Sensory Analysis for Food and Beverage Quality Control: A Practical Guide (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Sensory analysis for food and beverage quality control © Woodhead Publishing Limited, 2010 Related titles Chemical de...

Author: David Kilcast

121 downloads 1503 Views 3MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

Sensory analysis for food and beverage quality control

© Woodhead Publishing Limited, 2010

Related titles Chemical deterioration and physical instability of food and beverages (ISBN 978-1-84569-495-1) For a food product to be a success in the marketplace, it must be stable throughout its shelf-life. Changes due to food chemical deterioration and physical instability are not always recognised by food producers – who are more familiar with microbial spoilage – yet can be just as problematic. This book provides an authoritative review of key topics in this area. Chapters in Parts I and II focus on the chemical reactions and physical changes which negatively affect food quality. The remaining chapters outline the likely effects on different food products, for example baked goods, fruit and vegetables and beverages. Understanding and measuring the shelf-life of food (ISBN 978-1-85573-732-7) The shelf-life of a product is critical in determining both its quality and profitability. This book reviews the key factors in determining shelf-life and how it can be measured. Part I examines the factors affecting shelf-life and spoilage, including individual chapters on the major types of food spoilage, the role of moisture and temperature, spoilage yeasts, the Maillard reaction and the factors underlying lipid oxidation. Part II addresses the best ways of measuring the shelf-life of foods, with chapters on modelling food spoilage, measuring and modelling glass transition, detecting spoilage yeasts, measuring lipid oxidation, the design and validation of shelf-life tests and the use of accelerated shelf-life tests. Rapid and on-line instrumentation for food quality assurance (ISBN 978-1-85573-674-0) With its high volumes of production, the food industry has an urgent need for instrumentation that can be used on-line and which produces rapid measurements. As a result, there has been a wealth of research in this field, which is summarised in this important collection. The first part looks at techniques for the rapid detection of contaminants such as pesticides, veterinary drug residues and foreign bodies. The second part considers methods for identifying ingredients, including additives and measuring product quality. Details of these books and a complete list of Woodhead titles can be obtained by: • visiting our web site at www.woodheadpublishing.com • contacting Customer Services (e-mail: [email protected]; fax: +44 (0) 1223 893694; tel.: +44 (0) 1223 891358 ext. 130; address: Woodhead Publishing Limited, Abington Hall, Granta Park, Great Abington, Cambridge CB21 6AH, UK)


Woodhead Publishing Series in Food Science, Technology and Nutrition: Number 191

Sensory analysis for food and beverage quality control A practical guide Edited by David Kilcast

Oxford

Cambridge

New Delhi


Published by Woodhead Publishing Limited, Abington Hall, Granta Park, Great Abington, Cambridge CB21 6AH, UK www.woodheadpublishing.com Woodhead Publishing India Private Limited, G-2, Vardaan House, 7/28 Ansari Road, Daryaganj New Delhi – 110002, India www.woodheadpublishingindia.com Published in North America by CRC Press LLC, 6000 Broken Sound Parkway, NW, Suite 300, Boca Raton, FL 33487, USA First published 2010, Woodhead Publishing Limited and CRC Press LLC © Woodhead Publishing Limited, 2010 The authors have asserted their moral rights. This book contains information obtained from authentic and highly regarded sources. Reprinted material is quoted with permission, and sources are indicated. Reasonable efforts have been made to publish reliable data and information, but the authors and the publishers cannot assume responsibility for the validity of all materials. Neither the authors nor the publishers, nor anyone else associated with this publication, shall be liable for any loss, damage or liability directly or indirectly caused or alleged to be caused by this book. Neither this book nor any part may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, microfilming and recording, or by any information storage or retrieval system, without permission in writing from Woodhead Publishing Limited. The consent of Woodhead Publishing Limited does not extend to copying for general distribution, for promotion, for creating new works, or for resale. Specific permission must be obtained in writing from Woodhead Publishing Limited for such copying. Trademark notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation, without intent to infringe. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library. Library of Congress Cataloging in Publication Data A catalog record for this book is available from the Library of Congress. Woodhead Publishing ISBN 978-1-84569-476-0 (book) Woodhead Publishing ISBN 978-1-84569-951-2 (e-book) CRC Press ISBN 978-1-4398-3142-7 CRC Press order number: N10199 The publishers’ policy is to use permanent paper from mills that operate a sustainable forestry policy, and which has been manufactured from pulp which is processed using acid-free and elemental chlorine-free practices. Furthermore, the publishers ensure that the text paper and cover board used have met acceptable environmental accreditation standards. Typeset by Toppan Best-set Premedia Limited, Hong Kong Printed by TJ International Limited, Padstow, Cornwall, UK


Contents

Contributor contact details......................................................................... xi Woodhead Publishing Series in Food Science, Technology and Nutrition ............................................................................................... xv Preface.......................................................................................................... xxiii

Part I 1

2

Designing a sensory quality control program ..........................

1

Designing a sensory quality control program................................. M. A. Everitt, ME Consultancy Ltd, UK 1.1 Introduction ............................................................................ 1.2 Company culture and commitment to quality ................... 1.3 Establishing a sensory quality control (QC) program ...... 1.4 Key elements of a sensory quality control (QC) program ................................................................................... 1.5 Overview of approaches used to define sensory targets... 1.6 External support and consultancy ....................................... 1.7 References ...............................................................................

3

Selection and management of staff for sensory quality control ..................................................................................... E. De Vos, Tate & Lyle Food and Industrial Ingredients, EMEA, France 2.1 Introduction ............................................................................ 2.2 Personnel required for sensory quality control ................. 2.3 Setting up a quality control (QC) panel .............................


3 6 7 8 12 15 16

17

17 18 20

vi

Contents 2.4 2.5 2.6 2.7 2.8 2.9

3

Maintaining the quality control (QC) panel: performance, motivation and size ........................................ Possible issues ......................................................................... Case study: selection and management of staff for sensory quality control of cereal-based ingredients .......... Future trends .......................................................................... Sources of further information and advice ......................... References ...............................................................................

28 30 31 34 34 35

Proficiency testing of sensory panels ............................................... G. Hyldig, Technical University of Denmark, Denmark 3.1 Introduction ............................................................................ 3.2 Design and implementation of proficiency testing ............ 3.3 Panels ....................................................................................... 3.4 Analysis of data/validation of results .................................. 3.5 Panel performance ................................................................. 3.6 Glossary ................................................................................... 3.7 References and further reading ...........................................

37 38 43 44 45 46 46

Part II Methods for sensory quality control and analysis of results .......................................................................................

49

4

5

Sensory methods for quality control ............................................... L. L. Rogers, Consultant, UK 4.1 Introduction ............................................................................ 4.2 Descriptive specifications (DS) method.............................. 4.3 ‘In/out’ (or pass/fail) method................................................ 4.4 Difference from control (DFC) method ............................. 4.5 ‘A’ not ‘A’ method .................................................................. 4.6 Paired comparison methods (e.g. 2AFC, n-AFC, simple difference test) ........................................................... 4.7 Scaling method (including targeted scaling) ...................... 4.8 Ranking test ............................................................................ 4.9 Triangle test ............................................................................. 4.10 Quality scoring/grading/rating method ............................... 4.11 Magnitude estimation and duo–trio methods .................... 4.12 In-house and do-it-yourself (DIY) methods ...................... 4.13 References ............................................................................... Establishing product sensory specifications .................................... C. J. M. Beeren, Leatherhead Food Research, UK 5.1 Introduction ............................................................................ 5.2 Rationale using sensory specifications ................................ 5.3 Defining sensory specifications.............................................


37

51 51 55 60 62 65 66 67 69 70 70 72 73 74 75 75 78 78

Contents 5.4 5.5 5.6 5.7 5.8 6

7

Reference samples ................................................................. Implementation of sensory specifications ........................... Maintenance and follow-up .................................................. Case study ............................................................................... References ...............................................................................

Combining instrumental and sensory methods in food quality control ..................................................................................... D. Kilcast, Consultant, Food and Beverage Sensory Quality, UK 6.1 Introduction: the perceptual basis of food quality ............ 6.2 The role of instrumental measurement ............................... 6.3 Sensory analysis of quality .................................................... 6.4 Instrumental measurement of quality factors .................... 6.5 Analysis and validation of instrumental measurements ... 6.6 Future trends .......................................................................... 6.7 Sources of further information ............................................. 6.8 References ...............................................................................

vii 83 84 93 94 96

97 97 98 99 101 105 113 115 115

Statistical approaches to sensory quality control ........................... C. Findlay, Compusense Inc., Canada and A. Hasted, QI Statistics, UK 7.1 Introduction ............................................................................ 7.2 Statistics defined ..................................................................... 7.3 Managing risk ......................................................................... 7.4 Knowing your product........................................................... 7.5 Methods of measurement and practical examples ............ 7.6 Practical considerations ......................................................... 7.7 Assessor proficiency and validation..................................... 7.8 Sensory instrumental correlations ....................................... 7.9 Product matching ................................................................... 7.10 Conclusions ............................................................................. 7.11 References and further reading ...........................................

118

Part III Sensory quality control in practice .........................................

141

8

Using sensory techniques for shelf-life assessment ....................... L. L. Rogers, Consultant, UK 8.1 Introduction ............................................................................ 8.2 What is shelf-life? ................................................................... 8.3 Setting or confirming shelf-life? ........................................... 8.4 The case study: Setting up shelf-life confirmation studies for an ambient product ............................................ 8.5 References and further reading ...........................................


118 119 122 122 125 134 137 138 138 139 140

143 143 144 147 148 155

viii 9

10

11

12

Contents Sensory quality control for taint prevention .................................. D. Kilcast, Consultant, Food and Beverage Sensory Quality, UK 9.1 Introduction ............................................................................ 9.2 Chemistry of taint .................................................................. 9.3 Sources of taints ..................................................................... 9.4 Detection and analysis of taints ........................................... 9.5 Sensory testing procedures ................................................... 9.6 Diagnostic taint testing .......................................................... 9.7 Taint prevention ..................................................................... 9.8 The role of sensory quality control (QC) in taint prevention ...................................................................... 9.9 Ethical aspects ........................................................................ 9.10 Case studies ............................................................................. 9.11 Future trends .......................................................................... 9.12 Sources of further information ............................................. 9.13 References and further reading ........................................... Sensory quality definition of food ingredients ............................... A. Van Biesen, C. Petit and E. Vanzeveren, Puratos N.V., Belgium 10.1 Introduction ............................................................................ 10.2 Developing good quality ingredients in a consumer-oriented approach ................................................ 10.3 Case study 1: What’s your texture? ..................................... 10.4 Case study 2: A toast bread for Chinese consumers ......... 10.5 References ............................................................................... Sensory quality control in the chilled and frozen ready meal, soup and sauce sectors ....................................................................... M. Swainson and L. McWatt, University of Lincoln, UK 11.1 Introduction ............................................................................ 11.2 Sensory quality assurance (QA) in the recipe development process.............................................................. 11.3 Sensory quality assurance (QA) in the post-development product scale-up phase.......................... 11.4 Sensory quality assurance (QA) in the production process ................................................................. 11.5 Sensory quality assurance (QA) after product despatch .................................................................... 11.6 Conflicts of interest ................................................................ 11.7 Conclusions ............................................................................. 11.8 Acknowledgements ................................................................ 11.9 Sources of further information ............................................. Sensory quality control in the wine industry .................................. S. A. Langstaff, Applied Sensory, LLC, USA 12.1 Introduction ............................................................................ 12.2 Historical perspective ............................................................


156 156 159 160 164 165 173 175 178 179 181 183 184 184 186 186 186 188 193 201 203 203 204 206 209 232 233 233 234 234 236 236 237

Contents 12.3 12.4 12.5 12.6 12.7 12.8 12.9 12.10 12.11 12.12 12.13 13

14

15

European standards of wine quality .................................... The concept of wine quality ................................................. Attempts to standardize wine quality evaluation .............. Wine and the development of sensory evaluation as a science ................................................................................... Factors affecting wine quality ............................................... Levels of wine quality............................................................ Approaches to determining wine quality............................ Current sensory quality control practices in winemaking ......................................................................... Future of sensory evaluation in the wine industry ............ Sources of further information ............................................. References ...............................................................................

ix 238 239 242 245 246 248 248 249 257 259 260

Sensory quality control of distilled beverages ................................ J. R. Piggott, University of Strathclyde, UK and S. Macleod, John Dewar and Sons Ltd, UK 13.1 Introduction ............................................................................ 13.2 Origins of sensory quality control of spirits ....................... 13.3 Procedures and precautions .................................................. 13.4 Current industry practices..................................................... 13.5 Taints and off-flavours ........................................................... 13.6 Sources of further information ............................................. 13.7 References ...............................................................................

262

Sensory quality control of fresh produce ........................................ E. Costell, I. Carbonell, A. Tárrega and S. Bayarri, Instituto de Agroquímica y Tecnología de Alimentos, CSIC, Spain 14.1 Introduction ............................................................................ 14.2 The role of sensory analysis in quality control of fruit and vegetables ........................................................................ 14.3 A case study: Influence of storage temperature on the sensory quality of apples ....................................................... 14.4 Acknowledgements ................................................................ 14.5 References ...............................................................................

276

Sensory quality management of fish ................................................ E. Martinsdóttir, Matís – Icelandic Food Research, Iceland 15.1 Introduction: quality indices for fish.................................... 15.2 Guidelines for sensory evaluation of fish ........................... 15.3 Sensory evaluation of fish ..................................................... 15.4 Developing a quality index ................................................... 15.5 Using quality indices in storage management and production planning ...............................................................


262 263 264 266 270 273 273

276 277 280 290 290 293 293 295 296 303 305

x

Contents 15.6 15.7 15.8 15.9

16

17

Keeping fish under different storage conditions ................ Future trends .......................................................................... Acknowledgements ................................................................ References ...............................................................................

306 307 310 310

Sensory quality control in foodservice ............................................ P. G. Creed, formerly of Bournemouth University, UK 16.1 Introduction ............................................................................ 16.2 Aspects of sensory analysis in foodservice ......................... 16.3 Formal methods applicable to foodservice ......................... 16.4 Informal methods applicable to foodservice ...................... 16.5 Sensory quality control in foodservice – a case study ...... 16.6 Future trends .......................................................................... 16.7 Sources of further information and advice ......................... 16.8 References ...............................................................................

316

Sensory quality control of consumer goods other than food ....... A. Giboreau, Institut Paul Bocuse, France 17.1 Introduction ............................................................................ 17.2 General recommendations .................................................... 17.3 The control of sensory quality of non-food products: cases ........................................................................ 17.4 Conclusion ............................................................................... 17.5 Future trends .......................................................................... 17.6 Sources of further information ............................................. 17.7 References ...............................................................................

337

Appendix: Going forward – Implementing a sensory quality control program .................................................................... M. A. Everitt, ME Consultancy Ltd, UK A.1 Piloting the program .............................................................. A.2 Refinement and consolidation .............................................. A.3 Quality assurance (QA) ........................................................ A.4 The effectiveness of a sensory quality control (QC) program ......................................................................... A.5 Maintaining the effectiveness of a sensory quality control/quality assurance (QC/QA) program .................... A.6 Continuous improvement...................................................... Index .............................................................................................................


316 317 322 326 329 329 330 331

337 339 342 349 349 350 350

353 353 353 354 354 356 357 358

Contributor contact details

(* = main contact) Editor and Chapters 6 and 9

Chapter 2

D. Kilcast Consultant, Food & Beverage Sensory Quality

E. De Vos Tate & Lyle Food and Industrial Ingredients, EMEA Parc Scientifique de la Haute Borne 2, Avenue de l’Horizon 59650 Villeneuve d’Ascq France

E-mail: [email protected]

Chapter 1 and Appendix M. A. Everitt ME Consultancy Ltd Sunrise Nurseries Gretton Fields Cheltenham GL54 5HJ UK


Chapter 3


G. Hyldig National Food Institute (DTU Food) Technical University of Denmark Søltofts Plads, Building 221 DK-2800 Kgs. Lyngby Denmark E-mail: [email protected]


xii

Contributor contact details

Chapters 4 and 8

Chapter 11

L. L. Rogers Consultant UK

M. Swainson* and L. McWatt Department of Food Manufacture and Process Automation University of Lincoln Minerva House Park Road Holbeach PE12 7PT UK

E-mail: lauren.l.rogers@googlemail. com

Chapter 5 C. J. M. Beeren Sensory & Consumer Science Leatherhead Food Research Randalls Road Leatherhead KT22 7RY UK E-mail: CBeeren@leatherheadfood. com

Chapter 7


Chapter 12 S. A. Langstaff Applied Sensory LLC 5055 Business Center Dr. #108–126 Fairfield CA 94534 USA E-mail: [email protected]

C. Findlay Compusense Inc. 679 Southgate Drive Guelph, ON Canada N1G 4S2

Chapter 13


Chapter 10 A. Van Biesen*, C. Petit and E. Vanzeveren R&D Department Puratos N.V. Zone Maalbeek Industrialaan, 25 1702 Groot-Bijgaarden Belgium E-mail: [email protected]

J. R. Piggott* University of Strathclyde The Strathclyde Institute of Pharmacy and Biomedical Sciences 204 George Street Glasgow G1 1XW UK E-mail: [email protected] S. Macleod John Dewar and Sons Ltd 1700 London Road Glasgow G32 8XR UK E-mail: [email protected]


Contributor contact details Chapter 14

Chapter 16

E. Costell*, I. Carbonell, A. Tárrega and S. Bayarri Instituto de Agroquímica y Tecnología de Alimentos CSIC PO Box 73 46100 Burjassot Valencia Spain

P. G. Creed 2 Haxen Cottages Allowenshay Hinton St George TA17 8TB UK E-mail: [email protected]

Chapter 17 E-mail: [email protected]

Chapter 15 E. Martinsdóttir Matís ohf Vinlandsleid 12 113 Reykjavik Iceland

A. Giboreau Institut Paul Bocuse Château du vivier BP 25 69130 Ecully France E-mail: agnes.giboreau@ institutpaulbocuse.com

E-mail: emilia.martinsdottir@matis. is


xiii

Woodhead Publishing Series in Food Science, Technology and Nutrition

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

Chilled foods: a comprehensive guide Edited by C. Dennis and M. Stringer Yoghurt: science and technology A. Y. Tamime and R. K. Robinson Food processing technology: principles and practice P. J. Fellows Bender’s dictionary of nutrition and food technology Sixth edition D. A. Bender Determination of veterinary residues in food Edited by N. T. Crosby Food contaminants: sources and surveillance Edited by C. Creaser and R. Purchase Nitrates and nitrites in food and water Edited by M. J. Hill Pesticide chemistry and bioscience: the food–environment challenge Edited by G. T. Brooks and T. Roberts Pesticides: developments, impacts and controls Edited by G. A. Best and A. D. Ruthven Dietary fibre: chemical and biological aspects Edited by D. A. T. Southgate, K. W. Waldron, I. T. Johnson and G. R. Fenwick Vitamins and minerals in health and nutrition M. Tolonen Technology of biscuits, crackers and cookies Second edition D. Manley Instrumentation and sensors for the food industry Edited by E. Kress-Rogers Food and cancer prevention: chemical and biological aspects Edited by K. W. Waldron, I. T. Johnson and G. R. Fenwick Food colloids: proteins, lipids and polysaccharides Edited by E. Dickinson and B. Bergenstahl Food emulsions and foams Edited by E. Dickinson Maillard reactions in chemistry, food and health Edited by T. P. Labuza, V. Monnier, J. Baynes and J. O’Brien The Maillard reaction in foods and medicine Edited by J. O’Brien, H. E. Nursten, M. J. Crabbe and J. M. Ames Encapsulation and controlled release Edited by D. R. Karsa and R. A. Stephenson


xvi

Woodhead Publishing Series in Food Science, Technology & Nutrition

20 Flavours and fragrances Edited by A. D. Swift 21 Feta and related cheeses Edited by A. Y. Tamime and R. K. Robinson 22 Biochemistry of milk products Edited by A. T. Andrews and J. R. Varley 23 Physical properties of foods and food processing systems M. J. Lewis 24 Food irradiation: a reference guide V. M. Wilkinson and G. Gould 25 Kent’s technology of cereals: an introduction for students of food science and agriculture Fourth edition N. L. Kent and A. D. Evers 26 Biosensors for food analysis Edited by A. O. Scott 27 Separation processes in the food and biotechnology industries: principles and applications Edited by A. S. Grandison and M. J. Lewis 28 Handbook of indices of food quality and authenticity R. S. Singhal, P. K. Kulkarni and D. V. Rege 29 Principles and practices for the safe processing of foods D. A. Shapton and N. F. Shapton 30 Biscuit, cookie and cracker manufacturing manuals Volume 1: ingredients D. Manley 31 Biscuit, cookie and cracker manufacturing manuals Volume 2: biscuit doughs D. Manley 32 Biscuit, cookie and cracker manufacturing manuals Volume 3: biscuit dough piece forming D. Manley 33 Biscuit, cookie and cracker manufacturing manuals Volume 4: baking and cooling of biscuits D. Manley 34 Biscuit, cookie and cracker manufacturing manuals Volume 5: secondary processing in biscuit manufacturing D. Manley 35 Biscuit, cookie and cracker manufacturing manuals Volume 6: biscuit packaging and storage D. Manley 36 Practical dehydration Second edition M. Greensmith 37 Lawrie’s meat science Sixth edition R. A. Lawrie 38 Yoghurt: science and technology Second edition A. Y. Tamime and R. K. Robinson 39 New ingredients in food processing: biochemistry and agriculture G. Linden and D. Lorient 40 Benders’ dictionary of nutrition and food technology Seventh edition D. A. Bender and A. E. Bender 41 Technology of biscuits, crackers and cookies Third edition D. Manley 42 Food processing technology: principles and practice Second edition P. J. Fellows 43 Managing frozen foods Edited by C. J. Kennedy 44 Handbook of hydrocolloids Edited by G. O. Phillips and P. A. Williams 45 Food labelling Edited by J. R. Blanchfield 46 Cereal biotechnology Edited by P. C. Morris and J. H. Bryce 47 Food intolerance and the food industry Edited by T. Dean 48 The stability and shelf life of food Edited by D. Kilcast and P. Subramaniam 49 Functional foods: concept to product Edited by G. R. Gibson and C. M. Williams 50 Chilled foods: a comprehensive guide Second edition Edited by M. Stringer and C. Dennis 51 HACCP in the meat industry Edited by M. Brown


Woodhead Publishing Series in Food Science, Technology & Nutrition 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87

xvii

Biscuit, cracker and cookie recipes for the food industry D. Manley Cereals processing technology Edited by G. Owens Baking problems solved S. P. Cauvain and L. S. Young Thermal technologies in food processing Edited by P. Richardson Frying: improving quality Edited by J. B. Rossell Food chemical safety Volume 1: contaminants Edited by D. Watson Making the most of HACCP: learning from others’ experience Edited by T. Mayes and S. Mortimore Food process modelling Edited by L. M. M. Tijskens, M. L. A. T. M. Hertog and B. M. Nicolaï EU food law: a practical guide Edited by K. Goodburn Extrusion cooking: technologies and applications Edited by R. Guy Auditing in the food industry: from safety and quality to environmental and other audits Edited by M. Dillon and C. Griffith Handbook of herbs and spices Volume 1 Edited by K. V. Peter Food product development: maximising success M. Earle, R. Earle and A. Anderson Instrumentation and sensors for the food industry Second edition Edited by E. Kress-Rogers and C. J. B. Brimelow Food chemical safety Volume 2: additives Edited by D. Watson Fruit and vegetable biotechnology Edited by V. Valpuesta Foodborne pathogens: hazards, risk analysis and control Edited by C. de W. Blackburn and P. J. McClure Meat refrigeration S. J. James and C. James Lockhart and Wiseman’s crop husbandry Eighth edition H. J. S. Finch, A. M. Samuel and G. P. F. Lane Safety and quality issues in fish processing Edited by H. A. Bremner Minimal processing technologies in the food industries Edited by T. Ohlsson and N. Bengtsson Fruit and vegetable processing: improving quality Edited by W. Jongen The nutrition handbook for food processors Edited by C. J. K. Henry and C. Chapman Colour in food: improving quality Edited by D MacDougall Meat processing: improving quality Edited by J. P. Kerry, J. F. Kerry and D. A. Ledward Microbiological risk assessment in food processing Edited by M. Brown and M. Stringer Performance functional foods Edited by D. Watson Functional dairy products Volume 1 Edited by T. Mattila-Sandholm and M. Saarela Taints and off-flavours in foods Edited by B. Baigrie Yeasts in food Edited by T. Boekhout and V. Robert Phytochemical functional foods Edited by I. T. Johnson and G. Williamson Novel food packaging techniques Edited by R. Ahvenainen Detecting pathogens in food Edited by T. A. McMeekin Natural antimicrobials for the minimal processing of foods Edited by S. Roller Texture in food Volume 1: semi-solid foods Edited by B. M. McKenna Dairy processing: improving quality Edited by G Smit


xviii 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109

110 111 112 113 114 115 116 117 118 119 120 121 122 123

Woodhead Publishing Series in Food Science, Technology & Nutrition Hygiene in food processing: principles and practice Edited by H. L. M. Lelieveld, M. A. Mostert, B. White and J. Holah Rapid and on-line instrumentation for food quality assurance Edited by I. Tothill Sausage manufacture: principles and practice E. Essien Environmentally-friendly food processing Edited by B. Mattsson and U. Sonesson Bread making: improving quality Edited by S. P. Cauvain Food preservation techniques Edited by P. Zeuthen and L. Bøgh-Sørensen Food authenticity and traceability Edited by M. Lees Analytical methods for food additives R. Wood, L. Foster, A. Damant and P. Key Handbook of herbs and spices Volume 2 Edited by K. V. Peter Texture in food Volume 2: solid foods Edited by D. Kilcast Proteins in food processing Edited by R. Yada Detecting foreign bodies in food Edited by M. Edwards Understanding and measuring the shelf-life of food Edited by R. Steele Poultry meat processing and quality Edited by G. Mead Functional foods, ageing and degenerative disease Edited by C. Remacle and B. Reusens Mycotoxins in food: detection and control Edited by N. Magan and M. Olsen Improving the thermal processing of foods Edited by P. Richardson Pesticide, veterinary and other residues in food Edited by D. Watson Starch in food: structure, functions and applications Edited by A.-C. Eliasson Functional foods, cardiovascular disease and diabetes Edited by A. Arnoldi Brewing: science and practice D. E. Briggs, P. A. Brookes, R. Stevens and C. A. Boulton Using cereal science and technology for the benefit of consumers: proceedings of the 12th International ICC Cereal and Bread Congress, 24–26th May, 2004, Harrogate, UK Edited by S. P. Cauvain, L. S. Young and S. Salmon Improving the safety of fresh meat Edited by J. Sofos Understanding pathogen behaviour in food: virulence, stress response and resistance Edited by M. Griffiths The microwave processing of foods Edited by H. Schubert and M. Regier Food safety control in the poultry industry Edited by G. Mead Improving the safety of fresh fruit and vegetables Edited by W. Jongen Food, diet and obesity Edited by D. Mela Handbook of hygiene control in the food industry Edited by H. L. M. Lelieveld, M. A. Mostert and J. Holah Detecting allergens in food Edited by S. Koppelman and S. Hefle Improving the fat content of foods Edited by C. Williams and J. Buttriss Improving traceability in food processing and distribution Edited by I. Smith and A. Furness Flavour in food Edited by A. Voilley and P. Etievant The Chorleywood bread process S. P. Cauvain and L. S. Young Food spoilage microorganisms Edited by C. de W. Blackburn Emerging foodborne pathogens Edited by Y. Motarjemi and M. Adams


