantify the extent of non-linearity, or deviation from ideal elastic behavior (Blatz et al., 1974; Peleg, 1984). The BST equation (Blatz et al., 1974) has been most recently applied to biopolymer gels to describe and quantify non-linear behavior, but interpretations of the parameters estimated from this model have varied (McEvoy et al., 1985; Bot et al., 1996; Groot et al., 1996). 1.5 Relating fundamental rheological properties to sensory texture Although the goal of measuring fundamental rheological properties of foods is to measure physical properties of the system, and not to mimic the human sensory process (Foegeding et al., 2003), accurate measurement of physical properties of foods and determination of their relationships with the dynamic perception of texture can lead to a better understanding of structure-function relationships. Numerous studies have attempted to investigate relationships between fundamental rheological properties and sensory perception of texture. Diehl and Hamann (1979) used both torsion and uniaxial compression to study relationships between sensory texture and fundamental rheological properties of raw potatoes, melons and apples. Shear stress at failure in torsion appeared to be a stronger indicator of sensory texture than shear stress at failure in uniaxial compression for all three materials tested. Moisture release upon first bite was found to be negatively 21

37 correlated with shear stress at failure in torsion for all materials (r > 0.73, p < 0.05), which suggested that cell turgor pressure provides an important contribution to texture. Shear stress at failure in torsion also highly correlated with firmness of apples (r = 0.94, p < 0.01) and melons (r = 0.73, p < 0.01). Chewiness of apples correlated equally well with failure shear stress in torsion and in compression, and with the torsion shear modulus and uniaxial compression elastic modulus (r = , p < 0.05). In another study, Hamann and Webb (1979) used a shear/compression cell mounted to an Instron Universal testing machine to measure failure force for heat coagulated fish pastes, and found that the maximum cell force was a good predictor of sensory springiness (R 2 = 0.84, p < 0.01), firmness (R 2 = 0.82, p < 0.01), cohesiveness (R 2 = 0.83, p < 0.01) and gel strength (R 2 = 0.83, p < 0.01). Montejano et al. (1985) showed that torsion shear strain at failure correlated well with sensory perception cohesiveness of heat-induced protein gels (r = 0.87, p < 0.01), while torsion shear stress at failure correlated with sensory firmness (r = 0.72, p < 0.01). Hamann and Lanier (1986) reported high correlations between torsion failure stress and sensory firmness (r = 0.82), and between torsion failure strain and sensory cohesiveness (r = 0.84) for surimi gels. Munoz et al. (1986) found that sensory cohesiveness of gelatin gels correlated with compression yield deformation values (r = 0.88, p < 0.05), while sensory cohesiveness did not correlate with mechanical cohesiveness as determined through instrumental TPA. Sensory analysis and uniaxial compression methods were applied to feta cheese and sensory firmness correlated with fracture stress (r = 0.96) and fracture modulus (r = 0.95) values (Wium et al., 1997). 22

38 Drake et al. (1999) investigated cheese texture through evaluation of sensory texture and small-strain rheological methods (frequency sweeps, creep). Storage moduli (G ), loss moduli (G ) and percent creep recovery significantly correlated with sensory firmness, cohesiveness, stickiness and slipperiness (r = , p < 0.05) of cheeses. Gwartney et al. (2002) reported higher fracture stress values in reduced-fat cheeses than in equivalent full-fat cheeses. Reduced-fat cheeses were also perceived to be more hard, chewy, waxy, fracturable and springy, and less sticky, cohesive, meltable and smooth than full-fat cheeses. Fracture stress values correlated well with sensory hardness (r = 0.77, p < 0.01) and chewiness (r = 0.79, p < 0.01) scores, while fracture strain values did not distinguish the cheeses. This work suggested that compositional and processing factors of cheese manufacturing can be varied to match fracture stress and sensory hardness and chewiness properties (Gwartney et al., 2002). Steiner et al. (2003) used descriptive analysis and rheological analysis to evaluate caramel texture and reported correlations between storage moduli (G ) and sensory hardness (r = 0.90, p < 0.01), storage moduli (G ) or viscosity and number of chews required to break down the sample (r = 0.94, p < 0.01), and between tack force and sensory stickiness (r = 0.80, p < 0.05). Sensory texture and rheological properties of whey protein emulsion gels were investigated (Gwartney et al., 2004), and fracture stress values significantly correlated with sensory firmness (r = 0.88, p < 0.001) and number of chews required to prepare the sample for swallowing (r = 0.87, p < 0.001). Additionally, fracture strain values correlated with compressibility (r = 0.87, p < 0.001). Small-strain rheological methods, which probe the linear viscoelastic region, and fracture analyses, have yielded important information regarding structural properties of 23