Woodhead Publishing Series in Food Science, Technology & Nutrition 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156

xix

Benders’ dictionary of nutrition and food technology Eighth edition D. A. Bender Optimising sweet taste in foods Edited by W. J. Spillane Brewing: new technologies Edited by C. Bamforth Handbook of herbs and spices Volume 3 Edited by K. V. Peter Lawrie’s meat science Seventh edition R. A. Lawrie in collaboration with D. A. Ledward Modifying lipids for use in food Edited by F. Gunstone Meat products handbook: practical science and technology G. Feiner Food consumption and disease risk: consumer–pathogen interactions Edited by M. Potter Acrylamide and other hazardous compounds in heat-treated foods Edited by K. Skog and J. Alexander Managing allergens in food Edited by C. Mills, H. Wichers and K. Hoffman-Sommergruber Microbiological analysis of red meat, poultry and eggs Edited by G. Mead Maximising the value of marine by-products Edited by F. Shahidi Chemical migration and food contact materials Edited by K. Barnes, R. Sinclair and D. Watson Understanding consumers of food products Edited by L. Frewer and H. van Trijp Reducing salt in foods: practical strategies Edited by D. Kilcast and F. Angus Modelling microorganisms in food Edited by S. Brul, S. Van Gerwen and M. Zwietering Tamime and Robinson’s Yoghurt: science and technology Third edition A. Y. Tamime and R. K. Robinson Handbook of waste management and co-product recovery in food processing: Volume 1 Edited by K. W. Waldron Improving the flavour of cheese Edited by B. Weimer Novel food ingredients for weight control Edited by C. J. K. Henry Consumer-led food product development Edited by H. MacFie Functional dairy products Volume 2 Edited by M. Saarela Modifying flavour in food Edited by A. J. Taylor and J. Hort Cheese problems solved Edited by P. L. H. McSweeney Handbook of organic food safety and quality Edited by J. Cooper, C. Leifert and U. Niggli Understanding and controlling the microstructure of complex foods Edited by D. J. McClements Novel enzyme technology for food applications Edited by R. Rastall Food preservation by pulsed electric fields: from research to application Edited by H. L. M. Lelieveld and S. W. H. de Haan Technology of functional cereal products Edited by B. R. Hamaker Case studies in food product development Edited by M. Earle and R. Earle Delivery and controlled release of bioactives in foods and nutraceuticals Edited by N. Garti Fruit and vegetable flavour: recent advances and future prospects Edited by B. Brückner and S. G. Wyllie Food fortification and supplementation: technological, safety and regulatory aspects Edited by P. Berry Ottaway


xx

Woodhead Publishing Series in Food Science, Technology & Nutrition

157

Improving the health-promoting properties of fruit and vegetable products Edited by F. A. Tomás-Barberán and M. I. Gil Improving seafood products for the consumer Edited by T. Børresen In-pack processed foods: improving quality Edited by P. Richardson Handbook of water and energy management in food processing Edited by J. Klemeš, R. Smith and J-K Kim Environmentally compatible food packaging Edited by E. Chiellini Improving farmed fish quality and safety Edited by Ø. Lie Carbohydrate-active enzymes Edited by K-H Park Chilled foods: a comprehensive guide Third edition Edited by M. Brown Food for the ageing population Edited by M. M. Raats, C. P. G. M. de Groot and W. A. Van Staveren Improving the sensory and nutritional quality of fresh meat Edited by J. P. Kerry and D. A. Ledward Shellfish safety and quality Edited by S. E. Shumway and G. E. Rodrick Functional and speciality beverage technology Edited by P. Paquin Functional foods: principles and technology M. Guo Endocrine-disrupting chemicals in food Edited by I. Shaw Meals in science and practice: interdisciplinary research and business applications Edited by H. L. Meiselman Food constituents and oral health: current status and future prospects Edited by M. Wilson Handbook of hydrocolloids Second edition Edited by G. O. Phillips and P. A. Williams Food processing technology: principles and practice Third edition P. J. Fellows Science and technology of enrobed and filled chocolate, confectionery and bakery products Edited by G. Talbot Foodborne pathogens: hazards, risk analysis and control Second edition Edited by C. de W. Blackburn and P. J. McClure Designing functional foods: measuring and controlling food structure breakdown and absorption Edited by D. J. McClements and E. A. Decker New technologies in aquaculture: improving production efficiency, quality and environmental management Edited by G. Burnell and G. Allan More baking problems solved S. P. Cauvain and L. S. Young Soft drink and fruit juice problems solved P. Ashurst and R. Hargitt Biofilms in the food and beverage industries Edited by P. M. Fratamico, B. A. Annous and N. W. Gunther Dairy-derived ingredients: food and neutraceutical uses Edited by M. Corredig Handbook of waste management and co-product recovery in food processing Volume 2 Edited by K. W. Waldron Innovations in food labelling Edited by J. Albert Delivering performance in food supply chains Edited by C. Mena and G. Stevens Chemical deterioration and physical instability of food and beverages Edited by L. Skibsted, J. Risbo and M. Andersen Managing wine quality Volume 1: viticulture and wine quality Edited by A. Reynolds

158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187


Woodhead Publishing Series in Food Science, Technology & Nutrition 188

xxi

Improving the safety and quality of milk Volume 1: milk production and processing Edited by M. Griffiths 189 Improving the safety and quality of milk Volume 2: improving quality in milk products Edited by M. Griffiths 190 Cereal grains: assessing and managing quality Edited by C. Wrigley and I. Batey 191 Sensory analysis for food and beverage quality control: a practical guide Edited by D. Kilcast 192 Managing wine quality Volume 2: oenology and wine quality Edited by A. Reynolds 193 Winemaking problems solved Edited by C. Butzke 194 Environmental assessment and management in the food industry Edited by U. Sonesson, J. Berlin and F. Ziegler 195 Consumer-driven innovation in food and personal care products Edited by S. Jaeger and H. MacFie 196 Tracing pathogens in the food chain Edited by S. Brul, P.M. Fratamico and T.A. McMeekin 197 Case studies in novel food processing technologies Edited by C. Doona, K. Kustin and F. Feeherry 198 Freeze-drying of pharmaceutical and food products Tse-Chao Hua, Bao-Lin Liu and Hua Zhang 199 Oxidation in foods and beverages and antioxidant applications: Volume 1 understanding mechanisms of oxidation and antioxidant activity Eric A. Decker, Ryan J. Elias and D. Julian McClements 200 Oxidation in foods and beverages and antioxidant applications: Volume 2 management in different industry sectors Eric A. Decker, Ryan J. Elias and D. Julian McClements 201 Protective cultures, antimicrobial metabolites and bacteriophages of food and beverage biopreservation Edited by Christophe Lacroix


Preface

The last 20–30 years have seen many significant developments in the use of sensory analysis in both the food and beverages industries, and also in many non-food industries. This growth has been driven by many varied factors, among which can be listed: the growing (and in some sectors, belated) recognition that quality perceptions can be measured by humans; acceptance by regulators that quality standards must incorporate a strong element of human perception; wider ranges of test procedures, together with a better understanding of their advantages and limitations; development of both statistical and non-statistical methods for data analysis; and an explosion of interest in perceptual mechanisms at physicochemical, psychological and neurological levels. As a consequence, researchers with any interest in perceived sensory quality can draw on an enormous range of techniques. This has given the companies in the manufacturing sector the power to add more relevant focus to development programs and higher probabilities of meeting consumer requirements. Paradoxically, however, the increased opportunities available in the use of sensory techniques has been paralleled by moves throughout industry to increase production automation and to reduce employee numbers correspondingly. This has left many companies with real difficulties in implementing valid sensory testing systems in circumstances when staff numbers and time availability are at a premium. This situation can generally be managed within the larger companies that are able to support the required sensory facilities and staff skills, and medium/small companies with a need for sensory evaluation as part of their R&D functions can frequently call on contract laboratories for their needs. Specific problems can often arise when companies incorporate sensory evaluation into their quality control (QC) operations, however. Suppliers


xxiv

Preface

both of ingredients and finished products will commonly find that, whilst their needs for sensory evaluation in R&D can be contracted out, it is much more difficult to outsource sensory QC testing, for which there is a need for regular and rapid examination of product quality, together with consequential action plans, and which can in most cases only be carried out on-site. Demands for routine sensory QC come from many directions, in particular from national regulatory bodies, through accreditation procedures and from customer demands higher up the production chain. Consequently, all companies face the need to have in place some form of in-house sensory QC testing, even when sensory panels for R&D are not seen as essential. For such companies, which can range up through all sizes, the main difficulty to be faced lies in translating the vast amount of information in the scientific literature on setting up, operating and interpreting sensory tests into practical sensory QC systems that can be managed effectively within their own business, and delivering their objectives. One unfortunate consequence of the barriers facing companies in implementing reliable sensory QC testing is the belief that instrumental measurements can be used to replace sensory panels. Whilst there are many circumstances in which instrumental measurements of quality parameters are invaluable in complementing sensory information, a position must be found for appropriate sensory procedures. Small businesses that are unable to operate sensory protocols that conform to best practice must have as a minimum some form of sensory testing in place, coupled with an understanding of the limitations inherent in such non-ideal procedures. One important difference between the use of sensory evaluation in R&D and in QC environments is that whereas a high level of commonality of testing is possible across R&D functions in different industry sectors, sensory procedures in QC functions are adapted to a large degree to fit the requirements and operating constraints of individual companies. This results in highly disparate sensory procedures becoming established across the industry, in spite of the increasing use of standardisation. Valuable lessons can therefore be learned from the experiences of sensory specialists in a wide range of product sectors, and this book aims to record such experiences. The experiences within the food and beverage sectors have in recent years been exported to a wide range of non-food industries, and such nonfood applications are included in this volume. No reliable sensory QC program can be successful without adequate planning, and the first section of this book deals with the key aspects in designing a sensory QC program, and including the selection and management of staff for sensory QC, and important recent developments in the proficiency testing of sensory panels. The second part of the book covers sensory methods and data analysis, and together with an overview of the available methods includes how the setting up of sensory specifications can be approached, and how instrumen-


Preface

xxv

tal measurements can be used to complement sensory data. Issues associated with the statistical analysis of sensory data are covered in Chapter 7. In the third part of the book, the practical uses of sensory QC in a wide range of food and beverage applications are described, including specific applications to shelf-life assessment and to taint prevention. Finally, recent developments in the uses of sensory QC in non-food industries are described. Most companies have their own needs and constraints when setting up and operating sensory QC panels, and the aim of this book is to provide any company with some direction in establishing the basis for a reliable sensory QC program that meets the demands of a competitive business environment, and that will be able to adapt to changing future demands. David Kilcast


1 Designing a sensory quality control program M. A. Everitt, ME Consultancy Ltd, UK

Abstract: This chapter provides an overview of some of the main areas that need due consideration when planning to establish a sensory quality control program. Particular attention is given to the importance of management commitment, sensory specifications and common failings, approaches used to set sensory targets, and training regime. Key words: sensory specifications, consumer focus, sensory training program, quality grading.

1.1 Introduction Gaining and maintaining a quality advantage have become primary competitive issues within the food and drink industry. As the range within sectors plus the overall standard of food available to consumers continues to increase so has the need for businesses to review their policies towards quality versus production volume. Defining product quality and setting the parameters by which it is measured and controlled has been and remains a major challenge with regard to the sensory element of the product specification. The approach taken by a company in the way it defines and subsequently measures quality can have a major influence on the degree to which a sensory program is incorporated into the overall quality monitoring system. Gavin (1984) discusses five classical approaches that can be used to define product quality: 1. Transcendent approach based on philosophy which states that quality is recognised only through experience and cannot be precisely defined. 2. Product-based approach founded on economics which views quality as a precise and measurable property relating to variation in the amount of a specified characteristic or ingredient.


4


3. Manufacturing-based operations management approach which judges quality primarily by the level of conformance with a defined specification; any deviation is seen as a reduction in quality. 4. Value-based operations management approach which uses cost and price to define quality; a quality product would provide performance at an acceptable price or conformance at an acceptable cost. 5. User-based approach combines economics, marketing and operations management principles with the focus being on consumer satisfaction; a quality product would achieve greatest satisfaction for a specified group of consumers. A sensory quality control (QC) program will be most readily accepted and effectively incorporated into businesses that take the last approach. This approach also helps companies maintain a realistic balance between product quality, production volume and cost. A consumer-focused route is therefore advocated as most effective for setting up the sensory component of a QC program although other approaches based more on the use of internal business expertise are discussed.

1.1.1 Principle and objective of a sensory QC program The main aims of QC are to ensure: • • • • •

legal requirements are met; the product is safe and fit for use; there is compliance with nutritional guidelines and tolerances; the agreed/declared weight is provided; deviation from expected quality is kept within an acceptable tolerance.

The last area tends to be the most difficult to define and measure within a food or beverage program as the criteria typically used to judge ‘expected quality’ will primarily relate to a product’s sensory aspects (i.e. appearance, flavour and, texture) which are more prone to subjective interpretation due to personal perceptions and preferences. This can lead to the sensory part of a product quality specification being inadequately or in some instances inappropriately defined owing to insufficient appreciation of the sensory features that have most affect on the target consumers’ perception of quality.

1.1.2 Common failings in defining the sensory specification Visual characteristics Because visual characteristics are the most ‘tangible’ of all the sensory attributes these tend to receive the most attention with physical features (e.g. depth of a sponge layer, number of nuts on top of a Dundee cake)


Designing a sensory quality control program

5

often taking precedence over the true sensory attributes such as depth and uniformity of colour, degree of clarity and brightness. Flavour, texture and mouth-feel characteristics Descriptions relating to these modalities are often written in vague, generalised terms (e.g. ‘typical flavour’, ‘freshness of flavour’, ‘standard texture’) which can lead to varied assumptions being made across assessors about the interpretation. Even when individual attributes are listed, issues can still arise, especially if the author of the specification is not familiar with sensory evaluation principles and the need for precise, unambiguous terminology. Terms can remain bundled and confusing (e.g. confectionery flavour, ripe flavour) or be too technical for general application (e.g. diacetyl instead of ‘buttery’, dimethyl sulphide instead of ‘cabbagey’, 2,4,6-trichloroanisole instead of ‘musty’). Relevance and priority of sensory attributes To aid initial focus when defining the specification, the sensory quality attributes can be considered as falling into two main groupings: those which relate to the key characteristics that affect consumers’ acceptance of a product and those which are of internal importance to a business in order for it to maintain cost-effective control of the production process and avoid unnecessary wastage of materials. The relevance of the attributes under the ‘internal’ listing does not ordinarily pose an issue as the association within a business of a defined term with a specific processing factor or raw material is usually readily understood and its need accepted. Visual defects and also off-flavour characteristics are typically most relevant in the ‘internal’ attribute list. Particular visual defects can, for example, indicate the need to reset peeling, cutting or slicing equipment which if left unchecked would lead to increased waste of raw material. The monitoring of specific off-flavours can aid suitable control of factors such as processing temperatures and flow rates, water conservation, fats and oil quality, etc. Failings are more common regarding specification of the characteristics that relate to consumer acceptance. Attributes that are of no particular importance are often specified whilst those that are key to consumers’ degree of liking can be given low priority or completely overlooked. To ensure that the most appropriate attributes are specified and avoid the risk of subjective influences, the key features that relate to a target market’s expectations of the sensory quality need to be understood by those responsible for setting and applying the specification. The most reliable way to achieve this is for the information gained from consumer and sensory research conducted during a products development or re-development to be communicated to the Quality and Production teams before routine production commences. Unfortunately a common failing is that this information is not transferred, even after the running of detailed and lengthy


6


product development programs. This increases the risk of the sensory quality of routinely produced product deviating from its quality and hence eroding any quality advantage originally established at launch. Attribute weighting In addition to defining an appropriate set of sensory characteristics priorities can also be given to specific attributes to aid the most effective and efficient use of the specification. Again if a consumer-focused approach has been used to determine the main sensory criteria this will also enable weightings of importance to be designated more objectively for each and identify which are the most critical to control.

1.2

Company culture and commitment to quality

Without question companies that have established the most effective and successful sensory QC programs are those that have full management support from senior levels down, and view the control of sensory quality as an essential part of the overall QC system. These companies are commonly consumer-focused with a commitment to understanding customer/ consumer needs and achieving their total satisfaction, along with an aim for continuous improvement. The internal culture is usually one that encourages team working and ownership of quality across the workforce; it is not seen just as a function of the QC manager and operatives. Muñoz (2002) lists some of the main reasons that limit or prevent the development of a sound sensory QC program; the key issue running through relates to lack of support, interest and respect within an organisation, in particular from Plant, Quality and R&D management. 1.2.1 Winning management support As Lawless and Heymann (1998) state, ‘Without management support, especially from manufacturing, a QC program is bound to fail. In a typical case the program will amount to nothing more than “rubber stamping” of supervisory opinion thus supporting a management policy that maximises productivity at the expense of producing unacceptable products’. Production volume will remain a critical key performance indicator (KPI) for most if not all food manufacturing plants. However, quality systems that include standardised sensory assessment procedures are achieving greater acceptance as businesses increasingly strive to have a quality edge over their competitors. The opportunity therefore to promote the need for robust sensory QC programs has possibly never been greater. Selling the benefits It is important to show that developing a sensory component within a QC program will enhance it rather than detract. Management may be con-



7

cerned that in addition to the initial cost involved in setting up a sensory program, it will also lead to constant interruption of the production process with a subsequent drop in volume. The following list provides some examples that can be used to help promote the benefits of having sound sensory measures and how they can aid quality control: • reduce sensory quality variability both within and between production shifts by ensuring operatives have a common understanding of the key attributes of the target product; • avoid the development of serious quality issues, for example, by early stage off-flavour and taint detection; • resolve problems quickly by providing precise descriptions about the sensory quality enabling a specific stage, setting or ingredient in the production process to be targeted; • increase production efficiency by showing that the time allocated to a stage in the process, e.g. pre-mixes, blending, maturation, can be shortened without being detrimental to the finished product quality; • aid risk versus benefit analysis for operations management decisions. Management support needs to be sustained and commitment needs to be actively demonstrated. Companies that maintain successful sensory QC programs not only have a firm belief in product quality but provide financial support through investment in staff and their training plus incentive and reward schemes.

1.3 Establishing a sensory quality control (QC) program Once commitment is gained for setting up a sensory QC program it can be followed by over-ambitious implementation strategies due primarily to lack of comprehension of the time required to address adequately the various stages that are needed, in particular the training of assessors and validation of their abilities. If insufficient time is given to implementation it can be to the detriment of the final quality of the system and subsequently the respect it receives within a business. The author recommends the most effective approach, and also most efficient in the longer term, is to select one product line together with one production site to initially pilot the system and get it fully established. Once the overall approach is agreed and all associated procedures are defined it becomes quicker and easier to roll the system out to other plants and for other products. It can also be beneficial initially to apply the system along with the development of a new product to: • help encourage the collaboration required between Product Development and Quality Assurance to ensure a sensory specification is consumer-focused;


8


• overcome personal assumptions about the ‘ideal’ sensory quality which can be ingrained within a business especially for long-established product lines.

1.4 Key elements of a sensory quality control (QC) program Listed below are the main areas to address in preparation for implementation once management support has been gained and the product line agreed for initial roll-out: • • • • • • • • •

defined consumer-based sensory specification; program manager/coordinator; products for training purposes; training program; personnel to be trained; sensory assessment method; assessment area; data capture and dissemination; procedures and protocols.

1.4.1 Sensory specification As discussed earlier the specification needs to include the attributes that are important to consumers’ acceptance of the product plus those that provide information of internal importance. The specifications need to be precise and concise and cover only the key attributes so that time is not wasted assessing unnecessary characteristics either during panellist training or subsequent routine testing; typically the total number of attributes will be from five to nine. The attributes with their associated sensory targets need to be defined to enable appropriate training samples to be produced. Setting sensory targets is discussed further in Section 1.5. 1.4.2 Sensory program coordinator It is vital to have a coordinator of the system in order to maintain consistency in its application across the business. In most instances this individual will have a technical or quality control background rather than being a trained sensory professional. This should not pose an issue as long as they receive suitable sensory training prior to implementation and are given adequate authority to set up and subsequently audit the functioning of the program. The coordinator’s main responsibilities will be to: • organise and implement training programs; • work with Product Development and QC to define the sensory specifications;



9

• ensure adequate supplies of training samples are maintained; • ensure all relevant documentation is maintained and updated as appropriate; • ensure the system continues to be applied in a sound and professional manner; • provide reports and reviews for management of results plus the overall performance of the system.

1.4.3 Training program This is a critical element in achieving the successful operation of the program. Without sensory training, assessors’ judgements will be based on their own preferences and therefore be unreliable and variable; and may be skewed away from the sensory requirements of consumers in the target market. Using a reference product to illustrate the target quality, and also varying degrees of deviation around it, builds objectivity and consistency of assessment. As part of the training program, it can be beneficial to start with a basic overview of sensory evaluation principles and measurement largely via practical illustrations, to help panellists better appreciate the need for the training. A short exercise at the end is also advisable to validate each panellist’s competency and highlight their weakest areas. For instance four to five products that illustrate target, just acceptable and sub-standard product could be presented. Depending on the level of ability desired a ‘pass’ level would be defined. The amount of training time required will depend largely on the detail of the specification, whether a general, integrated ‘go/no go’ or a more diagnostic evaluation is wanted and the competency level desired for panellists. Pecore and Kellen (2002) discuss a sensory evaluation program established in General Mills where ‘each location customizes their training to meet the product needs, but in general, all panellists receive a basic overview of sensory evaluation, train on all available reference samples, then practice for 3 months’. As a general guide, 8 hours training minimum (the length of individual sessions can be tailored to suit) is advised for the initial introduction and reference sample orientation, and a follow-on practice period over at least a month to consolidate learning is highly recommended. To use assessors’ data without allowing them a consolidation phase in the formal test context can be risky and jeopardise the quality of the resulting information and hence confidence in the system.

1.4.4 Training samples The purpose of the training samples is to illustrate the range of variation in the intensities of the target product and the extent to which they can vary before the deviation becomes unacceptable, i.e. show the quality ‘grades’


10


that are to be designated. Off-flavours that could occur plus the range of possible defects also need to be demonstrated to develop panellists’ ability to detect and accurately describe and quantify them. Ideally the samples should be produced based on the information gained during product development to ensure consumer focus is retained. (Setting sensory targets is discussed further in Section 1.5.) A plentiful supply of the reference material, in particular the target standard, needs to be stored to aid re-calibration of panellists’ perceptions as well as for future training. If sufficient target product can be stored, it is advisable to have it available at the start of all evaluations to ensure assessor’s perceptions are freshly calibrated rather than letting them rely on memory. Optimal storage conditions should be applied to ensure minimum change over time. Frozen and ambient stable products (bottled, canned, dry packed, etc.) can usually be confidently held for 6 to 9 months in this way. Short shelf-life products pose more of a challenge and a written reference based on the descriptive profile plus the technical production specification set at launch may be the only option.

1.4.5 Selecting panellists Panellists involved in QC are typically recruited from the staff available at the manufacturing site/s. Recruitment is advised across all sections and shifts to encourage ownership of quality, commitment to the program and active problem-solving throughout the workforce. Guidance from manufacturing management, i.e. the factory manager, technical/quality manager and production director can prove valuable here by identifying the shift personnel who currently show most initiative towards quality control and problem solving. These people are most likely to become good ambassadors of the system and it is therefore beneficial to include them in the first round of training. By building an extended pool of trained panellists the assessments can be shared out, so helping reduce the risk of sensory adaptation and fatigue, and ensuring that an adequate number of panellists are available for each assessment occasion. Panellist screening and training are detailed further in Chapter 2.

1.4.6 Sensory assessment method Of all the aspects relating to sensory QC programs, comparison of the use of sensory methods is possibly the best documented. Muñoz et al. (1992) and Costell (2002) are but two publications that cover the topic in detail. Most popular methods tend be founded on a form of simplified descriptive analysis or a quality grading method. Various forms of grading (quality rating) methods have become very popular for online sensory control as they provide a quick, reliable (providing panellists have been fully trained



11

against reference products) and practical approach to measuring deviations from the target quality. One of the most popular of all is the red, amber, green ‘RAG’ system operated globally by a number of multinational companies. The method comprises three grades into which finished product can be allocated: green = target quality, amber = borderline quality, red = unacceptable quality. The method is often applied in two ways: an integrated approach where the sensory characteristics are judged and graded in entirety and a diagnostic approach where the key attributes for each sensory modality are graded individually, with the final grade allocated based on the lowest performing feature. The former approach provides a rapid, basic check of overall sensory quality while the latter enables focused troubleshooting and detailed tracking of quality fluctuations over time.

1.4.7 Assessment area Some significant compromises are likely to have to be made in relation to the assessment area as owing to restrictions on panellists’ time away from their primary duties, it usually needs to be within close proximity of the production line; the existing QC laboratory being an obvious choice. Air and light quality, noise levels, distraction and limited space can be particular issues. Facilities need to be standardised and controlled as far as possible in line with recommended sensory practice. ISO 8589 (2007) provides general guidance for the design of test rooms. Important assessments should always be conducted in a defined area.

1.4.8 Data capture and dissemination An important consideration in the capture of the data is how quickly any resulting actions can be communicated back to the production line and associated management. Individual assessments will most commonly be recorded on a paper ballot and collated by a QC operative and typically only involve a few panellists, three to five for example. Prompt communication of the resulting actions should not pose a problem. The longer-term benefit of a sensory QC program which also helps promote its perceived value is gained through trend analysis. This allows the sensory features causing the most frequent deviations from target quality to be identified which in turn can be linked back to specific aspects of the production process. To facilitate the analysis and communication of this information, data from the individual assessments are best stored electronically. Most manufacturing sites will already have computerised systems to store data from other quality and technical measurements. The software can often be adapted to accommodate the sensory information as well; if not, it is advisable to plan to have a program developed. Excel software may be more than adequate. Findlay (2002) describes a computerised system for sensory QC that operates both at local and Internet level and that can also be inte-


12


grated with other data systems such as a laboratory information management system (LIMS). 1.4.9 Procedures and protocols All associated procedures, protocols, assessment forms, etc., need to be standardised, with current versions maintained across the system in order to avoid bias and error from varied practices or out-of-date instructions being used. Key areas to address are: • sample collection, sampling point, e.g. directly off-line, after packing, after initial storage; • sample size; • preparation, cooking, serving procedures; • evaluation method; • assessment protocol: minimum portion, palate cleansing; • collating, recording, reporting results; • response procedures, hierarchy of actions dependent on result.

1.5 Overview of approaches used to define sensory targets Two approaches are advised which provide a sound route to achieving meaningful sensory specifications; a consumer-focused one, which advocates maximum use of consumer and sensory research information obtained during the product development process, and a producer-focused one, which utilises product knowledge from within a business. The former is advised as the most reliable and effective but it is not always possible due to time and budget constraints. 1.5.1 Consumer-targeted approach Figure 1.1 shows the key stages in the process of defining consumer-targeted sensory specifications. The development and refinement of consumer preference segmentation and preference mapping data analysis methods over the past two decades has enabled consumers’ widely varying individual preferences to be classified and measured in a more objective, reliable and consistent manner. As a descriptive sensory profiling technique typically supports the research, this approach provides the most comprehensive information about the key drivers of consumer acceptance along with the importance of each. The need for one or more target products to satisfy a market can be identified and information gathered to allow sensory targets to be set for each. By comparing the sensory profile of the most liked product (for the market in total or segmented groups) with ones for products that record a decrease in the degree of consumer liking, the acceptable range of deviation can be identified. This information provides the blueprint



Consumer & sensory research information

Key preference drivers identified – positive and negative

Target product attributes defined with tolerance limits

13

Sensory specification established

Fig. 1.1 Flow diagram illustrating the key stages in the process of defining consumer-targeted sensory specifications. Background image Copyright © 2007 Microsoft Corporation.

for defining the sensory grades and tolerance limits that will be set and also the reference samples to be used for training. The acceptance range can be set as tight or as loose as a business deems fit; either way with this approach it can be from a very informed perspective and appreciation of the business risks involved. Everitt (2009) details how to make effective use of preference mapping and cluster analysis data to define realistic sensory targets and establish practical consumer targeted sensory specifications. A sensory profile plot has the sensory panel mean intensity scores plotted on each attribute axis for each product. Intensity usually goes from low at the centre to the highest scores at the periphery as shown in Fig. 1.2. If the profile lines overlap or are in close proximity on a given attribute axis the difference between the scores is not statistically significant. Where the lines distinctly diverge, the differences are likely to be statistically significant (but always refer back to the associated ANOVA table from the original analysis); the larger the distance between the mean scores the greater the difference in the magnitude of intensity. Collaboration between departments Communication between Product Development, Manufacturing and Quality Control needs to be developed as an integral part of the development process to ensure that the sensory QC coordinator and supporting staff develop their knowledge of the desired sensory properties of the product as it develops in order to have a suitable specification ready for use from a product’s launch. This communication link is well established in some companies, particularly those where the sensory QC program is overseen and guided at a corporate level. For companies just starting or in the


14

Sensory analysis for food and beverage quality control Greasiness 100 Smoothness

80

Depth of colour

60

Stickiness

Uniformity of colour

40 20 0

Firmness

Yellow colour

Caramelised

Baked odour Vanilla

Burnt Sweetness

729 Target 194 257 438

Fig. 1.2 Comparison of sensory profiles generated by a trained sensory panel, showing three products selected to represent the sensory quality range, i.e. products 729, 257, 438. Product 194 illustrates an unacceptable deviation for key attributes depth of colour, baked odour, caramelised and burnt.

early stages of implementing a system it would be advisable to aim to establish this link as a priority. Level of consumer information Product development projects do not all have the budget or even warrant the need for consumer research on the scale of preference mapping. However, quantitative consumer information, even that from the simplest of product guidance studies, can be useful in establishing the sensory criteria for the QC specification. Again it is advisable to develop a culture whereby the departments that commonly obtain this type of data, i.e. Research and Product Development and Marketing, automatically feed information through to Manufacturing and, in particular, Quality Systems.

1.5.2 Producer-focused approach Some businesses primarily use an internally focused approach to set the target parameters of the sensory quality of their products. A crossfunctional team will be convened from Product Development, Marketing, Manufacturing and Quality Systems, for example. This team will initially grade a set of samples into a range of references for calibration purposes. The samples will either have been specifically formulated to illustrate



15

specific variations in sensory quality or have been sampled from production over different shifts and time points. In some instances the customer of the product will take the lead in defining the calibration range. The selected set will represent target quality plus acceptable and unacceptable deviations. These samples will then be used to train an assessment panel to recognise the defined quality grades. King et al. (2002) describe a sensory quality system developed on this basis that has proved to work very effectively at a global level. Although with this approach the target quality can be controlled very well to the defined limits, it may be skewed away from what the consumer ideally desires. A customer, for example from within Food Service, may use a combination of key criteria quite differently and set different priorities from those of consumers when judging the sensory quality. Consumer research on the calibration set is therefore advised at some point to check that the specified quality adequately meets with the expectations of the final user. Sales data and consumer complaint information If the production of a product is well established then sales and also consumer complaint data may be useful in helping ensure that the typical production quality satisfies consumers. Sales volume figures will indicate how well the product is being received in the market and show market share compared with key competitors. The comparison of product quality with that of competitors can be useful to help highlight which characteristics might benefit from modification. Tracking trends in consumer complaint data to identify which features of a product cause the most issues has limited use in the main for monitoring sensory quality, especially in relation to flavour and texture. Sensoryrelated complaints most commonly refer to visual factors and in particular defects as consumers can most readily identify and describe them. Flavour and texture descriptions tend to be vague and can often be misleading about the cause of the issue. The use of this type of data is recommended only as a secondary guide for checking the sensory quality as it is very much a retrospective approach so by the time any significant issues have been highlighted, product reputation may already have been harmed in the market.

1.6 External support and consultancy Not all business can invest in a dedicated sensory professional and therefore make use of external sensory science expertise. The advantages can be to help speed up the implementation process in particular for: • developing a training program; • planning the implementation stages;


16


• auditing current practice; • training the sensory coordinator plus pool of assessors; • defining suitable methods plus procedures. Expert advice can help ensure that all important criteria needed for the reliable functioning of a system are highlighted, the most effective methods are utilised and a suitable standard of training is provided. Continued support from a sensory professional is also recommended to provide ongoing guidance and help with future training and maintaining the robust functioning of the system as required. The availability of an external expert may be restricted due to other professional duties, therefore contact well ahead of a proposed implementation schedule is advised to agree responsibilities and ensure the required commitment can be provided.

1.7 References costell, e. (2002) A comparison of sensory methods in quality control. Food Quality and Preference, 13, 6, 341–353. everitt, m.a. (2009) Consumer targeted sensory quality. In Global Issues in Food Science and Technology (G. Barbosa-Canovas, A. Mortimer, D. Lineback, W. Spiess, K. Buckle and P. Coionna, Ed.), Vol. 8, 117–128. Academic Press. findlay, c. (2002) Computers and the internet in sensory quality control. Food Quality and Preference, 13, 6, 323–428. gavin, d.a. (1984) What does product quality really mean? MIT Sloane Management Review. 15 October, 25–35. international organization for standardization (2007) Sensory analysis: general guidance for the design of test rooms. ISO 8589 Edition 2. king, s., gillette, m., titman, d., adams, j. and ridgely, m. (2002) The Sensory Quality System: a global quality control solution. In Food Quality and Preference (H.J.H. MacFie and H.L. Meiselman. Ed. A.M. Muñoz. Guest Ed.) Vol 13, 6, 385–395, Elsevier. lawless, h.t. and heymann, h. (1998) Sensory Evaluation of Food: principles and practices, Vol. 16, 548–577. Chapman & Hall. muñoz, a.m. (2002) Sensory evaluation in quality control: an overview, new developments and future opportunities. Food Quality and Preference, 13, 6, 329–339. muñoz, a.m., civille, g.v. and carr, b.t. (1992) Sensory Evaluation in Quality Control. Van Nostrand Reinhold. pecore, s. and kellen, l. (2002) A consumer-focused QC/sensory program in the food industry. Food Quality and Preference, 13, 6, 369–374.


2 Selection and management of staff for sensory quality control E. De Vos, Tate & Lyle Food and Industrial Ingredients, EMEA, France

Abstract: This chapter will discuss the ‘human’ factor behind sensory quality control. It will start by indicating the different people needed for sensory quality control and their particular profiles and tasks. The steps to be taken for starting a panel will be detailed from recruitment, to screening and training, to final selection of suitable panellists. Health and ethics aspects of sensory activities will be considered. Also, the ways to maintain panel performance, motivation and panel size will be discussed, as these topics are key in ensuring reliable and consistent panel results. A number of issues that may occur when setting up and running a panel will be touched on briefly, and a case study on setting up panels for quality control of cereal-based ingredients will give an example of how panel implementation and maintenance are addressed at Tate & Lyle. The chapter concludes with some comments on likely future trends and an overview of possible sources of additional advice. Key words: sensory quality control panel set-up, sensory personnel, sensory panel performance monitoring, sensory panel motivation, cereal-based ingredients, sensory quality control.

2.1 Introduction Product quality and consistency have traditionally been monitored by analytical techniques in areas such as microbiology or chemistry. These techniques, however, usually do not give adequate information towards the organoleptic quality and the appreciation of this organoleptic quality by the end user or consumer, as they cannot measure human perception. For example, it is possible that a small amount of a chemical that causes taints at a very small dose remains unspotted via chemical analysis while it does give rise to an off-note. Also, a product may conform to specifications for, for example, sugar and salt, but if each of these components was present in levels just below the maximum specification level, synergy and interaction


18


between them may result in a product that has a too strongly perceived intensity of sweetness, saltiness or both. Analytical techniques alone are not enough to ensure a full product quality picture, and should therefore be complemented by sensory analysis using people as judges. As Simpson (2003) states, ‘the human “instrument” displays a response that in some cases is more sensitive, and in all cases more complex, than any chemical instrument yet developed’. Originally, often one single expert taster or grader decided on the organoleptic quality of a product. However, these people were not infallible, as their evaluation was often psychologically biased towards their own preferences or skewed for physiological reasons such as anosmia. Currently, only a few specialised branches of industry, such as tea, coffee and wine production, use this method, whereas elsewhere alternative sensory techniques have been established that try to minimise bias and improve reliability by using a group of people.

2.2 Personnel required for sensory quality control Conducting a sound sensory quality control (QC) program not only involves tasters, but also people for setting up and maintaining the program and for ensuring that results are fed back to the responsible quality and manufacturing staff in a timely, consistent and accurate manner, so that the necessary preventive or corrective actions can be taken. The number and background of people involved in sensory analysis and their method of working will differ between larger companies with multiple affiliates and a smaller company consisting of only one or two separately operating units. Larger companies with multiple affiliates typically have a sensory department linked to corporate R&D, consisting of one or more scientists dedicated to sensory analysis and/or consumer research and one or more panels. When implementing a QC program across the different affiliates of the company, the methodology is often developed by the corporate team in close communication with the affiliate. Corporate sensory people set up and train the local panels or at least assist in these tasks. Afterwards, the affiliates take the responsibility for gearing sensory tests, maintaining the panel and feeding back the sensory quality results while staying in contact with the corporate sensory department so that when issues occur, the corporate sensory department can be called upon for assistance. Smaller companies typically do not have a dedicated sensory department. In many cases a member of the QC department takes up the responsibility for sensory method development, panel set-up and panel maintenance as an additional task in his or her job description. A number of authors have reported on the particularities of implementing a sensory panel in small processing operations or in multiplant organisations; amongst them are


Selection and management of staff for sensory quality control

19

Carlton (1985), Ford (1991), King et al. (2002), Mastrian (1985) and Stouffer (1985). Whether in a smaller or larger company, a number of key people are required for establishing a sensory QC program. These typically include a sensory responsible, preferably a sensory professional, one or more technicians and suitable panellists. Additionally, other resources such as specialists in statistics or external consultants in sensory might be called upon ad hoc. The International Organization for Standardization provides guidance on sensory staff responsibilities (ISO, 2006a) and recruitment and training of panel leaders (ISO, 2006b), and is a good source for further reading on this topic. 2.2.1 Sensory responsible The first task of any sensory responsible starting a sensory QC program is to sell the program to the organisation (Rutenbeck, 1985), in order to demonstrate the goals, the functions and the added value of the program and to generate interest, as well as to obtain the necessary management support and commitment and sufficient budgets. The tasks of a sensory responsible, often referred to as ‘panel leader’ or ‘panel supervisor’, are setting up the required sensory facilities and providing the necessary equipment or tools, recruitment and training of panellists, and maintaining panel performance, panel size and motivation after implementation of the panel. This person is responsible for setting up the quality control program, selecting the most suitable test methodology and sample presentation, analysing and interpreting the results, deciding on how much information is disclosed to panellists, and disseminating results to the involved departments. It is important that a person who takes up the sensory responsible role has a certain profile, as not everyone is suited to performing this task. A sensory panel leader or panel supervisor preferably has a food science background and sensory science knowledge and expertise. Strong organisation and motivation skills are key. A panel leader should have an interest in people and should be able to take a leading role without biasing the panellists’ opinions. A good panel leader needs basic statistical knowledge and skills, is able to translate problems into sensory tests and is able to make correct interpretations of sensory results. Good communication and cooperation with management and other departments – for sensory QC mainly the production plant and the quality department – are key in order to trigger the right actions. 2.2.2 Support staff The assistant, often a lab technician, needs to make sure the samples are prepared and presented to the panellists according to the right procedure and is present in the panel room for the duration of the test. This person also manages adequate cleaning and maintenance of the panel room and


20


materials. Taking inventory of supplies such as cups, mineral water for rinsing, crackers for cleansing the palate and tissues is another important task, as this will allow for on-time and organised sensory testing. The technician is also responsible for record-keeping. The assistant, as well as the panel leader, need to be precise and accurate in performing their duties. The technicians should receive proper training on the methodologies that are used in sensory testing from the sensory responsible. Clearly, if an assistant is not available, the panel leader will perform the tasks of the assistant.

2.2.3 Panellists As for all other quality control measurements, the ‘instrument’ in sensory evaluation is critical. The ‘instrument’ in this case is the panel, made up from a number of panellists who are responsible for the day-to-day evaluation of raw materials and/or finished products. These people should be aware of the basic principles of good sensory practice and should be screened and trained in order to guarantee that they are suitable for the job at hand. Particular attention should be paid to their ability to recognise the presence of off-flavours or taints. Section 2.3 step by step illustrates the different phases to undertake when setting up a panel.

2.2.4 Other resources As sensory data acquisition often is automated via a software package, it is wise to involve the information services and information technology (IS/ IT) department or a computer specialist in the loop, to avoid issues that might arise with systems and networks. Also, an in-house or external statistician can come in handy in case specific questions regarding test set-up or data analysis need to be answered. External consultants in the sensory field may be called upon to help with initiating, implementing and validating the QC program, in case of need.

2.3 Setting up a quality control (QC) panel Setting up a QC panel implies a number of consecutive stages: pre-screening and recruitment, screening, training, final selection and validation/follow-up. The first stages aim at obtaining the required sensitivities and precision, and the last stage is needed to control and maintain the panel’s efficiency. Before starting any sensory QC program, management must provide full support for all stages of panel set-up, maintenance and activities of the panel. Hence, the first task of any panel leader starting a sensory QC program is – as mentioned earlier – to sell the program to the organisation.



21

Just as the quality of data obtained from instruments to a large extent depends on the instrument’s good functioning and correct calibration, similarly sensory data will depend greatly on the functioning of the panel which in turn depends on adequate recruitment, screening, training and selection. Controlled methods for panel recruitment, screening and training will allow biases that otherwise would influence panel results to be minimised. Obviously, panellists’ personal likes and dislikes should be put aside when participating in sensory QC panel work.

2.3.1 Pre-screening and recruitment Important questions arise when recruiting people to form a sensory panel (ISO, 1993): where should one look for the people, how many people should be selected and how shall they be selected? The health and ethics aspects of panel recruitment – think about allergy problems and problems with food avoidance by ethnic and religious groups – should obviously be considered during the process. Where to look for potential panellists Candidates for panels in general can be recruited within the company or externally from the neighbouring region, or both. There are advantages and disadvantages to employing external panellists for QC sensory. For quality control, answers are needed rapidly and at times linked to the production scheme in order to get quick guidance on product release. Although external panellists are employed solely for a sensory purpose and have no other commitments, the facts that they are not on the spot and are available only for fixed times during the week may become a drawback since test planning will be less flexible. The other disadvantage of using external panellists is the increased cost as they need to receive compensation and the follow-up of paperwork requires time and money (ISO, 1993; Lyon, 2002). On the other hand, the ‘real’ cost of running such a panel is relatively low since panellists are paid as casual staff, without overhead loading, and company staff are not diverted from their main roles (Kilcast, 1992). Another disadvantage of recruiting externally is the fact that people might leave at short notice, and that confidentiality of the results might be an issue. Since sensory QC typically requires people who know the products under evaluation thoroughly, internal panellists seem more suitable for QC purposes. The major disadvantage of using internal assessors is that their ability to participate in sensory sessions depends on their main job role, and their agenda. Time away from the main job may be a concern. Lawless and Heymann (1997) indicate that panel participation can, however, be a welcome break for workers, can enhance their sense of participation in corporate quality programs, can expand their job skills and their view of manufacturing, and does not necessarily result in a loss of productivity.


22


How many candidates should be recruited When setting up a panel it is generally recommended to recruit 1.5 to 3 times as many potential panellists as finally needed (Cross et al., 1978; ISO, 1993; Lawless and Heyman, 1997; Lyon et al., 1992) because the final number of panellists available for a test will most likely be brought down owing to insufficient sensitivity to quality defects or absence of panellists because of illness, abandon, lack of motivation or job migration. Recruiting enough potential panellists from the start increases the chances of ending up with a sufficiently large panel size. Unfortunately, not all companies have the luxury of starting with a sufficiently big pool of potential panellists. In reality, cost, resource and time constraints often limit the number of panellists to the available employees in the production plant or the quality department. When only a few people are available, it remains important to get an idea of the capabilities of these people and to screen and train them, even if it will be more difficult to turn down potential candidates at the end of the process. As multiple methodologies for sensory QC exist (cf. Chapter 4), the final panel size that is needed will be dictated by the applied methodology. The amount of training that panellists receive, the reproducibility of panellist results and the variability of the products under test will have an influence on the required number of panellists as well. It is clear that the panel should contain enough members to cater for and minimise variability in results. When resources are limited, it is important to thoroughly train the panellists and allow them to gain a lot of experience: smaller but highly trained and very sensitive panels usually lead to more reliable results than panels with more members that are less trained and less sensitive. How to screen and recruit Before recruiting candidates, the panel leader should first explain the aim of setting up a sensory panel and the merits it will bring to both the company and the panellist, what is going to happen during the screening and training process and during the actual evaluation sessions afterwards and what is expected from the potential panellists in terms of commitment and time requirements. When designing and executing sensory tests – including screening and training sessions – the IFST guidelines for ‘Ethical and Professional Practices for the Sensory Analysis of Foods’ (IFST, 2010) should be given full consideration. This document mentions the following general principles: • The scope of permitted tests using human subjects should be defined in a written Organisation Ethical Policy that will depend on the nature of each individual organisation, but should typically comprise an internal mechanism to define and monitor ethical procedures together with expert input from external sources where appropriate.



23

• Procedures should be carried out in such a way as to reduce any risk to the health of the participants by excluding individuals at risk (e.g. allergic candidates), to ensure that products under test are microbiologically safe and to provide an ingredient list where required. For QC sensory testing in particular, any information that might be relevant to posssible unidentified hazards should be explained. • Participants should be volunteers who should be able to withdraw from the testing at any time, without having to give reasons. When novel foods or ingredients, foods containing ingredients that are not approved in the country in which the test is carried out, or foods produced by novel processes are to be evaluated, a safety risk assessment has to be made and informed consent must be given by the panellists before starting any sensory tests. Reference can be made to the guidelines published by, for example, the Advisory Committee on Novel Foods and Processes (ACNFP, 1992). Once potential candidates are aware of what will be expected from them, a questionnaire or a personal interview will provide the panel leader with extra background information. The following aspects of the candidates are worth investigating (ASTM, 1981; ISO, 1993): • Availability: candidates should not only be regularly available to attend training sessions – as the learning curve is steepest during this time – but also during routine tasting sessions. It is wise to exclude candidates who travel frequently or have continual heavy workloads, temporary employees or students. • Health: candidates should be in good health, and should not suffer from any disabilities which may affect their senses, or from any allergies or illnesses, and should not take medication which might impair their sensory capacities and thus affect the reliability of their judgements. It may be useful to know whether candidates have dental prostheses, since they can have an influence in certain types of evaluation involving texture or flavour. Colds or temporary conditions (for instance, pregnancy) should not be a reason for eliminating a candidate. • Motivation, willingness and interest: panellists who are volunteering to participate and are interested in sensory analysis and the products to be investigated are to be preferred over panellists who are forced to participate and are not interested as the former are more likely to become better assessors. According to Howard (1972), ‘interest is linked to the increase of a panellist’s capacity while following the different stages’. • Character and personality: persons with extreme strong or weak characters often are not ideal candidates (Costell, 1983) as they either want to dominate the group or do not speak their minds. Candidates must show interest and motivation, must be punctual in attending sessions, reliable and honest and have a positive attitude to the use of sensory analysis.


24


• Dislikes and attitudes to foods: strong dislikes for certain foods or beverages that are under investigation need to be determined, as well as other reasons for not consuming certain products because of cultural, ethnic, health or other reasons. Panellists should be willing to taste and smell all test products as part of their learning experience. • Knowledge and aptitude: panellists require certain physical and intellectual abilities, in particular the capacity to concentrate and remain unaffected by external influences. Certainly for quality tests, a detailed knowledge of the product under test is beneficial. • Ability to communicate: panellists should have the ability to express themselves clearly so as to be able to communicate the sensations they perceive. Other factors such as educational background, experience in sensory analysis or smoking habits may also be recorded, but exclusion of candidates should not be based on these. When gathering information on panellists, the panel leader has to make sure that recording of data is in accordance with the provisions of relevant data protection legislation of the country concerned (IFST, 2010). In France for example, the ‘Commission Nationale de l’Informatique et des Libertés’ (French Data Protection Authority) has issued an Act (CNIL, 2009) relating to the protection of individuals with regards to the processing of personal data. The information provided by panellists and kept on file should be treated as strictly confidential. Panellists should have the option to change the provided information at any time. The information gathered via the questionnaire or the personal interview will allow for selection of the candidates who are deemed suitable to proceed to the screening phase. 2.3.2 Screening Screening should not be considered as part of the training stage, but rather as a tool to eliminate candidates who cannot detect large differences in attributes. ASTM (1981) defines screening as ‘the assessment of candidate potential and a precursor to training’. Basically, screening will allow potential sensory impairments to be checked and will give preliminary information on the candidate’s capabilities of recognising and describing basic odours, tastes and certain taints before starting the training. Even if a candidate has prior sensory knowledge, he or she should be screened to avoid including persons that are not able to discriminate between or are not sufficiently sensitive to flavours of interest. There is no specific screening method: the screening tests and the screening standards to be used will depend on the senses that will be required, and the food products and properties that have to be assessed. It is therefore recommended that the same product(s) are used during the screening as the one(s) that will finally be evaluated (Sidel et al., 1981). Also, some



25

scaling exercises (Meilgaard et al., 2006) might be included to assess the concept of ‘proportion’ to the potential candidates. Potential candidates should be screened on colour vision (Ishihara, 1971) and the incidence of ageusia or aneusmia (ISO, 1991). They should, depending on the requirements of the test situation, be subjected to matching tests, which will determine whether candidates are able to match test samples at well above threshold levels to standard samples, triangle tests for detection of a stimulus, ranking tests for discrimination between levels of intensity of a stimulus and/or tests to determine their descriptive ability, as described in ISO standard 8586-1 (1993). Based on the results of the screening tests, candidates with abnormal threshold levels should be excluded from the training, but candidates who show an ability to discriminate and who are also consistent and reproducible in their discrimination should be included. As ability and sensitivity increase with training, screening criteria should not be excessively harsh and should take into account candidates’ potential rather than their current performance. Candidates with high success rates at the screening stage are to be expected to be more useful than others, but those who show better results with repetition are likely to respond well to training.

2.3.3 Training Having passed the screening stage, suitable candidates proceed to the training stage. The aim of training is threefold (ASTM, 1981; ISO, 1993): 1. To provide assessors with rudimentary knowledge of procedures used in sensory analysis in order to familiarise the individuals with the test procedures. 2. To improve the individual’s ability to detect, recognise and describe sensory stimuli. 3. To train assessors to use this expertise and to improve the individual’s sensitivity to and memory for test attributes, so that they may become proficient in the use of such methods with particular products and that sensory judgements will be precise and consistent. Familiarisation with sensory test procedures Training should include an introduction to sensory analysis, the senses and how to use the senses. Candidates should be introduced to a number of general codes of conduct regulations, in order to ensure good sensory practice: • • • • •

be punctual; do not eat, drink, smoke 30 minutes before the session; do not use (strong) perfume, aftershave or lipstick; keep silent in the panel room; be objective and disregard personal likes and dislikes.


26


Additionally, panellists should be instructed to take into account the correct order of product evaluation, starting with colour/appearance, afterwards odour, texture, flavour (aroma and taste) and finally aftertaste (ISO, 1993). Indications should be given on the sample size that should be considered, whether swallowing is mandatory or spitting is allowed, whether to use, for example, crackers or just plain water for rinsing, the interval between sips/ bites, etc. (ASTM, 1981). Candidates should be made familiar with the methodology and format of the test, the forms to fill out or the computerised data acquisition system. Once assessors fully understand the sensory tests they have to perform, and are familiar with the specific knowledge required to perform the tests correctly, they can proceed to further training. Basic training: improving the ability to detect, recognise and describe sensory stimuli Whatever the methodology used for QC sensory testing, thorough training using the adequate reference and training standards must make sure that the variability in sensory response inherent in any group of people is reduced. This lengthy program of training is essential in ensuring that attributes causing quality defects are captured on time and that at the same time individual (over)sensitivities to certain attributes do not bias results so that defective products are not released for sale. It is obvious that the main aim of training is to familiarise potential panellists with standards, deviations from the standards and the limits of acceptable variation. Therefore, potential panellists should, again depending on the requirements of the test situation, be subjected to different training sessions. Training in detection and recognition of tastes and odours by means of various methods, for example matching, recognition, paired comparison, triangle and duo–trio tests, demonstrates tastes or odours at high and low concentrations and trains candidates in recognising and describing them correctly (ISO, 1991, 1993, 2006c). Training in the use of scales introduces candidates to the concepts of rating, classification, interval and ratio scales by initially ranking a series of single-odour, single-taste and single-textural stimuli with respect to the intensity of a particular characteristic. The various rating procedures are then used to attach meaningful magnitudes to the samples (ISO, 1993). Although profiling is less used for QC purposes because of its lengthy procedure to obtain results, an introduction to the development and the use of descriptors can be of interest to make candidates aware of the idea of profiling. Panellists are presented with a series of simple products and are asked to develop vocabularies for describing their sensory characteristics, in particular terms which allow samples to be differentiated (ISO, 1985, 1993). It is important to include sufficient time for exercises during the basic training sessions, in order to allow the candidates to gradually build up



27

experience. During the basic training stage, the panel leader carries out a follow-up evaluation to assess the interest and dedication of each panellist and double-checks their sensory capacities and correct application of the sensory methodology. Assessors should be encouraged and motivated by immediate feedback of results after each session. Extended training: building expertise After basic training, assessors should undergo a period of extended product training to familiarise themselves with the type of products that will be tested on a routine basis once the panel has been established, the relevant sensory parameters for the products at hand and the acceptable range of variation of these parameters and products. The type of training sessions will depend on the particular QC test situation, and may, for example, comprise difference assessments in which samples similar to those that will eventually be evaluated are presented to the candidates, who evaluate them using one of the different assessment procedures (ISO, 1993). In order to carry out effective quality control, panellists must have a depth of knowledge which can be gained only by long periods of exposure to the product range and defects which are likely to occur. Lyon et al. (1992) indicate that panellists must be able to make allowances for normal withinbatch variation, or batch-to-batch variation. Knowledge of normal product variation within a batch or between batches comes with experience over time. Repeated exposure to the products and the defects or off-taints that may occur will help build this experience. Once panellists have been tasting products for a while and show good repeatability, noteworthy acuity, or particular aptitude regarding specific attributes (e.g. a taint) or classes of materials, they can be taken to a higher level by improving their memory for sensory attributes, learning to keep clear and logical notes, gaining background knowledge on the range of products from lectures, books, trade press and technical contacts, gaining knowledge of technical aspects such as raw materials, production and distribution of the products concerned, increasing their communication skills with other experts and with non-experts and increasing their self-discipline (ISO, 2008). Additional training sessions that focus on these topics can help in further building the knowledge base and experience.

2.3.4 Selection Once the training stage is completed, it is time to select the most suitable candidates to form a panel based on the results that they have obtained during repeated training exercises. Selected candidates should perform consistently, should be able to differentiate the samples that are presented in difference assessments and should be able to rank the samples that are presented in ranking tests. The ability to detect adulterated samples at


28


decreasing concentrations can also be used as a criterion for selection. Candidates who perform these tasks less well than others should be rejected (ISO, 1993). Data obtained from rating and scoring exercises should be analysed by ANOVA to examine the individual results of each assessor. Assessors who have a high residual standard deviation, indicating inconsistency, or for whom the variation among the samples is not significant, indicating poor discrimination, should be considered for rejection (ISO, 1993). No additional specific selection procedure is advocated, amongst those already outlined, for qualitative descriptive analysis. Focus here should be on development and use of descriptors (profiles) and descriptive assessment (ISO, 1993). Final selection of members of the panel will be made according to panellists’ availability, sensitivity, reproducibility and their capability to provide valid sensory data. Failing to identify and eliminate non-discriminators before a test will increase the likelihood of missing real product differences (Sidel et al., 1981). The final number of panellists will be defined by the QC methodology that is applied. In any case, all suitable panellists should become members of the ‘pool’ from which to draw panellists in the future. In the event that after training only a limited number of suitable panellists remain, it is recommended not to include people who have less than satisfactory training results just to achieve a predetermined panel size (Cross et al., 1978).

2.4 Maintaining the quality control (QC) panel: performance, motivation and size 2.4.1 Monitoring panel performance After implementation of the panel and having started routine testing, it remains critical to monitor panellist performance regularly, as one needs to certify that panellists are able to repeatedly achieve accurate, precise and reproducible results (ISO, 1993; Muñoz et al., 1992). The monitoring methodology will depend on criteria derived from the particular method that is used for QC testing, but should be relative to the panel as a whole and to the panellist’s past performance. The monitoring program should make use of blind controls or blind deviating samples that are included in routine tests. Unacceptable variation or disagreement in the results of these samples should trigger retraining for individual panellists or the panel as a whole. Performance records should be kept and periodically reviewed. When panellists are retrained or training is given to potential new panellists, data should only be included in test results when they are in sufficient agreement with data from previously qualified panellists. Repeated review sessions and practice at recognising each parameter and the change in intensity of each parameter caused by spiking higher



29

levels of certain compounds in the sample is important to develop and maintain the skills required. Training and monitoring should be a continuing process that is repeated regularly during the normal operation of the panel. The best way to maintain the high competence level of the group is to keep on testing regularly. Frequent participation is necessary for motivation and to guarantee performance (ISO, 2005): panellists can indeed improve on their performance with repeated trials. When similar panels are used across different plants, in different countries, one has to ensure that for the same sample the (same) QC method gives the same result, even though the individual panellists are different. A nice tool to check for alignment is a round robin test, which is typically managed by the corporate sensory department. On a regular basis, typically monthly or bi-monthly, a blind sample is sent out to all participating panels, after which evaluation is performed locally and results are sent back to the corporate sensory department. Statistical tests are then performed on the received data and conclusions are drawn on the performance of the different panels. The conclusions are fed back to the participants, to ensure proper remedial actions, if required.

2.4.2 Panel motivation Even when good sensory practice is assured, panellists can start losing their motivation. They may become bored when they frequently participate in routine evaluation sessions. Lack of feedback on their results and lack of support or recognition from management are other factors that can decrease a panellist’s drive. In order to keep the panel performing adequately, it is important to maintain interest and motivation. Panellists should have faith in the group and should like to come to sessions. A ‘motivation maintenance’ program can include a number of activities: • Give information on why a particular test is needed and performed. In order not to bias panellists with details on the product samples, it is important to supply information on a test or a series of tests only after all related tests have been finished. • Inform the panel regularly on individual performance and the performance of the panel as a whole. Positive feedback will give a boost to the panellist’s self-confidence and motivation. Negative feedback, when given in a tactful, constructive manner, will help to make panellists aware of their weak points and can trigger individual retraining actions. • Give a small snack after sessions that include samples with a bad taste. • Reward the panellists who come to panel sessions regularly and/or perform well. Although money can be an incentive, most companies offer other incentives such as personalised ‘panel’ gift items such as mugs, umbrellas, travel clocks or pens, tickets for the movies, chocolate


30


goodie bags or gift vouchers, or organised special activities in or outside the company, for example, special panel Christmas lunches or wine tasting sessions. Distributing raffle tickets for a prize draw can be a nice incentive to encourage attendance at panel sessions. Mentioning panel achievements in the company newsletter can stress the importance of the panel to management and the rest of the business. Recognising noteworthy individual panellist contributions in terms of attendance or correct identification of off-notes or blind control samples is a means of applauding the good work of panellists. • The sensory responsible, the plant, the QC and the panellist’s management should appreciate and acknowledge the efforts that panellists undertake in order to be able to attend sensory sessions. Panellists should be reminded regularly that their contribution is key in guaranteeing the panel’s good functioning and subsequent problem solving and support achievements. Obviously, communication of achievements to the rest of the business is important too. • Provided a larger pool of trained panellists is available, rotating the panel regularly can improve motivation and fight boredom. 2.4.3 Panel size It is important to maintain enough panellists in the panel over time, whatever the recommended panel size as dictated by the applied QC methodology. The sensory responsible should make sure that a sufficiently high number of panellists is always available in the pool to ensure adequate numbers during the actual evaluations. There are many reasons why panellists may quit the panel and abandon their sensory task: illness, people leaving the company, time constraints, working day(s) off-site and shifts. If the normal panel size cannot be guaranteed over time, new recruitment and training should be scheduled to fill the gap of those who have left and to ensure that the panel can keep on performing as it is meant to.

2.5 Possible issues A QC sensory panel undoubtedly has its added value. However, it is clear that setting up and maintaining a QC panel requires substantial efforts, both from sensory staff and panellists. Panel motivation and maintenance of panel performance and panel size may pose some challenges, but a couple of other issues may make routine running of sensory panels even more cumbersome. With internal panellists, the time away from the person’s main job might be a concern as time constraints in the manufacturing environment may not always allow people to leave their workplace to participate in sensory sessions.



31

When the plant operates in shifts, even more complexity is added as availability of panellists is in this case typically limited during certain times of day or night, or during weekends. When the available shift workers are thoroughly trained and have a long-lasting experience in tasting the product(s) at hand, this may still lead to reliable results; however it remains advisable to use multiple panellists instead of only one or two. When standardising sensory QC procedures and coordinating sensory activities across multiple plants in different geographical locations, issues related to cross-cultural differences might occur. Language and effective communication can be an issue (Cardello, 1993) as there may be subtle differences in wording or definitions, and difficulties in expressing oneself or in understanding what is meant in another language. As Carlton (1985) indicated, local customs may also need attention: it is sometimes critical to do some groundwork to become familiar with the country’s culture before the sensory work begins. Carlton (1985) stresses for example that knowing the ranking rules and where people fall within the rankings is important, that power and authority within the group can affect interaction as well, that gender-effect may undermine the credibility of a female trainer in a male-dominated social structure and that dressing a bit more casually might promote group interaction and individual participation. Another issue that may arise is the fact that, strictly speaking, consumer studies are needed to establish the range of acceptable products or the cut-off points (Muñoz et al., 1992). This obviously adds an extra cost to the sensory quality program.

2.6 Case study: selection and management of staff for sensory quality control of cereal-based ingredients As an ingredient supplier, Tate & Lyle performs sensory quality control on ingredient batches before releasing them to customers. Once the appropriate sensory method has been established at the Global R&D Sensory level, the method is implemented across the different production plants. For most ingredients, a degree-of-difference (DoD) from control method with a defined cut-off point for rejection is used. The panel set-up process in each production plant involves the Global R&D Sensory responsible and the local panel leader, typically someone working in the local QC department who is taking up this additional task. Potential panel candidates are recruited internally, from the local plant, depending on their availability, interest and basic sensory acuity. As soon as the pool of candidates is established, training starts. Initial training is given by the Global R&D Sensory responsible so as to train both the ‘trainer’, i.e. the (local) panel leader, and the panellists. The training program starts with an introduction of the method to the panel leader and the potential panellists and is followed by multiple


32


sessions that have the aim of training both qualitatively and quantitatively, using spiked training standards, spiked test samples and good as well as deviating production samples.

2.6.1 Qualitative training The qualitative training sessions show panellists the ‘types’ of (off-)notes that can occur in a specific ingredient category. Panellists receive a bland reference sample and spiked training standards with the (off-)notes in a concentration at the upper range of the DoD scale. Although this high concentration will not be likely to occur under normal production circumstances, it is used for familiarising the panellists with what is meant by a specific (off-)note and will help in identifying the (off-)note when it occurs in lower concentrations afterwards. After having memorised the different training standards, they are put aside, and two or more coded unknown samples are provided for the panellists to identify within the (off-)notes of the previously received set. When the right (off-)note type is mentioned, the panellist is awarded a score of 100% for that sample.

2.6.2 Quantitative training The quantitative training sessions show panellists the ‘intensity’ of (off-) notes that can occur in a specific ingredient category. Per (off-)note, panellists receive a bland reference sample and spiked training standards with the given (off-)note in concentrations at the lower, middle and upper range of the DoD scale. After having memorised the different concentration levels for that (off-)note, the training standards are put aside and a test sample with unknown concentration is provided for the panellists to identify within the concentrations of the previously received set. When the right (off-)note concentration is mentioned, the panellist is awarded a score of 100% for that off-note. The higher the deviation from the right concentration, the more points are deducted (e.g. when using a 0 to 10 scale, each deviation from the right concentration will result in a decrease of 10% in score). The above procedure is repeated using blind samples at various appropriate concentrations for each (off-)note.

2.6.3 Qualitative/quantitative training The set-up of the qualitative/quantitative training sessions is similar to the set-up of the quantitative training sessions, with the difference that the (off-)note in the range of concentrations is now unidentified. The unknown sample now must be identified both in terms of its type (qualitative) and its concentration (quantitative). The panellist is now awarded an average score between the qualitative part and the quantitative part.



33

2.6.4 Spiked samples Whereas the previous sessions were performed on specific training sheets, the next sessions start making use of the routine evaluation sheet to familiarise the panellists with the normal evaluation procedure and form. The ‘spiked’ sessions consolidate the learnings from the previous training sessions. Panellists now receive a bland reference and three unknown samples, of different type and concentration, which they have to define and score on the form using the DoD scale. Again, panellists are awarded an average of the qualitative and quantitative scores.

2.6.5 Production samples The ‘production samples’ sessions are similar to the ‘spiked samples’ sessions, but the unknown samples are now not artificially spiked as before, but are real samples provided by the production plant. Again, panellists receive a bland reference and three unknown samples, for which they have to fill out the form using the DoD scale. The panellists’ result now is not a percentage, but relates to a benchmark score from earlier tests by, for example, Global R&D Sensory or a production plant.

2.6.6 Training results After having passed the full training program, panellists get a final score, which is an average over the scores of the different sessions. In order to become a member of the routine panel, a minimum score of 70% is required. Panellists scoring lower than this 70% are – at least temporarily – kept in the pool, but need additional training and re-testing to check if they are found to be fit for the panel after having gained more experience. In the meantime, their routine evaluation results are not taken into account in the final panel result for a given sample. During each of the training sessions, the result of the panel as a whole is checked also. Obviously this result should be in line with what is expected, but can deviate because of individual panellists’ scoring deviations.

2.6.7 Panel maintenance and follow-up All subsequent re-training and monitoring in the weeks/months following the initial training is the responsibility of the local panel leader and may involve Global R&D Sensory when needed. A yearly round robin scheme is maintained by Global R&D Sensory as a tool to follow-up the results of the panels in the different production units and to determine whether or not corrective re-training actions need to be taken. Therefore, one production plant sends out a sample to all the other plants in the scheme every 2 months. All panels evaluate the sample and the results are sent back to Global R&D Sensory, after which statistical


34


techniques are used to analyse panel and panellist reliability. Feedback is reported to the involved panels, after which corrective actions can be taken depending on the need.

2.7 Future trends People involved in product quality may be inclined to replace their regular ‘human’ panels by devices such as ‘electronic noses’ or ‘electronic tongues’. Properties of these devices should however be considered before deciding on this action. An electronic nose is a device that – after sufficient training with several training samples, both conforming to and deviating from the reference standard – allows differentiating samples on the basis of volatile components and to group similar samples (Gardner and Bartlett, 1999). An electronic tongue is a device with a similar working principle as an electronic nose, but is different in that the differentiation between the samples is done on the basis of water solubility instead of volatility of components (Legin et al., 1997). The major drawback of electronic nose and electronic tongue devices is that, although they are able to group similar samples together on the basis of differences in odour (volatile compounds) or taste (water-soluble compounds), they do not pinpoint what the cause of the difference is. This makes it difficult to address issues in the production process and to take remedial actions. This drawback could be countered by adding a mass spectrometer to the system, as this allows compounds to be separated and defined based on their mass. Electronic nose or tongue devices can have their place in quality control provided that their limitations are well known. However, they will never be able to replace a human panel, as they are not as sensitive as the human palate and are not able to evaluate complex matrices in the same way as humans.

2.8 Sources of further information and advice Most information in this chapter is based on the guidelines for panel set-up as defined by ISO (1991, 1993, 2006c, 2008), but has been adapted for use in a QC environment. ASTM standard 758 (1981) is an additional source of further, general panel set-up information. Specific methods to initiate a sensory QC/QA program and the issues related to this have been well documented by, for example, Beckley and Kroll (1996), Costell (1992), Jellinek (1985), King et al. (2002), Lawless and Heyman (1997), Lyon et al. (1992), Muñoz et al. (1992) and Yantis (1992).



35

Additionally, Campden and Chorleywood Food Research Association (CFRA) have published guidelines that are principally aimed at descriptive panels, but do include some useful information that can be adapted towards QC panels: Guidelines for the Selection and Training of Assessors for Descriptive Sensory Analysis (Lyon, 2002), Guidelines for the Motivation of Sensory Panels Within the Workplace (Kapparis et al., 2008) and Practical Guidelines for Monitoring On-going Job Performance of Sensory Descriptive Panellists (Pfeiffer et al., 2008).

2.9 References acnfp (1992), ‘Guidelines on the Conduct of Taste Trials Involving Novel Foods or Foods Produced by Novel Processes’, Advisory Committee on Novel Foods and Processes, available from http://www.acnfp.gov.uk/acnfppapers/inforelatass/ guidetastehuman/guidetaste. astm (1981), Guidelines for the Selection and Training of Sensory Panel Members, American Society for Testing and Materials, Special Technical Publication 758. beckley j p and kroll d r (1996), ‘Searching for sensory research excellence’, Food Technol., 50(2), 61–63. cardello a v (1993), ‘Cross-cultural sensory testing: a changing tide?’, Cereal Foods World, 38(9), 699–701. carlton d k (1985), ‘Plant sensory evaluation within a multiplant organization’, Food Technol., 39(11), 130–133, 142. cnil (2009), Act no. 78-17 of 6 January 1978 on Data Processing, Data Files and Individual Liberties, Amended by the Act of 6 August 2004 relating to the protection of individuals with regard to the processing of personal data and by the Act of 12 May 2009 relating to the simplification and clarification of law and lightening of procedures, Commission Nationale de l’Informatique et des Libertés, available from http://www.cnil.fr/fileadmin/documents/en/Act78-17VA.pdf costell e (1983), ‘El equipo de catadores como instrumento de analisis’, Rev. Agroquim. Tecnol. Aliment., 23(1), 1–10. costell e (1992), ‘Sensory analysis applied to quality control of citrus fruits’, Rev. Esp. Cienc. Tecnol. Aliment., 32(3), 269–281. cross h r, moen r and stanfield m s (1978), ‘Training and testing of judges for sensory analysis of meat quality’, Food Technol., 32(7), 48–54. ford a l (1991), ‘Sensory testing in a small company’, Food Australia, 43(6), 237. gardner j w and bartlett p n (1999), Electronic Noses. Principles and Applications, Oxford University Press. howard a (1972), ‘Taste panel technique. I. Reproducibility, reliability and validity’, Food Res. Quart., 32, 80. ifst (2010), ‘Ethical and Professional Practices for the Sensory Analysis of Foods’, The Institute for Food Science & Technology. Available from http://www.ifst.org/ documents/misc/practicesforsensoryanalysis1.pdf ishihara s (1971), Tests for Colour Blindness, Kanahara Shuppan Co.Ltd., TokyoKyoto, Japan. iso (1985), Sensory Analysis – Methodology – Flavour profile methods, ISO 6564:1985(E), International Organization for Standardization. iso (1991), Sensory Analysis – Methodology – Method of investigating sensitivity of taste, ISO 3972:1991(E), International Organization for Standardization. iso (1993), Sensory Analysis – General guidance for the selection, training and monitoring of assessors – Part 1: Selected assessors, ISO 8586-1:1993(E), International Organization for Standardization.


36


iso (2005), Sensory Analysis – Methodology – General guidance, ISO 6658:2005(E), International Organization for Standardization. iso (2006a), Sensory Analysis – General guidance for the staff of a sensory evaluation laboratory – Part 1: Staff responsibilities, ISO 13300-1:2006, International Organization for Standardization. iso (2006b), Sensory Analysis – General guidance for the staff of a sensory evaluation laboratory – Part 2: Recruitment and training of panel leaders, ISO 13300-2:2006, International Organization for Standardization. iso (2006c), Sensory Analysis – Methodology – Initiation and training of assessors in the detection and recognition of odours, ISO 5496:2006(E), International Organization for Standardization. iso (2008), Sensory Analysis – General guidance for the selection, training and monitoring of assessors – Part 2: Expert sensory assessors, ISO 8586-2:2008(E), International Organization for Standardization. jellinek g (1985), Sensory Evaluation of Food, Chichester, Ellis Horwood Ltd. kapparis e, pfeiffer j c and gilbert c g (2008), Guidelines for the Motivation of Sensory Panels Within the Workplace, Guideline No. 57, Campden & Chorleywood Food Research Association Group. kilcast d (1992), ‘New developments in sensory analysis’, International Food Ingredients, 2, 2–8. king s, gillette m, titman d, adams j and ridgely m (2002), ‘The Sensory Quality System: a global quality control solution’, Food Qual. Pref., 13, 385–395. lawless h t and heymann h (1997), ‘Sensory evaluation in quality control’, in Lawless H T and Heymann H, Sensory Evaluation of Food – Principles and practices, New York, Chapman & Hall, 548–584. legin a, rudnitskaya a, vlasov y, di natale c, davide f and d’amico a (1997), ‘Tasting of beverages using an electronic tongue’, Sensors Actuators B: Chem., 44, 291–296. lyon d h (2002), Guidelines for the Selection and Training of Assessors for Descriptive Sensory Analysis, Guideline No.37, Campden & Chorleywood Food Research Association Group. lyon d h, francombe m a, hasdell t a and lawson k (1992), Guidelines for Sensory Analysis in Food Product Development and Quality Control, London, Chapman & Hall, 47–57, 82. mastrian l k (1985), ‘The sensory evaluation program within a small processing operation’, Food Technol., 39(11), 127–129. meilgaard m c, civille g v and carr b t (2006), Sensory Evaluation Techniques, 4th Ed., Boca Raton, FL, CRC Press, 163–165. muñoz a m, civille g v and carr b t (1992), Sensory Evaluation in Quality Control, New York, Van Nostrand Reinhold, 28–40, 84–107, 126–139, 154–167, 179–198, 225–226. pfeiffer j c, kapparis e and gilbert c g (2008), Practical Guidelines for Monitoring On-going Job Performance of Sensory Descriptive Panellists, Guideline No. 58, Campden & Chorleywood Food Research Association Group. rutenbeck s k (1985), ‘Initiating an in-plant quality control/sensory evaluation plan’, Food Technol., 39(11), 124–126. sidel j l, stone h and bloomquist j (1981), ‘Use and misuse of sensory evaluation in research and quality control’, J. Dairy Sci., 64, 2296–2302. simpson w j (2003), ‘Sensory quality management and its role in brewery operations’, Scandinavian Brewers’ Review, 60(5), 18–23. stouffer j c (1985), ‘Coordinating sensory evaluation in a multiplant operation’, Food Technol., 39(11), 134–135. yantis j e (1992), The Role of Sensory Analysis in Quality Control, American Society for Testing and Materials, Manual Series: MNL 14.


3 Proficiency testing of sensory panels G. Hyldig, Technical University of Denmark, Denmark

Abstract: When designing a proficiency test for sensory panels, it is important to consider the type of sensory panel and the material that is selected as proficiency test item. Most sensory panels work with more than one method and often with both descriptive and discriminative tests. To illustrate the importance of choosing the methods and proficiency test item, some trials are described in the chapter. Univariate as well as multivariate statistical methods can be used for the data analysis. The results from proficiency testing provide a tool that enables laboratories to evaluate and demonstrate the reliability of the data produced. Key words: proficiency testing, proficiency test item, sensory panels, sensory analysis.

3.1 Introduction What is proficiency testing and why is it important for sensory panels? A proficiency test can be used to validate laboratory performance. The ISO definition of laboratory proficiency testing is ‘determination of laboratory testing performance by means of interlaboratory comparisons’. It is a comparison of a laboratory’s reported result for the analyte in question with the best estimate of the ‘true’ value of the analyte and hence not a validation of the analytical method. There is no difference in procedure to produce results for a single commodity, internal quality control and in research projects; it must be reliable, and the result may not depend on which laboratory has performed the analysis. Many laboratories conducting chemical or bacteriological analysis participate in proficiency tests on a regular basis and have done for many years. In work with sensory panels, numerous efforts have been made to validate the assessor’s performance, which is of course essential, but it is not enough to ensure that the panel performance is reliable: therefore it is necessary for sensory panels to participate in proficiency tests. For laboratories that have been accredited for a specific analytical method, the accreditation bodies demand that the laboratories participate


38


in proficiency testing at regular intervals. However, it is not only laboratories with an accreditation that can benefit from participating in a proficiency test. Big companies with several laboratories in different locations can also use proficiency testing to validate the performance of their laboratories and to prove that the results are independent of geography. Proficiency testing can be very useful in large research projects with participation of laboratories working with the same analysis often situated in different countries. There are general standards for proficiency testing such as ASTM E130195 ‘Standard guide for proficiency testing by interlaboratory comparisons’; EA-4/09 (2003) ‘Accreditation for sensory testing laboratories’; EN ISO/ IEC 17025 (2005) ‘General requirements for the competence of testing and calibration laboratories’; ILAC-G13:2000 ‘Guidelines for the requirements for the competence of providers of proficiency testing schemes’; ISO 17025 (1999) ‘General requirements for the competence of calibration and testing laboratories’; ISO Guide 35:2006 ‘Reference materials – General and statistical principles for certification’; ISO/IEC 17025:2005 ‘General requirements for the competence of calibration and testing laboratories’; ISO/IEC Guide 43-1 (1997) ‘Proficiency testing by interlaboratory comparisons – Part 1: Development and operation of proficiency testing schemes’; and ISO/IEC Guide 43-2 (1997) ‘Proficiency testing by interlaboratory comparisons – Part 2: Selection and use of proficiency testing schemes by laboratory accreditation bodies’. The standards for the proficiency testing of chemical methods can be useful when setting up proficiency test for sensory methods. Organizations such as the ISO (International Organization for Standardization/Organisation internationale de normalisation), NMKL (Nordisk metodikkomite for levnedsmidler/Nordic Committee on Food Analysis), EA (European Accreditation), ASTM (American Society for Testing and Materials), ILAC (International Laboratory Accreditation Cooperation) and FAPAS (part of the Food and Environment Research Agency, an agency of the UK Government Department for Environment, Food and Rural Affairs) have made standards and guidelines for proficiency testing, validation and many of these are used by the different national accreditation bodies. In the reference list there are additional guidelines and standards. For objective sensory analysis of food, there are several challenges regarding the methods and materials that are chosen for the proficiency test, especially if the samples are going to be sent over long distances. In this chapter the challenges for a proficiency test of objective sensory panels will be outlined and some suggestions for the proficiency test will be made.

3.2 Design and implementation of proficiency testing It is important to consider the type of sensory panel and the material that is selected as a proficiency test item when designing a proficiency test for


Proficiency testing of sensory panels

39

sensory panels. A sensory panel may not be trained or be accredited for it to evaluate all kinds of food material; such a specialised panel could be a sensory panel where the assessors are tested, selected and trained to have a high performance to evaluate wine (ISO 5496, 1992, ISO 6658, 1985, ISO 8586-1, 1993, ISO 8586-2, 1994, ISO 8589, 1988, ISO/CD 13300 – 1, Part 1 and Part 2, 2002, NMKL Procedure No. 6, 1998). It can be extremely difficult for such a panel to evaluate, for example, vegetables. This must be taken into consideration, otherwise it would be similar to using a pH electrode specially designed for measuring pH in cheese for measuring pH in wine. Another matter is how to choose the material for a proficiency test item. Sizeable amounts of material are needed for sensory evaluation, and it has to be stored and transported in a way that keeps the sensory quality constant over time. If the test is international and performed on food from the marketplace as the proficiency test item, it must be taken into consideration that the same product can be modified for each market in a way that the product differs in sensory quality from country to country, implying that the sensory characteristic is also different. Another consideration in international proficiency testing is the vocabulary used. It is vital that the definition of the different attributes is exactly the same in all languages. Terminological consistency is key here. The coordinators are responsible for organising all activities involved in the proficiency test from designing and setting up the proficiency test to the final report. In addition to this, confidentiality is something that has to be taken into consideration. The outcome of the proficiency test can be critical for some laboratories. Normally the identities of the participating sensory panels are kept confidential and only known to a minimum of persons involved in the coordination group. It can also be practical to get an agreement about the rights to use the results from the proficiency test. An overview of the different steps in designing a proficiency test is outlined in Table 3.1. The first two steps comprise the selection of a method and material for the proficiency test item. Sections 3.2.1 and 3.2.2 discuss the selection processes in more detail. Step 3 is a guideline for preparing the proficiency test item and the execution of the test. The guideline must be written in clear and simple language and contain a detailed description of the preparation such as sample size, heat treatment, serving temperature, coding, serving material (e.g. coloured glass, white porcelain bowl) and which order to serve the item for the assessors, number of replicates, which palate cleaner to use between samples, how to collect the results, the time frame of the session, numbers of samples and breaks in a session. If necessary, there can also be a guideline for setting up the software program for executing the sensory test. Guidelines for training the sensory panels are made in step 4. Before the proficiency test, the sensory panel has to be trained in the vocabulary and the scale. The training will depend on the method and the proficiency test


40


Table 3.1 The different steps in a proficiency test Step

Description

Step 1 Step 2 Step 3

Selection of method Selection and test of material for the proficiency test items Guidelines for preparation of the proficiency test items and the execution of the test Training material and guidelines for training Coding and consignment of the proficiency test items to the participating laboratories Proficiency test Collection of data from the proficiency test round Analysis of data and report of the results

Step 4 Step 5 Step 6 Step 7 Step 8

item. This can be compared with the calibrations that are necessary before many chemical analyses. In the guidelines, there must be a description of all the attributes, how to use the scale and the use of the training samples. Step 5 relates to coding and consignment of the proficiency test items. If the laboratory that is preparing and sending the proficiency test items is also participating in the proficiency testing, it can code and send the samples to another laboratory; this laboratory then makes a new coding and the proficiency test items can be sent to all the participating laboratories. Step 6, the proficiency test, includes all the participating laboratories conduct the proficiency test according to the guidelines. Each laboratory makes a report including all details about the progress of the proficiency test and all the results. The coordinator collects all the reports in step 7 and checks that everything has been carried out according to the guidelines. Step 8 is data analysis and a report of the conclusion of the performance of the participating laboratories (see also Section 3.4).

3.2.1 Methods Most sensory panels work with more than one method and often with both descriptive tests and discriminative tests (ISO 3972, 1991, ISO 4121, 2003, ISO 6564, 1985, ISO 8587, 1988). It can be useful to choose a simple test such as basic taste of liquids or a ranking test. Here, levels of different concentrations can be tested, so determining at which concentration levels the sensory panel is reliable and robust. On the other hand, if the sensory panel most often works with sensory profiling tests, it is necessary also to use these kinds of test in a proficiency test. When using a profiling test, it can be discussed if the panel should develop its own vocabulary, or if the vocabulary should be defined before the test. If the vocabulary is defined before the test, then all sensory panels must use the same set of attributes (ISO 5492, 1992). The first can be an advantage for the sensory panels, but



41

will complicate the collection of data and the data analysis. Scales can also give the same problem; if, however, continuous scales are used by all, the problem is smaller compared to some using category scale and some line scale. Keeping the method simple and using standard methods is key here.

3.2.2 Material The material that is chosen must be manageable for it to be sent to the different participating laboratories without delay or damage; also preparation for the sensory evaluation must be kept simple. At each step in the preparation, thawing, mixing, diluting, heat treatment, etc., there are opportunities to introduce bias and then the samples prepared in one laboratory will not have exactly the same sensory characteristics as for the same sample prepared in another laboratory. In that respect, the guidelines for preparation of the samples must be clear and easy to understand. If real food items are chosen, it must be checked that the batch is from the same production day. If the participating laboratories are asked to buy the item themselves, it must be checked that content and sensory quality are exactly the same for all participating laboratories. The easiest way to overcome this problem is to decide that one laboratory is responsible for buying or producing the materials and sending it to all the participating laboratories. The proficiency test item must be assessed for sufficient homogeneity and stability. It must be ensured that the proficiency test item is unchanged during storage or transport. If it is essential to keep the temperature constant or within a fixed interval during transport, it can be necessary to make a dummy shipping before the real proficiency test item is sent to the participating laboratories. Assessment of homogeneity will depend on the material. It can be liquid or solid, but it also has to be taken into consideration that the material may be mixed or diluted before the proficiency test. One way of assessing the homogeneity and stability of the proficiency test item is to prepare a package with samples of material and make a trial shipment, ensuring the distance and transport are realistic. When the samples arrive, it can be ascertained whether the sensory quality is the same as when it was sent. Since the sensory profile of many food materials will change over time even at a constant temperature, the timescale for the proficiency test in the different laboratories must be very tight, and the ideal solution would be that all the sensory panels do the proficiency test on the same day. The amount of material must be large enough to ensure that all assessors will get the right amount of sample material. There must also be material for training of the sensory panel. If the proficiency item is within the range of food material that the sensory panel normally evaluates, the panel needs one training session before the proficiency test and in other cases two. The guidelines for the training must include an account of the training samples and a detailed description of the definition of the sensory attributes.


42


The test material can be in a concentrated form to make it easier to send/ transport. In that case the water for diluting must be specified; a solution can be to use bottle water from a specified spring. Water is not sensoryneutral and therefore it must be standardised as to the type/brand of bottled water to be used in the proficiency test. Some examples of design for proficiency testing are now given. In proficiency tests described by Tomic et al. (2009), the proficiency item was a type of candy, wine gums, produced specially for the proficiency test by a large candy company. The formulas for the wine gums were designed to give five different samples with variation (low, high and medium) in sugar and acid content. Sample 1 had high sugar and low acid, sample 2 had high sugar and high acid, sample 3 had medium sugar and low acid, sample 4 had low sugar and low acid, and sample 5 had low sugar, and high acid. All wine gums had the same content of raspberry flavour and intensity of red colour. The methods were sensory profiling with nine attributes: sugar coat, transparency, acidic flavour, raspberry flavour, sweet taste, biting (strength used in the first bite), hardness, elasticity and stickiness (to the teeth in the mouth). The scale was a 15 cm unstructured line scale with anchor points. Tap water was used to clean the palate between each sample. Each of the five samples was evaluated in three replicates, resulting in a total of 15 samples to be tested by each panel. The results were collected in a spreadsheet according to the guidelines, sent together with the samples. In this proficiency test 26 sensory panels participated, 10 from Norway, 1 from the United Kingdom, 2 from Sweden and 13 from Denmark. The panels were industrial sensory panels that usually perform quality control and panels from research institutes. Another example of proficiency test item and set-up is from a proficiency test of sensory profile panels by McEwan et al. (2002). Here the proficiency test items were six commercially available red wines deemed to have notable and distinguishable characteristics, selected by a large professional wine importer in Sweden. The samples, 750 ml bottles, were wrapped in aluminium foil before coding to keep the samples anonymous. Twelve sensory panels participated and they were asked to generate their own vocabulary (up to 30 attributes), but were required to cover odour, taste, mouthfeel, aftertaste and the four basic tastes as defined by ISO. The sensory panels used different kind of scale, most of them a continuous scale from 0 to 100, but one panel used a category scale from 0 to 9. McEwan et al. (2002) found that wine was a very difficult proficiency test item to work with, the main reasons being that if the panels were inexperienced in profiling wine, they needed more training than just one day, which is the normal time for training in connection with a proficiency test. McEwan et al. (2003) also did a proficiency test where the method was a ranking test. Here the proficiency test item was apple juice with five mixtures of sugars added (glucose and fructose), diluted with bottled natural mineral water. Each mixture comprised 50 ml of apple juice, 50 ml of water



43

and 6.5 g of one of the five sugar mixtures. In the proficiency test, 14 sensory panels participated from eight European countries (the UK, Ireland, Spain, Italy, Denmark, Norway, Sweden and Finland). The assessors were asked to rank the apple juice samples according to perceived sweetness intensity. The ranking method followed the ISO standard, but with the exception that the panels were allowed to use their normal procedure: either 1 = most and 5 = least or 1 = least and 5 = most. Before data were analysed, the rank order was converted to follow the ISO standard. A ring trial with hard cheese was described by Hunter and McEwan (1998). Twelve different varieties of hard cheese were selected and the criteria for selection were: bovine origin; internationally traded; having different sensory characteristics; and being available to the project. The method used was sensory profiling. Seven sensory laboratories from Denmark, France, Germany, Italy, Norway, the UK and Switzerland participated in two ring trials. In the first ring trial, laboratories used their normal methodology for quantitative descriptive profiling. After the first ring trial the participating laboratories agreed on one vocabulary, which was communicated to all partners in English, with a definition for every attribute. The vocabulary consisted of the following attributes: • for odours: animal, strength, acid, fruity, creamy, yoghurt, ammonia and hay/grass; • for tastes: salt, acid, bitter, strength, ammonia, animal, creamy, sweet, fruity, toasted and pungent; • for texture: rubbery/elastic, crumbly, grainy, hardness, melting and coating/adhesive. This vocabulary was then independently translated into the language of the sensory panel by the sensory scientist at each laboratory. In the second ring trial in this study, the laboratories in Denmark, Norway and the UK required only a short period of training (2 to 5 h) on the new vocabulary which was not very different from the vocabulary they used in the first ring trial. In contrast, the laboratories in France, Germany, Italy and Switzerland required 5 to 10 h of training since they were asked to make large changes to their normal procedures. These examples illustrate the importance of choosing methods and the proficiency test item. When sensory panels are going to use new methods, materials or attributes it is necessary to give the laboratories more time for training the sensory panels.

3.3 Panels If the proficiency test is designed for sensory panels evaluating only one kind of food material, as for example wine, then the obvious choice for the


44


proficiency test item is wine. On the other hand, if the panels participating in the proficiency test are panels evaluating either different kinds of food or only one kind of food, then the proficiency test item must be chosen in such a way that the test will validate the sensory panels analytical performances and not focus on how good the panel is at including new matrices.

3.4 Analysis of data/validation of results First of all, the coordinator must go through all reports from the participating laboratories to see if the guidelines have been followed and look for errors: for example, the panel might have turned around the scale or used the scale incorrectly. The next step is to collect all the results in one spreadsheet and check for errors in the raw data, for example 102 instead of 10.2. In proficiency tests for chemical laboratories the analysis of performance is often expressed in the standardised form of a z-score. A z-score relates the error in a result to the designated standard deviation of the results for the analysis in question. In sensory analysis, the panels not only assess one analyte, but in sensory profiling the intensity of many different attributes or the panel make a ranking order, and therefore the use of a z-score is not very suitable for assessing the panel performance. There will normally be some variation among assessors in one panel and there will likewise be variation between different sensory panels. For welltrained sensory panels, there can be differences between the assessors in how they are using the scale and also for the different attributes even if there is a definition for each attribute. If sensory profiling is used in a proficiency test this can give problems in cases where the performance of the sensory panels only is validated by the value of intensity for each attribute. It is necessary to look at the pattern to see how well the panel is discriminating between the samples. Therefore the sensory distances between the material used for the proficiency test item is very important. If the differences are too big, the task will be too easy for the panels. Conversely, if the differences are too small it will create an even bigger problem, especially if some of the sensory panels also use different scales. In this respect it is difficult to talk about a true value and it would be more useful to have an expected profile for the proficiency test item. The ranking between the different attributes in the different samples must be the same for all the sensory panels participating in the proficiency test. Given that these sensory profiles cannot be known in advance, unless previous data are available, the expected profile can be derived as a consensus from all the data provided in the proficiency test round. However, there must be a sufficient number of panels to have confidence in such a consensus. Furthermore, one panel cannot be allowed to distort the consensus. Moreover, in a proficiency test



45

round, the panels could possibly all be good or bad. Therefore, the consensus approach is easy, but it does not always offer the best solution. One way to overcome this is to make the samples by mixing different kinds of ingredient, such as sugars with known sweetness, rancid oil, salt or different spices in different concentrations. Of course one should be aware of how the mixtures influence the intensity of each others. Univariate as well as multivariate statistical methods can be used in the data analysis. The univariate methods focus on the differences for each attribute while the multivariate methods look at differences at a more general level, such as a pattern, taking into account correlations between the attributes. A sensory panel only rarely assesses one attribute and therefore the multivariate methods are much more suitable for comparing the performances of sensory panels. The data analysis for the proficiency test described by Tomic et al. (2009) where 26 sensory panels participated was done with PanelCheck (see more details in Section 3.2.2.). PanelCheck software is open source and may be downloaded, distributed and used for free. The program is designed to make assessor validation, but it can be used for panel validation as well. Of course other statistical programs can be used. In the study, data analyses were carried out first at a global level, based on data from all 26 panels where each panel was treated as if it was an ‘individual’ assessor. This means that the performance of panels was visualised by the different plots. As a result from this process Tomic et al. identified 3 of the 26 panels that needed further analysis at a more detailed level. The methods they used were mixed model ANOVA, Tucker-1 plot, Manhattan plot, one-way ANOVA based F plot, MSE plot, p*MSE plot, profile plot and line plot. They argued that the reason for using multiple plots and their methods was that each of the plots contains unique information on panel and assessor performance. Plots from such an analysis with performance information can also be used by panel leaders as feedback to improve panel performance and performance of individual assessors.

3.5 Panel performance The results from proficiency testing are one of several tools that enable laboratories to evaluate and demonstrate the reliability of the data produced. In addition to validation and accreditation, proficiency testing is an important demand of the European Union and is increasingly important in laboratory accreditation. Accreditation bodies will also ask laboratories for more detailed information of the panel performance. To give a full picture of the panel performance one has to take a longer-term view and not look at one test only. Owing to the special problems with true values in sensory analyses it is recommended that the sensory panel participate in proficiency tests that use different kinds of sensory methods.


46


When setting up the significant levels or the levels of acceptable performance, they must be chosen for each proficiency test item. They should reflect the panel performance and not how adaptable the panel is to new products/materials. McEwan et al. (2003) show how too tough criteria can cause only the top expert panels to perform well in the proficiency test they took part in. As a supplement in the validation of the panel performance, the panel leader can use reference materials that are within the sensory range of the sensory panel’s normal working area. Such references can be used in the daily routine of the sensory panel.

3.6 Glossary Coordinator

The person responsible for coordinating all activities involved in the operation of a proficiency test. Proficiency test The determination of laboratory testing performance by means of interlaboratory comparison. Proficiency test item The material used in the proficiency test. Proficiency test round A single complete sequence of circulation of proficiency test items, to all participating laboratories, in a proficiency test scheme. Proficiency testing scheme The system for objectively checking laboratory results by means of an external agency. Reference material Material or a substance with one or more properties sufficiently homogeneous and well established to be used for calibration and/or training. True value The actual concentration of the analyte and for sensory analysis the intensity of the attributes in the matrix. Ring trial A single complete sequence of circulation of proficiency test items to all participants in a proficiency test scheme.

3.7 References and further reading astm E1301-95 Standard guide for proficiency testing by interlaboratory comparisons. astm STP758 Guidelines for the selection and training of sensory panel members. ea-4/09 (2003) Accreditation for sensory testing laboratories. en iso/iec 17025 General requirements for the competence of testing and calibration laboratories (2005).



47

hunter, e.a. and mcewan, j.a. (1998). Evaluation of an international ring trial for sensory profiling of hard cheese. Food Quality and Preference, Vol. 9 (5), 343–354. ilac-g13:2000 Guidelines for the requirements for the competence of providers of proficiency testing schemes. iso 17025 (1999) General requirements for the competence of calibration and testing laboratories. iso 3534 (1993) Statistics – Vocabulary and symbols. Part 1: Probability and general statistical terms. iso 3972 (1991) Sensory analysis – Method of investigating sensitivity of taste. iso 4121 (2003) Sensory analysis – Guidelines for the use of quantitative response scales. iso 5492 (1992) Sensory analysis – Vocabulary. iso 5496 (1992) Sensory analysis – Initiation and training of assessors in the detection and recognition of odours. iso 5725-1 (1994) Accuracy trueness and precision of measurement methods and results. Part 1: General principles and definitions. iso 6564 (1985) Sensory analysis – Flavour Profile Methods. iso 6658 (1985) Sensory analysis – General guidance. iso 8586-1 (1993) Sensory analysis – General guidance for selection, training and monitoring of assessors – Part 1: Selected assessors. iso 8586-2 (1994) Sensory analysis – General guidance for selection, training and monitoring of assessors – Part 2: Experts. iso 8587 (1988) Sensory analysis – Ranking. iso 8589 (1988) Sensory analysis – General guidance for the design of test rooms. iso guide 30 (1993) Terms and definitions used in connection with reference materials. In: International vocabulary for Basic and General Terms in Metrology, 2nd Edition, ISO, Geneva. iso guide 34 (2000) Reference materials – General requirements for the competence of reference material producers. iso guide 35 (2006) Reference materials – General and statistical principles for certification. iso/cd 13300 – 1 (2002) Sensory analysis – General guidance for the staff of a sensory evaluation laboratory – Part 1: Staff responsibilities. iso/cd 13300 – 1 (2002) Sensory analysis – General guidance for the staff of a sensory evaluation laboratory – Part 2: Recruitment and training of panel leaders. iso/iec 17025 (2005) General requirements for the competence of calibration and testing laboratories. iso/iec guide 43-1 (1997) Proficiency testing by interlaboratory comparisons – Part 1: Development and operation of proficiency testing schemes. iso/iec guide 43-2 (1997) Proficiency testing by interlaboratory comparisons – Part 2: Selection and use of proficiency testing schemes by laboratory accreditation bodies. mcewan j.a., hunter, e.a., van gemert, l.j. and lea, p. (2002) Proficiency testing for sensory profile panels: measuring panel performance. Food Quality and Preference, Vol. 13, 181–190. mcewan j.a., heiniö, r.-l., hunter, e.a. and lea, p. (2003) Proficiency testing for sensory ranking panels: measuring panel performance. Food Quality and Preference, Vol. 14, 247–256. nmkl procedure 6 (1998) General guidelines for quality assurance of sensory laboratories. nmkl procedure 14 (2004) SENSVAL: Guidelines for internal control in sensory analysis laboratories. nmkl procedure 16 (2005) Sensory quality control.


48


nmkl procedure 20 (2007) Evaluation of results from qualitative methods. panelcheck software. www.matforsk.no/web/sampro.nsf/webTemaPE/ PanelCheck!OpenDocument thompson, m. and wood, r (1993) The international harmonised protocol for the proficiency testing of (chemical) analytical laboratories. Pure and Applied Chemistry, Vol. 65, 2123–2144. tomic, o., luciano, g., nilsen, a.n., hyldig, g., lorensen, k. and næs, t (2009) Analysing sensory panel performance in a proficiency test using the PanelCheck software. Accepted by European Food Research and Technology, Vol. 230, No. 3, 497–511.


4 Sensory methods for quality control L. L. Rogers, Consultant, UK

Abstract: The chapter includes a reminder about the importance of choosing the right method, agreeing how the results will be used and the impact of choosing the wrong test for the objective. An introduction to the use of action standards is included. The chapter gives an overview of a large number of tests, giving advice on which are the best and most popular in the world of quality measurements. Each test section has: an introduction to the test including popularity, advantages and disadvantages, example uses; samples, including what format and how many, how much is required, balanced designs; panellists, number and level of training; example test set-up, for example, the ballot paper/test sheet; and references for further reading and information. Key words: sensory science, action standards, quality methods.

4.1 Introduction This chapter is concerned with the sensory methods available for quality control in the food and beverage industry. Several methods are described and discussed, including their popularity and applicability to quality control situations. Where methods are highly aligned to quality control, further detail is given in each method section: • introduction (including popularity, advantages and disadvantages, example uses); • samples (quantity, type, balanced designs); • panellists (number and level of training); • example test set-up (e.g. questionnaire/ballot paper/test sheet); • references for additional information. Many industries are still not using sensory science to its full capabilities in quality control. There seems to be a continued reliance on experts to judge sensory quality and in some industries, where they have moved away from experts, they have moved to the use of small numbers of panellists


52


making subjective rather than objective measurements. In the future it would be beneficial if the food industry paid more attention to the sensory attributes of their products and how the changes in these attributes can affect consumer liking. This change would see industries using recommended objective sensory methods, linked to statistical process control and agreed action standards. Consumer-led quality methods mean that the identification of the sensory characteristics most important to the consumers’ view of ‘quality’ is the way forward. Consumers will not continue to buy a product if it does not meet their expectations – and they may not complain: in today’s busy environment they are more likely never to buy the product again. These consumer-led quality methods could also be linked to instrumental measurements, to aid production facilities in making excellent products time and time again. It can be daunting when starting out in your sensory science career (and even later!) to see the huge number of sensory methods available. How can the right method be selected to meet the test objective? The objective of the test and the manner in which the results will be used are key to the selection of the correct test method. There may also be constraints due to facilities which mean that some methods will not always be available. Obviously this needs to be taken into account when deciding which method to use. One of the most important aspects to consider before selecting which method to use is to be clear of the objective of the sensory study. This will involve finding out different pieces of information to determine exactly why the sensory test is required so that the test can be designed to meet the objective. For example there is little point conducting a full sensory profile if the client only wishes to know if there is a texture change because of the introduction of a new ingredient supplier for a thickening agent. Another important consideration before deciding which method to use is how the results will be collected, analysed and subsequently used. A useful technique is the use of action standards as these can be incredibly helpful in designing the test. An action standard (AS) defines the aim of the overall experiment but also states the action (or next steps) to be taken dependent upon the results. In the example above, where a new supplier was under discussion for the thickening agent, the action standard might have read: AS1: ‘If the sensory test confirms that there is a textural difference between the new supplier and our existing supplier for product X, we will not proceed with the new supplier.’

The objective is clear: the client wishes to know if there is a difference in texture, but also will be rejecting the new supplier if there is a difference in texture. In this case a simple difference test could have been selected to determine if there was a texture difference. The action standard below might have resulted in a totally different sensory approach:


Sensory methods for quality control

53

AS2: ‘If the sensory test confirms that there is a textural difference between the new supplier and our existing supplier for product X, we will need to understand what the textural difference is and if there are any other changes to product X as a result of the supplier change. Our existing supplier will stop producing at the end of December so it is critical we find a new supplier.’

The sensory scientist may well have chosen to ask the client how likely they thought a difference might be between the new supplier’s ingredient and the existing supplier. If it was likely that there would be a large difference they may have decided to go directly to conducting a full profile to understand what the differences were. These differences would be critical in understanding the effect of this change on consumers’ reactions to product X. If the change threatened key drivers of liking, further discussions with the new supplier may be required to achieve a match. AS3: ‘If the sensory test confirms that there is a textural difference between the new supplier and our existing supplier for product X, we will need to understand what the textural difference is and if there are any other changes to product X as a result of the supplier change. We also need to understand how this difference is related to the natural variance in our product.’

In this case the sensory scientist may well decide to carry out some batchto-batch variability tests and determine where the new ingredient batch fits. The difference from control method might be a useful starting point to gather data for this test as several batches may be included in one test. AS4: ‘Is there a textural difference between the new supplier and our existing supplier for product X?’

The example above (AS4) is an example of a poor action standard. The next stage after understanding if there was a difference or not is not documented and therefore the choice of test is a difficult one. It is more likely that the wrong test would be chosen and then, when the results are reported, the client would be questioning what the next steps should be: the sensory scientist will probably not have the information at hand to guide the client. These examples indicate how knowing the objective and knowing what the next steps will be as an outcome of the sensory study, are vital in the choice of sensory test. Some of the method categories below are not suitable for day-to-day quality assessments but can be used in the set up of a QC programme, so have been included in this discussion (Costell, 2002). In each case it is not the test method alone which will result in the desired outcome, but also the manner in which the test is conducted and how the results are analysed and used (Costell, 2002). Table 4.1 gives an overview of all the tests and the relative complexity of each. Table 4.2 gives the number of panellists and the level of training and experience they might require for each method.



Method

Descriptive specification ‘In/out’ (or pass/fail) Difference from control A not A Paired comparison (e.g. 2AFC) Scaling (including targeted scaling) Ranking Triangle test Quality scoring/grading/rating Magnitude estimation Duo–trio In-house methods DIY

4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 4.10 4.11 4.11 4.12 4.12

Overview of sensory methods

Section

Table 4.1

High High High Medium Medium Medium Medium Low High Low Low High High

QC relevance Medium Low Medium Low to medium Low Low to high Low Low Low Low Low Variable, generally low Variable

Time to conduct test

High Medium Low to medium Medium Low High Low Low Medium to high Medium Low Variable, generally low Variable

Time to set up methodology

High Low Medium Low Low Medium to high Moderate Low Medium to high Low to medium Low Moderate Variable

Level of detail gained from results


55

Table 4.2 Recommended number of panellists Recommended number of panellists (highly trained panellists)

Panellist training and experience

(10) 25 (10) 30 (18) 20 (10) 30 (20)

High Medium Low to medium Medium Low

Variable

High

30 (5) 24 (18) 8–12 (5)

Low Low Medium to high

4.11 4.11 4.12

Descriptive specification ‘In/out’ (or pass/fail) Difference from control A not A Paired comparison (e.g. 2AFC) Scaling (including targeted scaling) Ranking Triangle test* Quality scoring/grading/ rating Magnitude estimation Duo–trio In-house methods

Variable 32 (15) Variable

4.12

DIY

Variable

Medium Low Variable generally low Variable

Section

Method

4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 4.10

* See ISO 4120:2004 for more details on number of panellists.

4.2 Descriptive specifications (DS) method The descriptive specifications (DS) method is also known as the comprehensive descriptive method (Muñoz et al., 1992) and descriptive analysis method (Lawless and Heymann, 1999). The basis for this popular method lies in the development of sensory specifications for finished products. A sensory specification is similar to other specifications and is a vital part of ensuring product quality. Specifications detail exactly what the product should look like, smell like and taste like and can easily be extended to texture measurements where necessary. An example semi-quantitative sensory specification is given in Fig. 4.1 and a fully quantitative example is given in Fig. 4.2. The sensory specification is built around those attributes which are known to contribute to consumer acceptance of the product and it is this aspect of the method which requires large resources during its conception. The method is very objective as it does not require the panellists to make any subjective judgements on the product’s quality as such. This judgement is made by the sensory scientist in the interpretation of the data. The method gives very actionable results which can be correlated to both instrumental and consumer measurements. A well-trained sensory panel of around 10 screened panellists is required to measure the levels of a selection of attributes, generally for finished


56


Date: Product:

Fruit drink 1

Pack type and size:

380 ml PET bottle

The product must be free from off-odours, taints and foreign particles Intensity scale:

None Slight Moderate Strong Very strong

Appearance

Evaluated by looking at the product in a clear sampling cup under artificial daylight before and after swirling. A very strong orange coloured liquid that is bright, clear and still. The liquid appears thin and does not leave a residue upon swirling.

Aroma

Evaluated by smelling from the sampling cup before and after swirling. A moderate lemon and moderate orange aroma that is moderately sweet. A slight aroma of vitamin C tablets and slightly floral.

Flavour

Evaluated by taking sips from the sampling cup. A moderate lemon, moderate orange flavour that is also moderately acidic and moderately sweet. A slight flavour of vitamin C tablets and also slightly bitter in flavour.

Mouth-feel/Texture

Evaluated at the same time as the flavour and by taking more sips from the sampling cup. Slightly drying mouth-feel. The liquid feels slightly thicker than water in the mouth.

Aftertaste/Afterfeel

Evaluated after swallowing – no extra sips taken. A moderate citrus aftertaste that is also moderately acidic and moderately bitter. Moderately drying and a moderate teeth/mouth-coating afterfeel. A slightly sweet aftertaste.

Glossary of terms example Flavour/aftertaste Orange A bitter orange flavour. Like the flavour of Seville oranges. Lemon A sour lemon flavour. Like fresh lemons Acidic The basic taste of citric acid solution. Sweet The basic taste of sucrose solution. Vitamin C tablets The flavour of vitamin C tablets. Like Haliborange. Bitter The basic taste of caffeine solution. Citrus The tangy flavour of general citrus fruits such as lemon, lime and orange.

Fig. 4.1 Example specification for a fruit drink.

products. Samples are usually taken from daily production batches. The quantity required is based on each panellist making one assessment of all the modalities under consideration in the specification. This can be done with a fairly simple paper ballot or using standard sensory software. The method can be semi-quantitative (see example questionnaire given in



57

Product 7, batch 123 Example data and specification (used for decision making and not seen by panellists) Attribute Appearance: colour intensity Appearance: brightness Aroma: lemon Aroma: orange Aroma: vitamin C tablet Flavour: lemon Flavour: orange Flavour: vitamin C tablet Flavour: acidic Flavour: sweet

Results for sample 123

Sensory specification

5.2 7.0 3.0 4.3 1.5 3.0 6.1 1.0 3.0 8.2

4.5–6.0 6.5–9.5 2.0–4.0 4.0–6.0 0–1.5 2.0–4.0 5.0–7.0 0–1.5 2.5–4.5 7.5–8.5

Fig. 4.2 Example data and specification for the fully quantitative descriptive/ specifications method.

Fig. 4.3) or fully quantitative by adding line scales for each attribute measured (an example questionnaire is given in Fig. 4.4). In the fully quantitative method, if the attribute measurements are out of specification then the product is deemed unacceptable by the sensory scientist. The panellists are not aware of the attribute intensity levels built into the specification and are therefore not making a judgement on product quality. They are acting as an instrument and the data they produce is used to help decide if the product meets specification. In the semi-quantitative method the panellist is checking if each attribute is present at the correct level. The information from each panellist is passed to the sensory scientist to decide if the product meets specification. The semi-quantitative method is useful for line-side assessments and can be particularly useful for checking preliminary products. The fully quantitative method is recommended for final product evaluation. The sensory specification can be set by management or by the additional use of consumer data. The setting of sensory specifications with input from consumer data is incredibly useful as it gives information about the attributes that drive product liking (and also disliking) but also gives information about the tolerances consumers have to changes in the product. This allows the quality team to make recommendations to management based on these tolerances, rather than rejecting products that may have been acceptable or not rejecting products that were unacceptable. For full details on setting up consumer-led specifications please see Muñoz et al. (1992). An additional dimension of this technique is that it lends itself very nicely to the use of statistical process control (SPC) (Oakland, 2007). This allows production to be monitored over time and drifts in quality highlighted before the product goes out of specification. An example of this is given in Fig. 4.5. For more details on the use of SPC please see Oakland (2007).


58


1. Refer to the sensory specification documentation before completing this assessment 2. Tick the box if the descriptor is present 3. Product must be assessed as per make-up instructions Appearance very strong orange coloured liquid bright clear still thin (like water) not leave a residue upon swirling Aroma moderate lemon moderate orange slight vitamin C tablet

Aftertaste moderate citrus moderately acidic slightly sweet moderately drying If there are any additional descriptors or if any descriptors listed here are not present, please consult your manager ................................................................... ...................................................................

Flavour moderate lemon moderate orange moderately acidic moderately sweet slight vitamin C tablet slight bitter

................................................................... ................................................................... ...................................................................

Texture in the mouth slightly drying feels slightly thicker than water in the mouth

Fig. 4.3 Example of the ballot paper for the semi-quantitative descriptive/specifications method. Product 7, batch 123 Instructions: Please rate each of the attributes below according to the standard protocol for assessment, and with reference to the attribute definitions list and intensity training programme. Appearance Colour intensity

|

|

0 Brightness

10

|

|

0

10

Aroma Lemon

|

|

0

10

Fig. 4.4 Example of the ballot paper for the fully quantitative descriptive/specifications method.



59

Orange

|

|

0

10

Vitamin C tablet

|

|

0

10

Flavour Lemon

|

|

0

10

Orange

|

|

0

10

Vitamin C tablet

|

|

0

10

Acidic

|

|

0

10

Sweet

|

|

0

10

The attribute list would continue with additional flavour, texture and aftertaste attributes Example definitions Lemon: the fresh lemon aroma/flavour as found in freshly peeled lemon segments Orange: the orange juice aroma/flavour as found in freshly peeled Jaffa orange segments Natural sweetness: basic taste of a sucrose solution Acidic: basic taste of a citric acid solution Example data and specification (used for decision making and not seen by panellists) Attribute

Results for sample 124

Appearance: colour intensity Appearance: brightness Aroma: lemon Aroma: orange Aroma: vitamin C tablet Flavour: lemon Flavour: orange Flavour: vitamin C tablet Flavour: acidic Flavour: sweet

5.2 7.0 3.0 4.3 1.5 3.0 6.1 1.0 3.0 8.2

Fig. 4.4 Continued


Sensory specification 4.5–6.0 6.5–9.5 2.0–4.0 4.0–6.0 0–1.5 2.0–4.0 5.0–7.0 0–1.5 2.5–4.5 7.5–8.5

60

Sensory analysis for food and beverage quality control 8

7

Intensity of raspberry flavour

6

5

4

Attribute intensity Upper control limit Lower control limit

3

2

1

0

1

2

3

4

5

6

7

8

9

10 11 12 13 14 15

Batch number

Fig. 4.5 Statistical process control (SPC) example.

4.3 ‘In/out’ (or pass/fail) method The ‘in/out’ method is widely used in quality assurance (QA) and quality control (QC) sensory tests due to its ease of setting up and its simplicity in analysis (Muñoz et al., 1992). It can be used for a wide variety of purposes: raw materials, interim products and finished products. A trained panel assesses whether each sample type is ‘in’ or ‘out’ of specification. An example ballot paper is given in Fig. 4.6. The specifications must be documented to limit personal subjectivity. The method differs from the previous DS method in one main factor: the panellists actually make the decision on whether or not a sample is suitable or not. This is one of its main disadvantages as this decision can be quite subjective in nature and can cause problems – especially when wrongly linked to production bonuses, for example. Another main disadvantage is the lack of information provided about the reason for the product failure although this may be built into the method where necessary. Although the method is simple there are many industrial examples where its use could be improved. In some production facilities only one, or sometimes up to five people, take part in these types of assessments. These are informal and conducted verbally based on each person’s experience of



61

Instructions Please evaluate the products below in the order shown. Evaluate each sample individually and mark whether it is ‘in’ or ‘out’ of specification in the box provided. Please use the specifications provided to help in your decision making. Sample 871 902 376 299

In

Out

Instructions Please evaluate the products below in the order shown. Evaluate each sample individually and mark whether it is ‘in’ or ‘out’ of specification in the box provided. Please use the specifications provided to help in your decision making. Where you have marked the product ‘out’ of specification please comment why you have made your choice Sample 871 902 376 299

In

Out

Comment .............................. .............................. .............................. ..............................

Fig. 4.6 Examples of the ballot papers for the ‘in/out’ or pass/fail method.

production quality and deviations. Problems can occur when these panellists do not agree on whether a product is in or out of specification (probably the reason why in some companies this job falls to one person) and can tend to lead to people making decisions based on their own personal preferences – not a recommended situation. Relatively easy additions and changes can be made to vastly improve the results. Firstly by documenting the production specifications, the deviations and personal preferences can be kept to a minimum (Carpenter et al., 2000). This is usually conducted by evaluation of a large number of samples and then determining important attributes: those which vary and those important for the product characteristics, and the limits of each of these attributes for successful production. The use of around 25 or more panellists will also improve the data collected by this method. Panel training, particularly in the form of examples of products both in and out of specification, can hugely increase the analytical nature of this method. Documentation of sample preparation and consistent serving methods can also go along way to improving the use of this method. As for many QA and QC methods, the use of action standards vastly improves the decision-making process. Generally the percentage of panellists rating each batch in or out is used for decision making.


62


Obviously in an ideal situation the panellists would all rate an individual sample in the same manner, for example 100% of panellists would rate the batch as ‘in’ specification. However, in practice this does not happen and therefore action standards often take the form of ‘60% or more panellists accept the batch’ and therefore the batch is deemed to be in specification, or ‘40% or less accept the batch’ and the batch is deemed to be out of specification. Any batches falling in the area between 40 and 60% would be sent for further analysis to determine next steps (Muñoz et al., 1992). The use of panel monitoring techniques is critical for this type of method and can be easily implemented by the use of ‘hidden control’ products – usually kept from previous rejected batches for this purpose. Further additions, perhaps more complicated and expensive, can be made by the addition of consumer information. However, specifications for this method are generally prepared by management, although the addition of any consumer data where available would be beneficial.

4.4 Difference from control (DFC) method The difference from control (DFC) method is another popular method in daily quality assessments (Muñoz et al., 1992; Lawless and Heymann, 1999) and can also be used for the assessment of batches on a regular basis for ambient products. However, one of the disadvantages is the need for a ‘control’ product. For some food products this is less of an issue as control products can be stored effectively for several months at a time and a new control selected at regular intervals. Where a control cannot be stored the control for each test must be representative of standard production. One solution for this can be to use the descriptive specification method (DS method – see Section 4.2 above) to select the control product and then use this control to assess several batches over the usability period of the control using the DFC method. As the panellists for this test do not need to be as highly trained as those for the DS method, the DFC and the DS methods can be used in conjunction for resource and time saving in a production environment. The test is fairly straightforward to set up and screened panellists require only a short training period to get used to the test: but it is recommended that panellists are monitored in each test by the use of a hidden control (see below). Data analysis can be difficult where statistical significance is employed; however, many companies rely on the mean scores alone for each judgement. The test can be run by a technician but is best analysed with input from a sensory scientist where statistical analysis is required. Although a greater quantity of the control is required there only needs to be enough sample of each batch for the panellists to assess once. If statistical tests such as analysis of variance are to be conducted, the number of panellists required is around 18. However an understanding of batch conformity



63

can be gathered with around ten or more highly trained panellists if the data are validated by the measurements for the hidden control. Each panellist is presented with the control product and several other batches coded with three-digit numbers. The number of batches that can be assessed in each test will depend on the product type, but if there is little carry-over or aftertaste issues, up to five different batches could be assessed in one test. The panellist assesses the control product first and is then asked to determine how different each individual sample compares with the control on a scale: generally 0 to 10 where 0 is no difference and 10 is an extreme difference. See Fig. 4.7 for an example questionnaire. The batches themselves are not compared – only compared with the control – therefore making it very resource friendly as there are two to five comparisons to the control in just one test. The DFC method relies on the use of ‘hidden’ control to prove its validity. This hidden control is another portion of the control batch but instead of being identified as such, it is hidden along with the various batches by a three-digit code. The hidden control should be rated by the panellists as being the same as the control with a score of 0 or 1, or perhaps 2 depending on the production variability between batches and between controls. If the hidden control is rated outside these limits the test results are rejected. The panellists must never be aware of the existence of the hidden control nor must its correct identification be used as panel monitoring in a feedback situation. The reason for this is that the panellists

Instructions Assess the sample marked control first. Assess the first sample marked with the 3-digit code. Assess the overall sensory differences between the two samples using the scale below – mark the scale to indicate the size of the overall difference. Difference Scale Code: 123

No 0

2

4

Mod

Difference Scale Code: 456

No 0

2

4

Difference Scale Code: 789

No 0

2

4

6

8

Extreme 10

6

8

Extreme 10

6

8

Extreme 10

Mod

Mod

Notes For a computerised system, each sample would be presented on a different ‘page’ or screen. For each coded sample a question about the difference may also be asked and this can also be presented in the form of differences about all modalities if desired.

Fig. 4.7 Example of the ballot paper for the difference from control method.


64


will be focused on identifying the hidden control rather than identifying any differences between batches: not the objective of the test at all. This method can be used for production where there is inherent product variability due to its components, in preparation or in serving: for example baked and snack products. The use of standard difference tests would result in significant differences where in fact the difference is just due to the product’s inherent variability. Where the control batches can also be variable, the method can be adapted (Pecore et al., 2006) and two controls are presented. For this example there are four pairs for each panellist: Control 1 versus Control 1 (also called the hidden control), Control 1 versus a second control, control 1 versus the test batch and the second control versus the test batch. The first three comparisons are simply part of the original difference from control test and it is only the fourth comparison that makes up the control variability test. This allows for the test batch to be within the controls’ batch variability and easily detects where the test batch is outside the control batches variability. A further adaptation of this method also considers the variability of the test product, for example in ingredient substitution, by the introduction of a balanced design to eliminate panellists’ fatigue and the need for more sampling (Young et al., 2008). The DFC can also be used as a targeted DFC (TDFC). This is particularly useful if there is prior knowledge about the attributes or modalities that change within production batches. For example if differences are generally seen in sweetness level, then a TDFC can be employed to determine the difference in sweetness between a range of batches. See Fig. 4.8 for an example questionnaire.

Instructions Assess the sample marked control first. Assess the first sample marked with the 3-digit code. Assess the sensory difference in sweetness between the two samples using the scale below – mark the scale to indicate the size of the difference in sweetness. Difference in sweetness No Scale 0 Code: 123

Mod 2

4

6

8

Extreme 10

Difference in sweetness No Mod Extreme Scale 0 2 4 6 8 10 Code: 456 Notes For a computerised system, each sample would be presented on a different ‘page’ or screen.

Fig. 4.8

Example of the ballot paper for the targeted difference from control method.



65

4.5 ‘A’ not ‘A’ method The ‘A’ not ‘A’ method is another popular method for QC and QA specialists because it is simple to train panellists for, conduct and analyse. It is particularly useful for production facilities where a small number of different products are made on a regular basis, as panellists become very familiar with these products and hence very familiar with ‘A’. It is also useful where the two samples cannot be exactly the same in appearance (obviously where this modality is not essential to the product quality) but the differences are subtle and only obvious if the two samples were presented together (Lawless and Heymann, 1999). This method is only useful where the inherent variability is low, otherwise it results in the rejection of too many production batches (Muñoz et al., 1992). The method only really gives the answer that the batch is different but does not give information as to the degree of difference or in what format the difference takes – for example, that the batch is sweeter or with a higher flavour intensity. However, the test is very simple to set up and train panellists and easy to analyse the results. The panellists are presented with a sample labelled ‘A’ to familiarise themselves with and then presented with a series of three digit coded samples, some of which are A and some of which are test batches. In training the panellists must become familiar with A prior to taking part in the tests so that the ‘sensory profile’ of A is familiar to them and the initial assessment during the tests just serves as a reminder (BSI, 1988). The presentation to each panellist should be random and different for each assessor. See Fig. 4.9 for an example questionnaire. The number of panellists required depends upon the test objective and the required significance level, but around 20 panellists would be used for

Instructions Assess the sample marked ‘A’ first, then pass back to the test coordinator. The coded samples consist of ‘A’ and ‘not A’ in a random order. All the ‘not A’ samples are identical. The respective number of each of the two kinds of samples is unknown to you. Assess the coded samples one by one and complete the form below Sample code

The sample is ‘A’

123 456 789 234 678

Fig. 4.9

Example of the ballot paper for the ‘A’ not ‘A’ method.


‘Not A’

66


a typical situation and each would assess five ‘A’ and five ‘not A’ samples. The results can be tabulated to indicate the number of panellists identifying ‘A’ and ‘not A’ correctly and the ratios are then analysed using χ2 (BSI, 1988).

4.6 Paired comparison methods (e.g. 2AFC, n-AFC, simple difference test) The paired comparison test is another simple method to set up, train and analyse but not always ideal for quality environments (Muñoz et al., 1992) as it can be very sensitive to small differences. The test can determine if two samples are different in a particular attribute (directional paired comparison or 2-alternative forced choice (2-AFC) method) or simply if the samples are different (simple difference test). For example in a test with biscuits, the sensory scientist may know that they differ in texture and therefore the panellists would be asked which biscuit is softer in texture: this would be a 2-AFC method (Lawless and Heymann, 1999). The simple difference test has limited usefulness as generally the triangle test or duo–trio are more suitable. However, it can be useful where there is limited sample quantity or where the presentation of three samples is not possible, for example for chewing gums or certain curried products.

4.6.1 2-AFC method There are two presentation orders (AB and BA) and the test is designed so that both orders are presented an equal number of times. The samples are both presented at the same time and the panellist is asked to identify the sample which is higher in the specified attribute. Figure 4.10 gives an example questionnaire for this test. The panellists must understand the attribute under consideration to be able to judge the difference effectively. The results give an indication of the direction of difference between the two samples; however, if the difference in one attribute affects several other attributes (for example the sugar level in biscuits can affect the sweetness and hardness) then this would not be the test of choice.

4.6.2 Simple difference test The samples are again both presented at the same time. Little training is needed as people generally find it easy to decide if the samples are the same or different. Around 20 to 50 presentations are required (Meilgaard et al., 1999) but each panellist assesses each sample only once. Therefore around 60 panellists for a QC application are sensible. There are four presentation orders in this example (AA, BB, AB, BA) and these will be randomised across panellists with an equal number of each order presented. Figure 4.10



67

2-Alternative forced choice method Instructions There are two samples for you to assess: please assess them in the order shown below. Please assess the force required to bite off half the biscuit. Please ring the code of the hardest biscuit. 123

456

Thank you for taking part.

Simple difference test method Instructions There are two samples for you to assess: please assess them in the order shown below. Please bite off half of each biscuit. Are the two samples the same or different? Please tick the relevant box. Pair 123 and 456

SAME

DIFFERENT


Fig. 4.10 Example of the ballot paper for the paired comparison methods.

gives an example questionnaire for this test. The results can only tell the sensory scientist if the samples are different or not – no reason for difference or intensity of difference can be obtained from this method.

4.7 Scaling method (including targeted scaling) Scaling can be useful where a quality unit is closely assigned to a Research and Development (R&D) unit and therefore has use of the R&D quantitative descriptive profiling panel. The method can be used with a quality panel if the time is available for training and panel monitoring. The method is closely linked to quantitative profiling but generally the sensory scientist selects only the attributes that are known to change in production for daily monitoring of the key sensory characteristics. The method is not very popular due to the amount of panel training required when there is no access to an R&D panel, and the statistical data analysis element requires time and statistical training for the test coordinator. It can be useful though for the measurement of particular attributes that are known to change but yet are key to consumers’ liking of the product. This is


68


particularly helpful where the product’s characteristics may be changed by blending or reworking as there is quantitative data to help with these adjustments. Screened and trained panellists rate the specific attributes on line scales for each sample. Samples are coded with three digit numbers and presented monadically to each panellist in a randomised balanced design. The rating can be performed using paper questionnaires but where sensory data collection systems are in use this method is much easier to gather and analyse data. An example questionnaire is given in Fig. 4.11. The panellist assesses each sample and marks on the line scale the intensity of the given attribute. For panellists that are not part of a quantitative panel it can be useful to give a warm-up sample with a given intensity and a further known-intensity sample can be used for panel monitoring. A useful adaptation of this method is the scaling of only one attribute. This is known as targeted scaling. The method can be especially useful where an R&D panel is not available as QC panellists can become very skilled in the understanding of the attribute and its scaling.

Instructions: Please rate each of the attributes below according to the standard protocol for assessment Appearance Colour intensity

|

|

0

100

Brightness

|

|

0

100

Aroma Lemon (the fresh lemon aroma/flavour as found in freshly peeled lemon segments)

|

|

0

100

Orange (the orange juice aroma/flavour as found in freshly peeled Jaffa orange segments)

|

|

0

100

Vitamin C tablet (the aged orange aroma as found in Co-Op brand Vitamin C tablets)

|

|

0

100

Fig. 4.11 Example of the ballot paper for the scaling method.



69

4.8 Ranking test The ranking test is very simple to set up, panellists need little training and the analysis is simple, but in terms of quality assessments it is not especially popular. This is due to the manner in which the sensory question is asked in this test. Panellists are asked to put the samples in order (rank) of some attribute (BSI, 1989). For example they may be asked to rank the samples in order of sweetness or creaminess. As potential quality issues may be linked to several attributes (and not just creaminess alone, for example) this can limit the usefulness of this test. However, if the production variability is known, and that generally creaminess variability is the main issue, then this method can be very useful. Screened and trained panellists are presented with four or five samples in a random order and asked to rank them in order of the specified attribute or give the samples equal rank. See Fig. 4.12 for an example questionnaire. The use of data collection software can be very useful here as the panellists can simply drag and drop each sample in rank order or drop equal ranked samples into the same ‘box’. For quality methods one of the samples must be the verified control and it can be useful to have a hidden control on certain occasions. The hidden sample can also be useful to check panel performance. Please see notes on the use of a hidden control in Section 4.4 above. The minimum number of panellists required for this test is five (BSI, 1989), but this is not recommended as the more panellists you have the

Instructions Assess the samples in the order shown below. Note the intensity of the creaminess of each sample. Write ‘1’ in the box of the sample which is the least creamy. Write ‘2’ for the next creamy and ‘3’ for the next and ‘4’ for the most creamy. If two samples appear the same please rank them with the same number. Handy tip: as you assess each sample, place it in front of you in the order of creaminess – this makes it easier to fill in the boxes below when you have finalised your decision. Sample code

Rank Order

123 456 789 234 Thank you for taking part.

Fig. 4.12

Example of the ballot paper for the ranking method.


70


greater the quality of the data. Carpenter et al. (2000), recommend a minimum of 30 panellists and analysis of the data with the Friedman rank test (O’Mahony, 1986).

4.9 Triangle test Muñoz et al. (1992) state that the triangle test is not an ideal method in QC as it can be very sensitive to small differences and can therefore create too many false positives. However, where a product can withstand very little difference, and consumers require low variability, it can be useful. Screened and trained panellists are presented with three coded samples. They are told that two of the samples are the same and one is different and then asked to identify the ‘odd’ sample. There are six possible presentation orders (AAB, ABA, BAA, BBA, BAB, ABB) and therefore it is recommended to conduct this test with groups of six assessors (6, 12, 18 . . .) to include all the presentation orders in each case. The ISO standard (ISO4120:2004) for triangle tests gives a very detailed overview of the number of panellists required dependent upon the objective of the test. An example questionnaire is given in Fig. 4.13. An additional question may Instructions There are three samples for you to assess. Two of the samples are the same and one is different. Please assess them in the order shown below. Please ring the code of the odd sample. 123

456

789


Fig. 4.13

Example of the ballot paper for the triangle test method.

be added to the questionnaire to gain information about the nature of the difference. It is not recommended to give direct feedback on whether the panellist detected the ‘correct’ odd sample, as in the situation where samples are generally not different (as decided by the whole panel), this can lead to the panellists feeling that they always fail in this test situation.

4.10 Quality scoring/grading/rating method The quality scoring method is a common method for quality control and is often developed for the company’s specific product(s). The training and



71

experience levels of the panellists taking part in these tests are generally high as the method relies in part on the panellist’s memory of the ideal product (often seen in scales of ‘typicality’) and also the panellist generally needs experience of the day-to-day issues and changes occurring within the product. Another factor in the panellist’s training requirements with this method is the fact that they will be making the decision about the product quality: usually by scoring, grading or rating the effect of the changing attributes on the end quality of the product. This type of method can often be seen in use for commodity products such as milk and fish and are often supported by an industry consensus. For example the American Dairy Science Association developed a scale used for milk products: 10 and 9 (excellent), 8 (good), 7 (fair), 6 (poor), to less than 6 being ‘unacceptable, a probable consumer complaint’. In some cases specific modalities or attributes are measured and the scores or grades summed to give an overall indication of product quality. Figure 4.14 gives an example of a quality score based on ‘typicality’. There can be many disadvantages of this method. Firstly, because it relies on the expertise of the panellist, the results may not be directly linkable to consumers’ opinions of the products. Panellists can also drift into their own scoring system based on their own likes and dislikes. For new product development it can cause issues as there is generally little understanding of consumers’ opinions of the new product at the first production stage. In some cases the method is very poorly used with small numbers of panellists without the necessary training experience – this tends to lead to panellists making their own personal judgements on the products. If an overall scoring system is used this does not give the information required to fix the problem for the next production run or if the product might be re-blended or used for a different product as there is no information about which sensory

Please assess each coded sample below and score according to the table below. Scale

Definition

1 2 3 4 5

Fresh, typical, full flavour, no aged, no stale, no rancid flavours Fresh, typical, slightly lacking flavour, no off notes Relatively fresh, typical, however dull flavour Flavour slightly unbalanced with ageing, stale notes Aged, stale, rancid, not typical Code

Score

123

_____

456

_____

789

_____

Fig. 4.14 Example of the ballot paper for a quality score method.


72


characteristics are causing the quality issues. With the correct controls in place (Muñoz et al., 1992) this method can be fast and economic to use and, when backed by industry standards and linked to consumer acceptability scores, can be useful for management in making quality decisions. However Muñoz et al. recommend that unless the rules are adhered to, companies would be better placed if they chose another recommended method (such as DS or in/out) owing to the inherent disadvantages of quality scoring and grading.

4.11 Magnitude estimation and duo–trio methods The magnitude estimation method is very simple but rarely used for dayto-day quality ratings except for specific products such as the assessment of chilli peppers. The method is based on Steven’s law. Panellists are given a reference sample and told, for example, that the reference would score 50 for a specific attribute (e.g. sweetness, crispiness, chilli heat). Subsequent samples are then scored in comparison to this reference. For example if the next chilli was twice as hot as the reference, the sample would score 100. For more information see the ISO standard 11056:1999. Muñoz et al. (1992) state that the duo–trio test is not an ideal method in QC as it can be very sensitive to small differences and can therefore create too many false positives. The test was developed by Peryam and Swartz in 1950 for quality control in distilleries. It was proposed as an advantage over the triangle test as it was thought to be easier for panellists to match rather than compare three unknowns (Stone and Sidel, 2004). The test measures if there are any differences between two products but three products are assessed. This method is similar to the ‘A’ not ‘A’ and the triangle test in that a reference sample (A) is given to the panellist and then they are presented with a pair of samples and asked which of the pair matches ‘A’. Unlike the ‘A’ not ‘A’ test, all three samples are presented simultaneously. The test does not give any indication as to the nature of the difference. An advantage of this test is that the analysis of the data from this method is very simple: the scientist looks up the number of correct answers in statistical tables and reports the result depending on the significance. More than 15 panellists are required: the ideal minimum number being around 30 with equal numbers of each possible combination. There are two formats to the duo–trio test: where the reference is the same for each panellist (constant reference) and where the reference is a balanced representation of both of the samples in the test (balanced reference). Where panellists are familiar with the reference sample the first option appears more sensitive (Lawless and Heymann, 1999). An example questionnaire for both methods is given in Fig. 4.15. The constant reference method, where panellists are familiar with the reference, can be very sensitive to differences attributable to ingredient substitution.



73

Constant reference duo–trio Instructions There are three samples for you to assess. One of the coded pairs is the same as the reference. Please assess the reference first, then the two coded samples in the order shown below. Please ring the code of the sample which is most similar to the reference Reference

456

789


Balanced reference duo–trio Instructions There are three samples for you to assess. One of the coded pairs is the same as the reference. Please assess the reference first, then the two coded samples in the order shown below. Please ring the code of the sample which is most similar to the reference Reference

456

789


Fig. 4.15 Example of the ballot paper for the duo–trio test method. The questionnaire is identical for both tests; it is the presentation of the samples that differs.

4.12 In-house and do-it-yourself (DIY) methods There are many in-house sensory methods developed for quality control and some of these have become very popular and used by other industries: for example the duo–trio test. Many companies have internal grading systems based on typicality, product quality (using scales such as bad to good) and even undocumented assessments where the line-staff simply assess the product and tick a box to say if the product is satisfactory or not. As mentioned in the introduction to this chapter, these subjective methods are not ideal, and the use of the objective, consumer-led methods are recommended. However, some in-house methods have been developed over many years, have detailed Standard Operating Procedures and Work Instructions and are based on sensory specifications and consumer data, resulting in more objective and controlled results. Hybrids of the recommended methods can be used to produce a detailed programme of tests for the different situations a quality control sensory scientist might find themselves in. For example the sensory scientist might


74


choose the difference from control method for monitoring batch variation and for ingredient substitution tests, the descriptive specifications method for passing daily batches and the ‘in/out’ method for passing raw materials.

4.13 References bsi (1988) British Standard Methods for Sensory Analysis of food, Part 5 ‘A’ not ‘A’ test, BS 5929: Part 5. bsi (1989) British Standard Methods for Sensory Analysis of food, Part 6 Ranking, BS 5929: Part 6. carpenter, r p, lyon d h and hasdell, t a (2000) Guideline for Sensory Analysis in Food Product Development and Quality Control, Aspen. costell, e (2002) ‘A comparison of sensory methods in quality control’, Food Quality and Preference, 13, 341–353. iso Sensory Analysis – Methodology – Triangle test, 4120:2004. www.iso.org. iso Sensory Analysis – Methodology – Magnitude estimation method, 11056:1999. www.iso.org. lawless, h t and heymann, h (1999) Sensory Evaluation of Food. Principles and practices, Aspen. meilgaard, m, civille, g v and carr, b t (1999) Sensory Evaluation Techniques, CRC Press. muñoz, m, civille, g v and carr, b t (1992) Sensory Evaluation in Quality Control, Van Nostrand Reinhold. oakland, j (2007) Statistical Process Control, Butterworth-Heinemann. o’mahoney, m (1986) Sensory Evaluation of Food: Statistical methods and procedures, Marcel Dekker. pecore, s et al. (2006) Degree of difference testing: a new approach incorporating control lot variability, Food Quality and Preference, 17, 552–555. stone, h and sidel, j l (2004) Sensory Evaluation Practices, Academic Press. young, t a et al. (2008) ‘Incorporating test and control product variability in degree of difference tests’, Food Quality and Preference, 19, 734–736, doi:10.1016/j. foodqual.2008.04.002.


5 Establishing product sensory specifications C. J. M. Beeren, Leatherhead Food Research, UK

Abstract: To obtain a consistent product quality, with characteristics as desired by the consumer, detailed product specifications are vital. Product specifications can be created by using different methods, but will always need the input from consumers and from different disciplines within the organisation such as product development, manufacturing and quality control. Any product attributes identified should be well defined, unambiguous and well understood by assessors to ensure consistent use and results. Once a reliable sensory specification has been developed, production, competitor and development samples can be compared against the specification to check sensory quality. All aspects from development, to implementation, use and follow-up of the product sensory specifications are discussed in this chapter. Key words: product specification, sensory characteristics, quality, key attributes.

5.1 Introduction A consumer’s first purchase of a food or beverage product is probably influenced by information gathered from different sources, such as data available from commercials, recommendations by family or friends and details from promotion in store, or the decision may be influenced by the price. At this point of the first purchase, little information would be available about the sensorial characteristics; a consumer would not know yet whether he or she personally would find the product as tasty as a friend made it sound, or whether the product really looks as good on the table as the commercial made it look. A few exceptions, though, exist; as for some products the human senses of sight and or smell may play an important factor influencing this first purchase. With our sense of sight we measure visual aspects of products, and as such many products benefit from an appealing product packaging, and position on the shelf. In addition, for a


76


smaller number of products, the product itself is visible to the consumer, for example a pizza may be seen through a window in the packaging, or a cheese could be seen at the counter. For these occasions, sight would play a major role in the first purchase decision. The aroma of the food or beverage plays an important factor during a first purchase for only a minority of products, but where it does, it will be vital, such as the smell coming from a bakery aisle or from a roasted chicken in a delicatessen. When a food or beverage is re-purchased, the influence of the product’s sensory characteristics is obvious; the consumer knows how the product looks, smells, tastes and feels, and will have certain expectations about the sensorial characteristics of the food or beverage. Consequently, a consistent sensory quality is inherent for continued consumer support and thus to make a product successful. Naturally, consumers vary from individual to individual, and also within individual; thus they will make their food choices based on numerous factors, including their mood state, the time of day, the social surroundings, etc. The total food concept is therefore not only based on the sensorial product characteristics, but is a combination of these individual product characteristics and the environment in which the product is bought and eaten (Earle et al., 2001). The different factors influencing the purchase introduce many possibilities for manufacturers, but also a need to clearly communicate the food brand, characteristics and values to inform consumers and ultimately to convince them that the product is the right choice. To develop and position a product accurately, the involvement of the target consumer is vital. The consumer’s needs or wishes in terms of product ideas, concepts and potential product characteristics must be known by the product developer. Prototypes should be evaluated and once a product is launched and in the marketplace, ongoing consumer research should be carried out, ensuring continuing product fulfilment and measuring any possible desired changes to the product characteristics. Ideally, consumer data should be correlated to specific emotional or environmental conditions as this will give an enhanced insight to the consumers’ view of the product and may give potential for further or new developments.

5.1.1 Instrumental measurements Usually quality control (QC) uses a combination of different analyses, likely to include both instrumental and sensory analysis. The exact test types used will be dependent on different factors such as product type, amount of samples to be evaluated, staff availability and experience. In a study by Gimeno et al. (2000), the use of colour measurement by the L a b (lightness, red/green, blue/yellow) system and instrumental pH measurement is demonstrated in chorizo samples; the b variability of the colour measurement indicated different amounts of paprika used and


Establishing product sensory specifications

77

the pH correlated with a textural sensory property; the cohesiveness of the chorizo samples. This example illustrates that for some attributes an instrumental method could be successfully used to measure a sensory characteristic of the product. The advantages of using an instrumental specification for colour measurements above sensory measurements is described by Hutchings (1999); the instrumental method is not susceptible to eye fatigue, poor colour memory, suitability of assessment condition – including uniform lighting – availability of trained graders and in addition, instrumental colour measurements may also be less time consuming than measurements carried out by sensory assessors. The effectiveness of the instrumental technique should, however, always be validated first by correlating the results obtained from the instrument with the perception of the consumer. Often a combination of sensory and instrumental measurements is used to define product quality; this is illustrated by Bruwer et al. (2007) who carried out a study for a tortilla chip producer to characterise the textural properties and variations of the food product. Instrumental data and data obtained from a trained sensory panel were used, resulting in two robust quality variables for developing online sensors for process measurements.

5.1.2 Sensorial measurements Although for some characteristics it may be appropriate to use instrumental measurements, it should always be ensured that the instrumental method correctly reflects the character as perceived by the human senses and thus by the consumer. Instruments and the human senses may respond in different ways to the same input, resulting in a different output. Additionally different product characteristics will interact with each other, resulting in a different perception. For example, the colour perception of white and yellow corns appears affected by the variation of the pericarp thickness and the glossiness and thus instrumental methods appear to be not as effective as evaluation by sensory evaluation to distinguish subtle colour differences of this product (Hutchings, 1999; Floyd et al., 1995). Another example pointing out the importance of sensorial measurement is described by Paganuzzi and Carozzi (2000). The results of their paper describe the significance of the sensory measurements and the subsequent request for EU approval to change the product specification of an extra virgin olive oil DOC (denominazione d’origine controllata). Muñoz (2002) indicates that companies that do not use sensory practices in their QC operations and have not confirmed the correlation between sensory and instrumental methods, either are unable to detect the sensory issues in their products, and may thus fail to check products comprehensively, or may not have had a major quality problem involving sensory issues to acknowledge the value of sensory measurements in QC. Furthermore, Feria-Morales (2002) indicates that many foods are sold according to their


78


sensory quality, which is not easily measured by conventional analytical techniques. More information on the correlation between sensorial and instrumental methods can be found in Chapter 6 of this book.

5.2 Rationale using sensory specifications The importance of the sensory characteristics of food and beverage products for consumers clearly shows the requirement for the manufacturer to produce products with consistent characteristics meeting the demand of the consumer. It is not only end consumers who expect a consistent sensory quality; food and beverage manufacturers also expect a reliable, constant quality for the bought ingredients. Product specifications are often supplied by the ingredient suppliers to communicate the specific product quality characteristics between businesses. Moreover, specification sheets appear the most common service given by ingredient suppliers and are also one of the most important factors in choosing a vendor for the buyer (Berglind, 2003). Not only are product specifications created by ingredient companies supplying products to other food manufacturers, product sensory specifications are an essential tool to aid with product consistency and represent one of the most important tools of a QC program. Well-defined product sensory specifications can be reliably used for all subsequent assessments to monitor the sensory quality, comparing real product characteristics with the set specifications (Metheringham and Rodway, 2001). Short-term benefits of using sensory product specifications include the provision of fast, objective and focused results, whilst long-term benefits include helping reduce quality fluctuations and the identification of the sensory critical control points within the production process. The use of sensory specifications also promotes awareness of the sensory quality of products by staff, tracks product consistency and aids communication with the industry (Metheringham and Rodway, 2001).

5.3 Defining sensory specifications It may be difficult for consumers to identify and articulate the specific sensory attributes that influence their acceptability for particular food products. The interaction of different, even simple tastes, such as sweet and acidic taste, or colour and fruit flavours can make product evaluation and identification of exact characteristics complex (Earle et al., 2001). Developing terms to describe products in aroma, appearance, flavour and texture is therefore often carried out by trained sensory panels or by panels of product experts. The terms derived should be specific enough to be useful for



79

product development and for efficient usage for QC purposes; to merely describe a tomato sauce as ‘red’ will not give enough information, as there are many varieties of a red colour (Staniforth, 2004) or to identify a ‘typical’ flavour does not define the product’s flavour and thus more detailed descriptions are required. After creation of well-defined product descriptors, the descriptors should be correlated with the product liking of consumers. This will create objective measurement criteria for consumers’ ideals and will identify the key quality attributes required for product development, quality control testing and to aid with shelf-life determination of products.

5.3.1 Target product When creating the sensory specification, the following factors should be considered: • • • • • • • • • • •

target product; available consumer information; product ingredients; production process; storage conditions; packaging; brand; function of product; product quality requirements; product transportation; product marketing.

The basis of the creation of the sensory specification will be the target product as this was created by product developers based on the desires of the consumer. This target product should be sensible and viable within production, cost and shelf-life constraints and the produced and marketed food or beverage should mirror this reference product. Information gathered from consumers during the development stage may also prove very useful as guidance for consumers’ requirements. This could include the desired consumer characteristics obtained from product brainstorming and from any consumer product research carried out on concepts, prototypes and/or final versions of the newly developed product. The process that a food or beverage will undergo plays a major role on the end target product, and as such (key) product ingredients, production processes, packaging and storage conditions should be studied when creating the sensory specification, as these process factors may have an influence on the sensory profile of the end product. Ideally, several product variations would be made available during the generation of the product sensory specification, using minimum and maximum levels of particular


80


key ingredients or processes, storing samples under different conditions, (if relevant) in different packaging materials and products of different ages. During the creation of the product specification, all factors influencing the total quality of the target product should be taken into account; product, price, place and promotion, referred to as the four Ps in the marketing mix. The product in this mix is in a condition that consumers expect to receive it (Blythe, 2008). Within the four Ps, the following factors should be included in the considerations: brand, function, quality, packaging, safety, pricing decisions, storage, transportation, market coverage, advertising and promotional strategies (Internet Center for Management and Business Administration, Inc., 2009).

5.3.2 Identification of critical consumer attributes Attributes used for the sensory product specification should be the ones that are key to consumers’ acceptance to ensure that the product reflects the consumers’ desire. Hence, consumers should be involved when setting up the product sensory specifications. Data generated during product development and product testing may be considered first, which may avoid excessive expenses. It should be noted, though, that even a minor change in product, packaging, promotion or price may have affected consumers’ expectations and thus consumer insight should be obtained from the target product in case earlier testing was carried out on a different product. Furthermore, consumers’ liking is an evolving process: consumers may change their mind, at shorter or longer time frames, as other products or variants come on the market or as people change and thus frequent testing of consumers’ opinion is recommended. Identification of critical consumer attributes includes also removing any obsolete attributes, which are the attributes that are not important to consumers. Too extensive product sensory specifications may lead to overtesting and a waste of time and resources.

5.3.3 Creation of the product sensory specification Key quality attributes defined should be based upon human perception, capturing the different consumption stages. Visual attributes are generally evaluated first, followed by aroma, taste/flavour, texture/mouth-feel and at the end of the consumption, after swallowing, specific aftertastes and an afterfeel may be perceived. For products where aroma can change rapidly or where the aroma can be perceived readily, for example, hot served products such as coffee, the aroma may be evaluated prior to the appearance to ensure all important aroma characters are captured. Creation of the sensory product specification by highly experienced product specialists may not always be most optimal. The product experts



81

could be inclined to incorporate their own likes and dislikes, biasing the outcome. Trained, pre-selected assessors objectively describing the main product characteristics and setting realistic, measurable sensory standards may therefore be more appropriate as the main method creating the attribute lists (Metheringham and Rodway, 2001). It is, however, recommended that different disciplines are involved when creating the product sensory specification, including, but not exclusively, product development, ingredient purchasing, production, process development, and sales and marketing. Input from client, retailer or food manufacturer may be essential at this stage too. The different experts are able to identify possible product variability. Ingredient variability should be taken into account also; some ingredients subjected to variability will inherently alter the end product, also indicating the need for proper raw material/ ingredients QC and thus ingredient sensory specifications. Quality limits for raw materials must be set to allow for the finished product to meet its specification; however, limits should be broad enough to allow for some flexibility in purchasing (Matz, 1992). Sensible quality limits for changes occurring during shelf-life and storage should also be set and measured against. Any probable deviations of the standard product quality should be captured during the evaluation of the product and the degree of variability of products’ sensory characteristics should be known. When these are covered in the sensory specification, it would be less likely that these would be overlooked and assessors can be trained on acceptability levels. Unfortunately unexpected events happen, leading to quality deviations such as taints or off-flavours which were not considered at any stage. Including an attribute named ‘other’, ‘off-note’ or the possibility of giving open comments prevents assessors’ dumping the perceived difference into another attribute. The final product specification however, should reflect ‘conformance to customers’ requirements’ and the unit operations should function to be in compliance with this specification, as a product that fails to reflect the features required by the customer will lead to rejection of the product and a product that exceeds expectations may be compromising economical viability (Bonnel, 1994) or raising consumer expectations, which may not be met at a re-purchase of a standard quality product. To summarise, the generated list would; • contain the key quality attributes vital to consumers; • contain attributes which may vary (due to ingredient or process variation); • ask assessors to describe any perceived characteristics not mentioned in the standard list, if relevant; • be written in a sensible order, similarly to the consumption experience of these products.


82


Attribute definitions and ranges All product attributes generated should be clearly defined, to avoid any ambiguities and to ensure all assessors would have a clear understanding of the characteristic to be evaluated. Attributes such as fresh, natural or typical can easily be misinterpreted and should thus be avoided. Also simple attributes such as hardness of a chocolate sample could be misinterpreted if this was without any further clarification; some respondents may, for example, assume that the attribute evaluates the hardness when the product is first bitten into, while other assessors could assume that a hard chocolate would be a chocolate which would stay hard, even after repetitive chewing. An example of a product attribute list, including definitions, is shown in Table 5.1. Following the creation of critical measurable sensory attributes, viable acceptability ranges for each of the attributes must be established. The

Table 5.1 Sensory attribute list – chocolate product Appearance Brown colour Gloss Evenness of surface Air

Brown chocolate colour intensity Light scattering on surface The presence of pits/damage on the surface The presence of air after breaking the chocolate

Aroma Overall aroma intensity Cocoa Stale Off

Total aroma strength Aroma of cocoa powder Musty, cardboard aroma Atypical aroma, e.g. rubber

Texture Hardness on first bite Smoothness Mouth-coating Body

Force required to shear sample on first bite Initial smoothness assessed when tongue yields chocolate Coating of mouth Thickness of sample

Flavour Overall flavour intensity Sweet Bitter Cocoa Nutty Stale Off

Total flavour strength Basic taste of sucrose Basic taste of caffeine Flavour of cocoa powder Flavour of almonds/nut skins Musty, cardboard flavour Atypical flavour, e.g. rubber

After-effects Sweet Bitter Cocoa Stale Off flavour

Sweet aftertaste Bitter aftertaste Cocoa powder aftertaste Musty, cardboard Atypical aftertaste, e.g. rubber



83

range should be viable within production and buying constraints and reflect consumers’ expectations of the product. Similarly to the definition of the attributes, the acceptable ranges for each attribute should be unambiguous. ‘Few to many’ raisins in an apple pie may be perceived in a different way by different assessors and should be more accurate, unless dealing with very trained assessors evaluating the products. 5.3.4 Role of quality control during establishment of product specification The main roles for sensory testing within the QC department are to maintain consistent product quality, the prevention of taints and the assessment of product shelf-life. As part of the development of the sensory product specification, QC should oversee the progress and ensure a complete list of attributes and ranges capturing all critical elements, whilst creating a workable specification for the manufacturer. Usable and actionable results are required to ensure follow-up of any products not falling within the specification. Product results should also be observed and specifications be amended as necessary.

5.4 Reference samples To assist with the evaluation, suitable reference samples will aid the understanding of sensory attributes and ranges. Reference samples will guide as a baseline and reduce assessment variability. Suitable reference samples often include (frozen) factory obtained samples, different aged, ‘spiked’ or abused samples and chemicals to illustrate particular attributes and visual references, illustrating the range of product variability. Identification of visual references, such as colour charts or photographs, is appropriate to assist with the appearance attributes, exact colours, hues and intensities are often difficult to remember for assessors and are easily referred to with colour charts. Good quality charts and images would remain unchanged and allow faultless use over time unlike the use of standard photographs/prints which may be subject to variability with processing and change over time. Any attributes with which assessors may be less familiar would benefit from demonstration using reference products, preferably shown at different intensities, e.g., different cooking times and different levels of ingredients. 5.4.1 Gold standard Reference products may be a gold standard of the product, where assessors can compare the product ‘like for like’ and compare directly for any differences. If this method of gold standard is used, QC must ensure a consistent reference, which may be freshly prepared or stably stored, e.g. frozen.


84


Refreshment of gold standards on a frequent basis could lead to a drifting reference sample; a very gradual change may not be perceived on a one-to-one comparison between successive batches, but could add up to a change in quality over time. More robust testing of replacement of the gold standard could prevent this gradual drift. 5.4.2 Competitive material samples as references When a gold standard is not at hand, ‘equivalent’ competitive products could be acquired with which to compare production samples. It should be noted that this comparison, although often helpful, is relative to the consistency of the competitive material. Advantages of comparing with competitive products include building up knowledge of differences between manufactured and competitive products on the market, establishing a benchmark. 5.4.3 Rejected or manipulated samples as references To illustrate different ranges of attributes, rejected or manipulated products may prove very valuable. Any rejected or borderline products should if possible be kept for training or reference purposes of the sensory evaluation. These products are direct illustrations of non-ideal products, and assessors will immediately understand the product limits. To demonstrate particular product ranges, specifically to clarify attribute ranges possibly occurring during manufacturing or from ingredient supply, manipulated products may be very useful. Smaller or larger volumes of specific ingredients could be added, products could be exposed to specific aromas in an enclosed environment, additional ingredients can be added or products can be prepared differently. It will speak for itself that only food grade materials and products with a known history can be used and that any products, including any rejected or manipulated samples given to assessors, should be safe to consume and be within national regulations. 5.4.4 Daily products as references To describe specific product characteristics, such as certain aromas, flavours or textures, well-known daily products could be referred to. The daily reference products do not have to be similar to the evaluated product; for example, vanillin sugar to illustrate the vanilla aroma note in cookies and cinnamon powder to illustrate a spicy flavour characteristic in a beverage.

5.5 Implementation of sensory specifications To optimally use the created product sensory specifications, good sensory procedures should be used for the assessment of products. A simple but



85

robust method to measure the desired product characteristics would normally be favourable within the QC environment. Naturally, availability and viability of instrumental and sensorial measurements and the company’s capability to use the measurements should be taken into account, using solely the most appropriate systems for measurement of product quality.

5.5.1 Sensory assessors Several respondents are required for sensory product assessments. These respondents could be recruited internally or externally. The advantages of using internal assessors may include their presence in case of urgent need for evaluation and being more cost effective, and the disadvantages are likely time constraints and subjectivity. Focusing on externally recruited assessors, a personal interest in the outcome of the test would be less probable and they would spend more time focusing better at the job at hand. Sensory screening potential assessors Assessors used for sensory evaluation of quality control checks are typically screened and selected for their ability to distinguish small differences in appearance, aroma, taste, flavour and texture and to verbalise these differences. Generally, screening tests include a basic taste recognition test, whereby four or five basic tastes would be presented and evaluated, consisting of sweet, salt, sour, bitter and possibly umami. In some instances astringent could be included as an additional sensory characteristic; for example, for chocolate or wine panels. As not all respondents would be familiar with all basic tastes, a prior exposure to all components is appropriate. Other usual screening tests include detection and recognition of aromas, whereby generally approximately five different components are evaluated and assessors try to recognise and describe the perceived smell. To be selected, assessors would at least be able to smell the samples and to describe products in the direction of the exact material. The ISO standard (ISO 8586-1) suggests using materials such as benzaldehyde for almond/ cherry aroma and vanillin for a vanilla aroma. Incorporating common taint samples, such as trichloroanisoles (TCA) and trichlorophenols (TCP) is common too, as about 70% of taints are caused by these two components. Taint issues should be picked up during the QC sensory evaluation and it is thus important that assessors are sensitive to these components. To ensure that assessors would also be able to describe any deviations from the sensory specification, term derivation of products is also a common part of a screening procedure. Term derivation could be of any specific food, beverage or illustration. Other tests that may be used for assessor screening are dependent on the role of the assessors, and could include specific discrimination tests, such as triangle or paired comparison, memory tests, tests to identify visual


86


impairments, ranking tests, etc. ISO standards do give further details on screening of assessors (ISO 8586-1). It is important that the sensory analyst knows any personal details of assessors, such as allergy information or strong likes or dislikes. Further, any assessor who is part of a sensory panel should be motivated and not be forced into it. Sensory training potential assessors To fully comprehend the test products and entirely understand the test methods and procedures, sensory training for assessors is required. Any assessors not trained prior to the assessment in the methods or products used may be less confident and are unlikely to deviate from any standard results, and thus score any product as acceptable. Exposure to products will help identifying the varieties of the products concerned; Heckel and Wilson (2002) describe the recognition of deviation from the norm in terms of appearance and flavour by a trained person; samplers not exposed to several samples may not catch the potential variation and problems. Training may include background details on the product itself, such as details on ingredients, and the process to discover the sources of specific product characteristics and exposure to the several test products and variants, including tainted test samples. The test method should be explained and practised, and test procedures should be discussed. Any customer feedback, case studies and unacceptable samples could be shared and discussed to gain a better understanding of the product quality. During training, it is important to evaluate and discuss acceptability ranges, to ensure all assessors would fully understand the acceptable levels. It is useful to compare training results with results from already trained panels and with instrumental data. Correlation between the sensory and instrumental data is not always straightforward, as human perception may be influenced by different factors. Any correlations or deviations both with already trained sensory assessors and with instrumental data, would be valuable information for discussions between assessors and the panel leader(s). When carried out properly, training also tends to work very well to motivate assessors.

5.5.2 Sensory test methods Test procedures chosen to maintain product quality should consider the viability within company constraints. Testing should be practical and consider factors such as the assessors’ availability, test frequency, and the test facility to provide useful and reliable sensory data. To evaluate the product against the created sensory product specification, different types of sensory testing could be applied.Whilst Chapter 4 described different test methods which can be used in more detail and specific test



87

methods are extensively described in textbooks (Lawless and Heymann, 1998; Brinkman, 2002; Meilgaard et al., 2007), a brief overview of specific methods relevant for testing against sensory specifications is now given. Tests commonly used within QC departments for product comparison against sensory product specifications include: • • • • •

in/out method; grading; quality ratings; difference from control; descriptive testing.

In/out method The in/out method is a simple method, benefiting from its ease of use by assessors. When using the in/out method, also referred to as In-spec/out of spec or pass/fail, assessors are asked whether a specific sample is in or outside the defined sensory specification. Examples of ‘In-specification’ criteria for potatoes are uniform off-white colour, potato flavour, earthy flavour and soft texture after cooking for a specific time. Out-of-specification examples for a potato variant may be blackening/greying as illustrated with photographs, glassy appearance and off-flavour. Results can be tracked visually over time. Grading method The grading method is an extension of the in/out method, whereby samples are classified in different grades; generally three different grades, A, B, C, are employed, with ‘A’ usually describing a product within the defined specification, ‘B’ a product that is ‘just acceptable, but needs improvement’ and ‘C’ descibing a products which is ‘out-of-specification’ and should be refected. Table 5.2 shows an example of a grading of fish samples. Illustrations can be created to to view trends over specific periods (Figs. 5.1 and 5.2). Quality rating method Quality rating methods are used to rate specific quality attributes. Assessors are trained to recognise different quality levels of important attributes, defined in the product sensory specification. Each attribute will be rated by the assessors, based on its perceived sensory quality. Samples will be rated on line (unstructured scale) or category scales, examples of which are shown in Fig. 5.3 (line scale) and Fig. 5.4 (category scale). The quality rating method is a semi-quantitative method and statistical treatment of rated data is possible. Difference from control The difference from control test is used to establish whether a difference exists and to get an indication of the size of the difference between two or


88


Table 5.2

Aroma

Raw fish grades – Atlantic ground fish A

B

C

A detectable raw fish odour

Neutral odour

Presence of taint or off-odour, such as sour, ammonia, bilge Appearance Colour of bled fish Colour of bled fish Yellow or brown (chart A) (chart B) colour (chart C) No single blood clot or Blood clot(s) Blood clot(s) up to 0.5 cm in total with dimension exceeding 4 cm in maximum dimension between 0.5 and total dimension 4.0 cm Bruising/discoloration Bruising/ Bruising/discoloration not exceeding 2.0 cm discoloration exceeding 5.0 cm in total dimension between 2.0 and 5.0 cm Mouthfeel/ Firm Slightly soft Excessively soft Resilient; up to 10% Between 10 and More than 40% of texture surface area may 40% of surface the surface shows show gaping area may show gaping gaping Adapted from Bonnel (1994).

25 A

B

C

Frequency

20 15 10 5 0 1

2 Period

Fig. 5.1 Grading – visualisation over time period (1).

more products. A consistent control is required as a reference sample; this control would be the ‘gold standard’ as described in Section 5.4.1. Some statistical treatment is possible to identify whether the test samples are different from the reference sample. Descriptive testing With descriptive testing the nature and size of the difference of specific sensory product attributes are compared. Different types of descriptive



89

Sensory scoring

Grade

3

2 Year 1 Year 2

Dec

Oct

Nov

Sep

Aug

Jun

May

Apr

Mar

Feb

Jan

1

Month

Fig. 5.2 Grading – visualisation over time period (2).

Sweetness None

Very

None

Strong

Dry

Very moist

Lemon Flavour

Moistness

Hardness Soft

Fig. 5.3

1 None

Hard

Examples of line scales (adapted from Meilgaard et al., 2007).

2 Very Slight

3 Slight

4 5 6 7 8 9 Slight/ Moderately Moderately Strong Strong– Very Strong Moderately /Strong Very Strong

Fig. 5.4 Example category scale.

testing are available, such as quantitative descriptive analysis (QDA), flavour profiling (FP), texture profiling (TP), free choice profiling (FCP), spectrum descriptive analysis, consensus profiling and time-intensity methods. Most methods require extensive training and are usually more time consuming than earlier described test methods. The information


90


Off-flavour* Honey flavour

Yeast flavour

Fruit aroma* 60 50 40 30 20 10 0

Citrus aroma* Yeast aroma*

Overall flavour*

Sweet taste*

Citrus flavour* Fruity flavour*

Sour taste* Bitter taste 1

2

3

*Indicates a statistical significance difference between samples at 95% confidence level.

Fig. 5.5 Spider graph displaying sensory attributes.

obtained from descriptive testing, however, will be more in-depth, showing full profiles of evaluated samples and identifying very specific differences, visually (Fig. 5.5) and in terms of statistical significance.

5.5.3 Sensory test protocols Efforts must be made by food and beverage companies to measure and deliver a product with a consistent sensory quality; clearly defined test procedures and protocols will assist in keeping consistent quality and reducing bias during evaluation. Assessment area, number of assessors, frequency of assessment, product sampling procedures, product storage, sample size, sample presentation, sample treatment, data analysis and actions should all be considered carefully. Firstly, the testing objectives should be clearly defined, and agreed upon with all involved parties. Production, quality control, processing, product development and sales departments should be in agreement to the purpose and thus the objectives of the tests carried out. Test protocols should then be developed with the aim of reaching the set objectives. Sensory environment The environment in which the sensory testing is carried out has an influence not only on product perception, but also psychologically, indicating the importance of the testing. A dedicated room for testing shows the company’s commitment to sensory testing. Using the same dedicated area



91

for all tests would also lead to more consistent results; the room has an effect on the presentation of products and the appearance of the visual characteristics, and can be a comfort to assessors. Influences of the environment should be considered (ISO8589, 1988), such as: • consistent lighting, i.e. artificial light, and same types of lights throughout the room, ideally artificial daylight/north light fluorescent should be considered for optimal colour assessments; • noise may distract assessors and thus influence performance; • odours coming from production or preparation area should be kept to a minimum; • decoration may influence the perception of the visual appearance – neutral colours, such as matt off-white or light neutral grey are recommended; • individual testing booths to limit distraction and to avoid communication between assessors; • furthermore, preparation and storage area adequate for testing purposes should be in place to present samples appropriately and the testing area should be easily accessible. Assessors Screened and trained assessors can take part in the evaluation of products. Panel leaders should keep records of assessors’ sensitivity levels. In particular sensitivity results for any potential taint materials would be of interest in case an off-note were identified during the evaluation. Assessors should be told not to eat, drink or smoke prior to an evaluation session, not to use any (strong) fragrances and not test when they have a heavy cold. Assessors may respond differently towards test products depending on the individual, likes/dislikes, sensitivity, knowledge of the test (products), mood, familiarity of products, time of day, etc. A panel of assessors should therefore be employed to reduce personal factors. Training will aid establishing more objective evaluation of products and thus, generally, the more trained assessors are, the fewer assessors are required. However, panel leaders should always consider that however well trained, some variability will still exist between assessors, not least in sensitivity levels, knowledge of test and motivation. Other factors influencing the panel size include the amount of test samples, consequences of wrong decisions, etc. A panel of approximately six people for each assessment is not atypical. Samples The presentation of samples should be consistent; using consistent preparation methods delivering the same volume of sample(s), at the same


92


temperature and in the same manner during every evaluation. Assessors should handle the samples in the same way during each test. The test procedures to sample products from production, to store, prepare and present products should be documented. Within one test, samples are ideally prepared by the same person, avoiding differences due to preparation. Test procedures on how to handle and evaluate the food or beverage should be readily available for the assessors, ideally within the product specification. For example, the texture of products may be assessed by using the fingers, cutlery or in the mouth; each one may lead to different results. Records should be kept during testing, factors such as length of storage, product preparation, serving temperature and evaluation time can be utilised if any abnormalities appear or any questions arise at a later stage. The amount of samples tested should be kept to a minimum, particularly samples with a strong flavour, or acidic products. It may also be difficult for assessors to spend too much time away from their normal activities, and thus amount of time spent should be considered. More frequently a panel with fewer samples would be favoured above larger sample numbers. Samples should be presented blind, reducing biased judgements. Appropriate palate cleansers should be used to refresh assessors’ palates between each test sample. Water should always be available prior to a new sample. Further useful palate cleansers could include water biscuits or bread, lemon juice for fatty products, cucumber, yogurt for spicy products, milk for acidic products or carrots for soft textured products. Assessors should always drink water after any other palate cleanser and before evaluating the next product. The weakest flavoured samples should be presented first, and during preparation it should be ensured that all components of products are given to each of the assessors when multi-component products are evaluated. To reduce carry-over effects and adaptation, samples – of a similar strength – are randomised. Test methodology and protocol To avoid confusion by assessors, the same methodology should be utilised and well explained and understood by assessors. Sampling at regular times, ideally appropriately to the product – but not at a time that assessors are hungry or have just eaten – will help establish a consistent product assessment. Exact methods and protocols used during testing should remain as consistent as possible and assessors should be comfortable using these. When products are shared, good hygiene practices should be considered by assessors, and sufficient tools should be available for assessors, i.e., a sufficient amount of cutlery to use a new set for each assessment. The way the products are handled by assessors, including how products are manipulated in the mouth, should be clear from the sensory specification as this may affect the sensory perception. Some attributes may be judged on a first bite, others



93

after a specific amount of chews. Products are ideally swallowed to ensure all characteristics are well perceived. In some cases this may not be practical; then it would be advisable that all assessors should sip and spit and spittoons must be available.

5.6 Maintenance and follow-up When a product is recognised as out of spec, immediate action should be taken. Follow-up actions should thus be formulated at the time that the quality system is set up. It may be possible to reconsider the intended use of the product: products could be blended or the further processed for improvements. Heckel and Wilson (2002) describe the possibility of rebleaching or re-refining oils with high peroxide values: this process is then followed by the stabilisation by hydrogenation, and the alteration of oil blends in cases of different melting points. In other cases, however, products cannot be reworked and should be disposed of. Any form of deviation from the specification should be logged and investigated, to prevent reoccurrence. The investigation should focus on the root cause, and commonly will include further sensory evaluation with expert assessors to identify irregularity of the product and possible further analytical testing to identify the root cause. Seasonal variation can be an issue for natural products. Specification tolerances may need to be amended depending on the season. Analysis of trends will help identifying the variation: Fig. 5.2 shows a grading scheme of a particular product, whereby the first months in a year are liable to decline in quality. Trend analysis is also useful for identification in gradual reduction in quality. Figure 5.2 shows an indication of dropping quality in year 2. Management support and commitment are vital in decision making and ensuring subsequent actions are carried out. Tracking of sensory data over time may highlight potential issues in an earlier stage and will give a good overview on the performance of products. Table 5.3 illustrates different options in relation to test types. Accurate detailed records of all testing and of actions performed should be kept as evidence to prevent or reduce claims based on presumably defective products (Matz, 1993). Records should show that appropriate action is taken on products not meeting the sensory product specification. As consequences can be very serious, senior management should be involved in any decision making and creation of action standards on rejected products. Successful QC programs require ongoing technical support from R&D sensory professionals, support from the factory and factory management for guidance, training, creation of awareness on the importance and decision making (Muñoz, 2002).


94


Table 5.3 Tracking data by sensory methodology Sensory methodology

Tracking data

In/out method Grading Quality ratings Difference from control Descriptive testing

Visualisation of in or out over time Visualisation of classification over time (Fig. 5.1) Statistical difference per attribute over time and visually Statistical difference or size of difference from reference over time Statistical difference between products and visually (Fig. 5.5)

5.7 Case study Consistent, high-quality products are vital in the competitive market in which retailers operate. To ensure optimum and consistent quality of products, and to assist the suppliers in defining and measuring product quality, a tool to measure product quality is used. The product quality tool is a sensory product specification comprising a list of key attributes and a quality grading levels of each attribute. The product quality tool is part of the wider product specification agreed between the supplier and the retailer.

5.7.1 Set up product quality tool To create the product quality tool, the following process is applied; 1. Key attributes are identified by the supplier based on extensive product and process knowledge. 2. The list of attributes is rationalised based on consumer input. 3. Graphical images of different stages of the product are included, e.g. prepared and unprepared, frozen and defrosted, fresh and end of life product, start and end of season. 4. Graphical images are provided showing different quality levels of all visual attributes of the food or beverage and of the packaging, including preferred and inferior product quality. 5. Definitions and quality ranges of all aroma, flavour, texture and aftereffects attributes are created. 6. The supplier and retailer will discuss all available information and agree on the final product quality tool.

5.7.2 Using the product quality tool Once agreed, the product quality tool will be used by both the supplier and the retailer. The supplier will evaluate the finished products using the product quality tool before sending the product out to the retailer. The



95

retailer will test the products upon arrival to ensure the transport has not affected product quality. At each evaluation, a panel of assessors will score each attribute in one of the following three criteria: acceptable (A), borderline (B) and reject (R). The total score will be the lowest given score. If a borderline or reject score is given, an investigation will be carried out to find the cause for a (slightly) lesser quality. If the tests result in a borderline or reject score, the supplier and retailer will discuss these findings and agree on corrective actions. Assessors A minimum of three assessors carry out the evaluation individually. Each of these assessors is screened in their sensory perception, is familiar with the product quality tool and trained in the products, attributes and scoring system. Protocols Specific protocols for evaluation will be defined ensuring testing is carried out using the same principles, at supplier and at retailer and also over time. The protocol will include the following items: • exact product preparation instructions 䊊 equipment used, 䊊 length of time and temperature in microwave, oven, etc. • precise product serving instructions 䊊 serving temperature, 䊊 size of sample, 䊊 display of sample; • location of testing; • instructions for tasting 䊊 palate cleansers to use, 䊊 swallow sample or spit.

5.7.3 Results from product quality tool If a product is scored as acceptable, the product is ready to be distributed to the retailer. If a product is scored as borderline, a discussion with the supplier will take place and corrective actions will be taken. If a product is scored as rejected, it cannot be dispatched. The problem has to be solved first, and re-sampling will have to take place. Over time, the results are collected and frequent updates are issued to see trends occurring in the data. The results will highlight any trends over time, e.g. gradual increase or decrease in quality, seasonal variability and differences in ranges. The results kept are useful in case of consumer complaints and provide in-depth information on product quality.


96


5.8 References berglind, v. (2003). ‘Value-adding vendors’, Prepared Foods, vol. 172 (1), pp 69–70, 72 blythe, j. (2008). Essentials of Marketing, 4th Edition, Harlow, Pearson Education bonnel, a.d. (1994). Quality Assurance in Seafood Processing: A Practical Guide, New York, Chapman & Hall brinkman, j. (2002). Proeven van Succes, Sensorisch Onderzoek: Technieken, Procedures en Toepassingen, Houten, Keesing Noordervliet BV bruwer, m.-j., macgregor, j.f. and bourg jr., w.m. (2007). ‘Fusion of sensory and mechanical testing data to define measures of snack food texture’, Food Quality and Preference, vol. 18(6), pp 890–900 earle, m., earle, r. and anderson, a. (2001). Food Product Development, Cambridge, Woodhead Publishing Limited feria-morales, a.m. (2002). ‘Examining the case of green coffee to illustrate the limitations of grading systems/expert tasters in sensory evaluation for quality control’, Food Quality and Preference, vol. 13, pp 355–367 floyd, c.d. rooney, l.n. and bockholt, a.j. (1995) ‘Measuring desirable and undesirable color in white and yellow food corn, Cereal Chemistry, 72(5), pp 488–490 gimeno, o., ansorena, d., astiasarán, i. and bello, j. (2000). ‘Characterization of chorizo de Pamplona: instrumental measurements of colour and texture’, Food Chemistry, vol. 69, pp 195–200 heckel, c.b. and wilson, e. (2002). ‘Characterization of fats and oils, specifications and technical bulletins’, The Manufacturing Confectioner, vol. 82(6), pp 43–49 hutchings, j.b. (1999). Food Color and Appearance, Gaithersburg, Aspen Publishers internet center for management and business administration, inc. (2009). ‘The marketing mix (the 4 P’s of marketing), [Online], Available: http://www.netmba. com/marketing/mix/ [02 Feb 2009] iso 8586-1 (1993). Assessors for Sensory Analysis, Part 1. Guide to the Selection, Training and Monitoring of Selected Assessors, International Organization of Standardization (ISO) iso 8589 (1988). Sensory Analysis – General Guidance for the Design of Test Rooms, International Organization of Standardization (ISO) lawless, h.t. and heymann, h. (1998). Sensory Evaluation of Food, New York, Chapman & Hall/International Thompson Publishing matz, s.a. (1992). Cookie and Cracker Technology, 3rd Edition, New York, Van Nostrand Reinhold/AVI matz, s.a. (1993). Snack Food Technology, 3rd Edition, New York, Van Nostrand Reinhold/AVI meilgaard, m.c., civille, g.v. and carr, b.t. (2007). Sensory Evaluation Techniques, 4th Edition, Boca Raton, CRC Press metheringham, t. and rodway, l. (2001). ‘Quality in practice using proven sensory techniques to aid quality control’, New Food, issue 2, pp 19, 21–23 muñoz, a.m. (2002). ‘Sensory evaluation in quality control; an overview, new developments and future opportunities’, Food Quality and Preference, vol. 13, pp 329–339 paganuzzi, v. and carozzi, s. (2000). ‘Organoleptic characteristics of the extra virgin olive oil of the “Riviera ligure” designation’, La Rivista Italiana Delle Sostanze Grasse, vol. 77(1), pp 5–10 staniforth, j. (2004). ‘A question of consistency and quality’, New Food, Issue 2, pp 20–22


6 Combining instrumental and sensory methods in food quality control D. Kilcast, Consultant, Food and Beverage Sensory Quality, UK

Abstract: Sensory evaluation is the primary method of acquiring valid quality information; however, for practical reasons instrumental methods are frequently used, but are also commonly misused. This chapter describes the various types of instrumental measurements that are relevant to appearance, texture and flavour, and then outlines how relevant data can be extracted and correlated with sensory data, using different types of statistical procedures. The chapter concludes with descriptions of more advanced testing methods, including in vivo methods, non-destructive procedures and also the use of electronic noses and tongues. Key words: sensory-instrumental, statistical analysis, correlation, method validation, electronic noses and tongues.

6.1

Introduction: the perceptual basis of food quality

Consumers fortunate enough to live in prosperous societies have the choice of an enormous and ever-increasing range of foods, and manufacturers find themselves in an intensely competitive situation. In less well-developed societies, hunger will be the constant driving force, and our diet will be determined by availability of any food that satisfies our basic nutritional needs. It is increasingly clear that if we are to understand what drives consumer choice of food, no single factor can be considered in isolation from other factors. For some years, psychology researchers have been developing models to understand consumer behaviour (e.g. Shepherd and Sparks, 1994). Although there are many circumstances under which non-sensory factors such as price, advertising and nutritional image can have strong effects, delivering the sensory characteristics of foods is required by consumers central to continued purchase of foods. The importance of a holistic approach is also becoming more clear when the components of sensory perception are considered. During the sequences


98


of actions that constitute food consumption we perceive a whole range of different characteristics relating to the appearance, flavour and texture of the food. Traditionally, it has been common industrial practice to consider these characteristics individually when analysing and designing food sensory quality, and this can be seen in the development of sensory methods that are specific to certain characteristics, for example the flavour profile method and the texture profile method (details of both methods can be found in Lawless and Heymann, 1998). Current sensory measurement systems, however, are increasingly focused on assessing all sensory factors that are likely to be important to perceived quality, and on understanding how these interact at both physiological and psychological levels. Numerous sensory methods have consequently been developed to assess various aspects of sensory quality for both research and quality control purposes, and these have been described in more detail in this volume. The relevance of these sensory quality measurements (made by trained panels) to likely consumer response should ideally also have been established by carrying out appropriate correlation studies. The information secured by such research is vital in maximising product success, but can be out of reach of smaller companies operating on limited budgets, and there is a danger of falling prey to the temptation of extrapolating too far from a limited number of non-validated quality measurements. Similar considerations can also lead to the uncritical use of instrumental measurements that are assumed to be relevant to sensory quality.

6.2 The role of instrumental measurement The development of applied sensory techniques for evaluating the quality of consumer goods has been most extensive in the food and beverages industry, almost certainly reflecting the intimate contact that users have with the finished product. In contrast, until relatively recently other manufacturers of consumer goods have relied almost exclusively on using various types of appropriate instrumental measurement methods to ensure that any important perceived sensory characteristics of the product are as intended. The extent of the use of instrumental methods in different industries therefore reflects the difficulties inherent in the availability of validated sensory techniques. Given the wide range of sensory techniques available to the food and beverage sector then, why is there such extensive usage of instrumental measurement methods in quality control functions, and why is there a growing demand for the development of new methods? Many possible answers to these questions can be proposed. For example: • Logistical difficulties (especially in terms of time and cost) in setting up sensory panels, especially in small companies. • Staff downsizing policies, giving rise to difficulties in securing adequate panellist numbers.


Combining instrumental and sensory methods

99

• The realisation that in order to maintain even a basic sensory quality control (QC) system, resources in terms of facilities and panellist training require investment. • Instability of results from sensory panels over long time periods. • Possibility of contamination (accidental or malicious) of product by toxic chemicals, especially when investigating consumer complaint returns. • The manufacturing business produces large numbers of small batches of different products, and key customers demand 100% batch testing. • An unfounded expectation that there will be a simple and invariant 1 : 1 correlation between an instrumental parameter with a key sensory characteristic. • Lack of appreciation of the power and relevance of formal testing procedures, and a failure to recognise that uncontrolled informal sampling procedures are not an adequate substitute. • A naive faith in data that is generated by modern electronic instrumentation. Lawless and Heymann (1998) have also pointed out that instrumental measurements should be used for evaluations that are repetitive, fatiguing and dangerous, and when decisions made with the data are not business critical – again, providing that a correlation can be established. Irrespective of these concerns, consumer enjoyment of foods and beverages will be determined principally by a wide range of responses from the senses, and no instrument (or set of instruments) will be able to mimic these in the foreseeable future. However, the concerns listed above are not trivial, and although all companies must take all possible steps to employ sensory methods in QC, instrumental methods will continue to provide valuable quality input, provided that steps are taken to establish that the measurements relate to relevant sensory characteristics.

6.3 Sensory analysis of quality Details of the use of sensory methods in a QC environment are covered in other chapters in this volume, and this section will be restricted to an overview of the type of sensory characteristics that instrumental methods are designed to relate to. 6.3.1 The human senses It is generally accepted that humans have five senses in operation, namely sight, smell, taste, touch and hearing, although warmth, cold, movement and pain may also be considered as senses of importance in a food context. Foods are complex mixtures of chemical compounds, arranged into structural units. The perception of the sensory characteristics of foods results


100


from the stimulation of all our senses to some extent by the physicochemical properties of the foods. The sensory characteristics of food are generally grouped into three categories, namely appearance, flavour and texture. These categories are, however, not independent of one another. For example, colour, which is obviously an important appearance characteristic, can be shown to have an influence on flavour perception; consumers will assign higher scores for flavour intensity to darker foods than to lighter foods. The interaction between appearance and flavour is referred to as ‘visual flavour’. Similarly, textural characteristics such as viscosity can influence the perception of flavour, and some flavour characteristics, e.g. acidity, can affect textural characteristics. One means of defining flavour, texture and appearance is by taking into account the fact that each can be attributed to the stimulation of one or possibly two of the senses. On this basis the International Organization for Standardization (ISO, 1992) has proposed working definitions for flavour, texture and appearance, as given below. • Appearance: sensory characteristics of foods perceived largely by way of the visual sense. Input from other senses, especially smell, may contribute. • Flavour: the combination of taste and odour. Pain, heat, cold, tactile and visual sensations may also contribute. • Texture: sensory characteristics perceived largely by way of the senses of movement and touch. Input from other senses, especially vision and taste, may sometimes contribute.

6.3.2 Sensory test procedures The main sensory test procedures available to sensory analysts are shown schematically in Fig. 6.1. There is a fundamental distinction between the analytical methods, which use trained panels as an instrument to measure sensory properties, and the hedonic methods, which measure consumer responses to the sensory characteristics. Whilst the availability of such a wide range is of great value to sensory researchers, for practical reasons those used for QC purposes need to be chosen and adapted to the restrictions imposed within most manufacturing environments (See Chapter 4 and Muñoz et al., 1992; Costello, 2002). The methods most commonly used for QC purposes range from simple procedures that are easy to operate, but which have limited information content, through to complex systems with high information content, but which are expensive and time-consuming. A schematic diagram of the relationship between some key methods is shown in Figure 6.2. Whilst all these methods have their place in sensory QC, if they are to be complemented, or even replaced, by instrumental methods, then the more quantitative methods (especially quality ratings and profiling) are of the greatest importance if numerical relationships between instrumental



101

Sensory testing procedures

Analytical tests

Difference tests

Hedonic tests

Quantitative tests

Paired comparison

Simple descriptive

Duo–trio

Profiling

Triangle

Time–intensity

Preference Acceptability Relative-to-ideal

R-index

Fig. 6.1 A classification of sensory testing procedures.

• In/out method

Simple: low information content

• Quality ratings methods • Difference from control

• Quantitative profiling

Complex: high information content

Fig. 6.2 Key sensory test methods used in quality control.

and sensory data are to be established. The range of sensory techniques that are of potential value has been summarised by Hugi and Voirol (2001).

6.4 Instrumental measurement of quality factors 6.4.1 General principles It is possible to identify an enormous range of measurement variables that can be used to quantify aspects of product quality and product safety, in addition to those used to quantify factors such as production efficiency and compliance (Kress-Rogers, 2001a). Uncritical choice of the types of instrumental measurements to be used will generate substantial difficulties in data handling and interpretation. Further, as a consequence of the development of sophisticated software used for instrument control and data analysis, a given instrument will frequently generate a large number of numeric parameters. If confusion is to be avoided, and erroneous conclusions are to


102


be minimised, careful consideration must be given both to the choice of instrumental measurement and to the selection of measured parameters. Steps that should be taken in the use of instrumental measurements are outlined below. 1. Identify the sensory quality factors that the instrumental measurements are intended to mimic. 2. Select appropriate sensory methods for quantifying the sensory quality factors, and any necessary data manipulation and formatting. 3. Select appropriate instrumental test procedures and measured parameters, and any necessary data manipulation and formatting. 4. Correlate the instrumental measurements against sensory measurements using appropriate methods. 5. Validate any sensory–instrumental relationships developed for their value as a predictive tool. One important distinction between instrumental and sensory assessments is that the instrumental measurements comprise measurement of discrete, well-defined physicochemical properties, whereas sensory perception is rarely discrete, and different stimuli (within or across different sensory modalities) interact at both physiological and psychological levels. As a consequence, whilst an instrumental measurement of flavour components can usually be assumed to be free from the influence of other product characteristics, it has long been established that the sensory perception of flavour can be strongly influenced by product colour (for example see Zampini et al., 2007). One notable and important exception, however, lies in the measurement of flavour, in which case the physicochemical structure of the products can influence the release of flavour components, with profound effects on both perceived flavour and on the measured flavour components (for example, Taylor, 2002; Taylor and Roberts, 2004).

6.4.2 Appearance measurement For many food products, the visual senses are the first to be used by purchasers, consumers and trained sensory assessors. If a negative impression is communicated at point of sale, then purchase might not go ahead. More subtle influences can also influence the perception of non-visual sensory attributes through interactive mechanisms. Even if the product is packaged at point of sale and not directly visible, then visual information associated with the packaging system, including product images, product descriptions and ingredient lists will generate an expectation of product quality. Colour is usually regarded as the most important visual product characteristic (Francis and Clydesdale, 1975), and many instrumental systems have been developed for colour measurement, varying in sophistication from colour reference atlases to highly sensitive electronic instruments (detailed information can be found in, for example, MacDougall, 2001,



103

2002). The most common ones in practical use that are capable of generating quantitative data are based on the Hunter Lab system, in which colour is measured in terms of three parameters: L (lightness), a (red/green) and b (blue/yellow). Many foods carry important visual cues other than colour. For example, the glossiness of chocolate, turbidity of beverages and the visual composition of prepared meals can all influence consumer liking, and many systems have been developed for measuring this wider range of visual characteristics (Kress-Rogers and Brimelow, 2001). A more holistic view has also been taken in appearance assessment, through the concept of total appearance (Hutchings, 1999). In addition to direct visual information, the importance of indirect visual information about the product should also be considered, commonly available in printed form on packaging as photographic images, product labelling and ingredient information.

6.4.3 Texture measurement Texture perception is complex, with two major components: a tactile, surface response from skin (somesthesis) and a deep response from muscles and tendons (kinesthesis or proprioception) (Kilcast, 2004). In addition to perception in the mouth, manipulation of products by fingers and hands can generate textural responses, together with visual information (visual texture) and information arising from sounds released when handling and chewing products. As a consequence, many types of instrumental measurements have been devised to cover food categories (for more detailed information on measurement methods, see Rosenthal, 1999; Bourne, 2002; McKenna, 2003; Kilcast, 2004). The majority of these methods measure a wide range of mechanical characteristics of food which, although related to texture, do not give a complete picture of textural characteristics. As an example, the frictional properties of foods that are related to the perception of attributes such as roughness and creaminess have received relatively little attention, and measurement methods have not been developed to the extent of those used for other textural characteristics, but in recent years the importance of understanding such processes has been stressed (de Wijk and Prinz, 2006; Engelen and van der Bilt, 2008). Although the incomplete nature of instrumental texture measurement is widely recognised, the use of texture measurement in QC protocols is extensive, and this in part results from the greater degree of difficulty often encountered in using sensory panels for texture assessment in comparison with other sensory modalities. The types of texture measurement employed have been categorised as empirical, imitative and fundamental. Empirical methods measure often ill-defined variables that are indicated through practical experience to be related to some aspect of textural quality, and are frequently dedicated to a specific product type. Imitative methods mimic conditions that the product


104


is subjected to during eating. Fundamental methods measure well-defined physical properties of the product which can be independent of the measurement method. Both imitative and fundamental methods can usually be applied to a wide range of food types, and instrument manufacturers supply a wide range of test cells for this purpose. In general, empirical methods generate a single measurement parameter, whereas imitative and fundamental methods can generate a wide range of measurement parameters, some of which might be correlated. Selection of appropriate parameters is important if valid correlations with sensory data are to be achieved.

6.4.4 Flavour measurement Flavour is conventionally regarded as a combination of sensations derived from several distinct types of chemical stimuli. Tastes, detected by receptors on the tongue and other oral surfaces, are involatile chemical stimuli that are carried in solution by saliva from the food to the receptors. It is now widely accepted that there are five basic tastes – sweet, salt, bitter, acid, savoury (umami) – although this list is sometimes extended to include other sensations. Odours (aromas) are volatile chemical stimuli detected by receptors located in the olfactory epithelium in the nasal cavity. These are transmitted to the receptors directly through sniffing (orthonasal route), or from the mouth during eating (retronasal route). The odour response is complex, with around 2500 odorous chemicals found in food (Taylor and Roberts, 2004). A third component of flavour, a chemical sense that stimulates the trigeminal nerves, is responsible for sensations such as burning and cooling. Trigeminal sensations can arise from both chemicals in dissolved in saliva, for example the tingling sensation from carbonic acid in fizzy drinks, and from volatile chemicals, for example pungent thiocyanates in mustard and horseradish. Measurement of flavour components is consequently strongly influenced by the widely differing volatilities of flavour-active chemicals. The relatively large number of volatile chemicals contributing to flavour has been reflected in the wide range of instrumental methods that are now commonly used for volatile analysis, in particular gas chromatography/mass spectrometry systems (for example, see Kress-Rogers, 2001a). More recently, considerable publicity has been given to the development of so-called ‘electronic noses’, which are more correctly volatile sensors operating on a pattern recognition basis. Although these are finding numerous uses in other fields, relatively few routine uses have been recorded within the food industry (Röck et al., 2008). Measurement of taste chemicals has relied predominantly on traditional methods of chemical analysis for salty and acidic stimuli, with high-pressure liquid chromatography being used for less volatile chemicals such as sugars. However, ‘electronic tongues’ have now appeared on the market, and whilst considerable research is being carried out, very few practical applications have been reported.



105

6.4.5 Other measurements In addition to physicochemical measurements that can be related directly to the sensory quality of products, other measured data can be used to help build a model of likely sensory quality. This can take the form of data such as solution concentrations of components, pH, process temperature and emulsion droplet size. Depending on the type of data incorporated, care needs to be taken in the use of correlation methods: this is discussed further in Section 6.5.4. 6.4.6 Selection of instrumental measurement methods When developing instrumental–sensory relationships, careful consideration must be taken in both R&D and QC environments in selecting the instrumental methods to be used. Researchers often fall prey to the temptation to list all the instruments that might generate data relevant to sensory perception, leading to probable problems in data analysis and interpretation. On the other hand, QC methods are often those that are inexpensive, rapid and convenient, but which are not necessarily the most appropriate. In either situation, an additional danger is that methods will be selected on the basis of outdated information (or, worse, hearsay) regarding their relevance to perceived sensory characteristics, and correlations assumed rather than being checked and validated. In analytical investigations of aroma, Reineccius (2006) has stressed the importance of giving careful consideration to the sample, volatiles of interest, analysis time and study objectives in selecting analytical procedures, and has pointed out that analytical objectives such as those listed below will strongly influence the choice of procedures: • Obtain a complete aroma isolate to accurately identify and quantify all aroma constituents. • Identify key components responsible for the characteristic aroma. • Identify any off-notes. • Monitor aroma changes with time. • Predict sensory attributes. • Determine if a food flavouring is adulterated. The validity of all instrumental–sensory data correlations found in published or internal company literature should always be questioned, especially if product design factors such as ingredient composition, physical structure, processing conditions, storage conditions and packaging have changed substantially since the reported investigations.

6.5 Analysis and validation of instrumental measurements Section 6.2 listed some of the driving forces underlying the use of instrumental measurement of food quality. In a QC environment, the most


106


pressing requirement is to find instrumental methods that are rapid and inexpensive and which can reduce the dependence of the company on sensory panels (or even replace the use of sensory panels, although fortunately regulations in most developed countries recognise the importance of sensory quality assessments). One consequence is that all too often instrumental data are used uncritically, and several steps are needed to ensure that any instrumental measurement(s) used are valid.

6.5.1 Data inspection Most authorities on statistical data analysis stress the need to carry out appropriate visual inspection of numeric data before any statistical or correlation analysis is carried out. This applies not only to instrumental data, but also to sensory data used for correlation studies. The primary purpose of this stage is to check for any anomalies in the data that would compromise the quality of any data associations achieved. This could take the form, for example, of an instrument recalibration during an experiment, an uncorrected temperature change, or a simple transcription error. Inspection of small data sets in tabular form is feasible, and for many instruments, such as pH meters and empirical texture measuring instruments generating just a single-point measurement, this will give a good indication of data anomalies. Increasingly, however, multi-purpose instruments are used that carry out a continuous recording during a test, for example deformation-force measurements in texture assessment. Instrumental software will then often calculate a summary parameter that experimentation has shown to relate to sensory characteristics. Although convenient to the user, care should always be taken to inspect the form of the data recording to ensure that valid parameters are being measured. As an example, testing of gels and solid foods for firmness usually involves penetrating the product with a probe of defined geometry, and recording the force continuously during penetration. By convention, firmness is usually measured as the force recorded at a set penetration distance. Relatively minor changes to the gel structure and the probe geometry can result in distinctly different force–deformation curve shapes, as shown in Fig. 6.3. In the case of a simple brittle gel (e.g. gelatin) penetrated by a cylindrical probe, a break in the gel structure occurs at a short penetration distance, giving a discontinuity to the smooth curve. This has two consequences in practice. First, this initial break occurs at the 4 mm penetration distance conventionally used for gel firmness testing, and unwanted variability on this measurement. Secondly, this initial break results in chaotic breakdown patterns at higher penetration distances, and as a consequence high variability in any parameters measured at these penetration distances (Kilcast et al., 1984; Kilcast, 2001). This initial break does not occur when using a hemispherically ended probe, and firmness measurements associated with the simpler breakdown pattern show lower variability.


107

Force

Force


Initial break point

Penetration

Penetration

Brittle gel: cylindrical probe

Brittle gel: hemispherically ended probe

Fig. 6.3 Effect of penetration test probe geometry on gel breakage patterns.

Probe movement

Adhesive failure

Necking

Start

Necking

Cohesive failure

Fig. 6.4 Adhesive failure and cohesive failure in stickiness measurement.

A further example of the importance of visual observation of product behaviour during instrumental testing can be seen in research on devising instrumental measurements to measure stickiness in foods (Kilcast and Roberts, 1998; Kilcast, 2001). The perception of oral stickiness during sensory testing relates to the force needed to remove product from the teeth (ISO 5492, 1992). Instrumental testing of stickiness commonly uses a procedure in which the product is placed between two plates, is compressed, and then the force recorded as the plates are separated (Fig. 6.4). Perceived stickiness can then be related to the force when the product separates from the plate. In some situations, however, the product remains stuck to the plates and undergoes an internal failure, termed cohesive failure. Whilst this is an important characteristic in some contexts, for example in the sticking


108


of unwanted material to equipment surfaces, it is less likely to relate to perceived stickiness. As the behaviour of products during such tests can be influenced by a range of factors, such as product rheology, test conditions and the surface energy of the materials used for testing, observation can help to minimise the risk of misinterpretation.

6.5.2 Correlation analysis The most common objective in the use of instrumental relationships is to set up an empirical statistical model that relates the intensity of a sensory characteristic to a measured instrumental parameter, or to a set of instrumental parameters. Another relationship that is sometimes considered is to use instrumental data to directly model consumer liking. This requires, however, reliable consumer liking data that is relevant to the intended market, and an understanding of consumer segmentation patterns. A stepwise approach is therefore usually taken, first to relate instrumental data to key sensory attributes, and then to relate the sensory attributes to consumer liking (examples of different approaches to modelling consumer liking can be found in MacFie, 2007). An important prerequisite to carrying out any statistical analysis of instrumental data is to carry out a visual inspection of the data using scatter plots, usually by plotting one measure on the x-axis against a second measure on the y-axis. The visual form of the resulting plot will often give useful information on the data relationship, and guidance on further data analysis. In addition, the plots will often highlight problems with the data set. Examples of the form of plots that might be seen are shown in Fig. 6.5. Figure 6.5a shows a plot in which it is difficult to discern any structure in the data set, and a significant correlation is unlikely to be seen in such a plot. In Fig. 6.5b, there is sufficient indication of a possible linear relationship that would warrant further investigation. Evidence for a relationship can also be seen in Fig. 6.5c, but in this case the curvilinear form of this plot points to using non-linear modelling. The form of the plot in Fig. 6.5d is found occasionally, and indicates a possible change in the product structure (especially in texture testing) or some environmental factor such as temperature during the test. In this case (sometimes called a broken-stick model) two different linear relationships are evident, with the intersection occurring at the presumed change. If the scatter plots reveal possible linear relationships, then the next step is usually to calculate Pearson product moment correlation coefficients (r). A perfect positive correlation gives r = +1, a perfect negative correlation gives r = −1, and no correlation gives r = 0. However, it should be noted that these coefficients are relevant only to linear correlations, and strong data relationships can exist in which the correlation coefficient is very low. This is illustrated in Fig. 6.6. The scatter plot shown in Fig. 6.6a will give a correlation coefficient that is close to zero. The plots shown in Fig. 6.6b indicate



x

x x x x xx x x x xx x x

(a)

x

x

x x x x

x

x

x xx x x xxx x xx x x x x x xx

(c)

x

xx

(b)

109

xx x x x x x x x x x x x xx xx x x xx xx x

x xx

(d)

x xxx x xx x

x xx

Fig. 6.5 Different forms of scatter plots: (a) no relationship; (b) linear relationship; (c) curvilinear relationship; (d) broken-stick model.

near-perfect positive and negative correlations, and will give correlation coefficients close to +1 and −1, respectively. The scatter plot shown in Fig. 6.6c shows a very strong non-linear correlation, but which will give a very low correlation coefficient. (This inverted-U relationship is very commonly encountered in relationships between consumer liking and sensory attributes.) Figure 6.6d shows a situation in which this is a strong linear correlation, but an outlying point reduces the correlation coefficient. This situation often occurs through data transposition errors, and can also indicate a step change in a measurement. The square of the correlation coefficient (r2, or coefficient of determination) gives a measure of the data variance accounted for by the linear correlation. For example, a correlation coefficient of 0.7 indicates that 49% of the data variance is accounted for in the correlation. Statistical software packages commonly available will often associate a significance value to the correlation coefficient. Lawless and Heymann (1998) have described the use of the so-called Anscombe data sets in demonstrating the dangers of using correlation coefficients without first examining the form of the data.


110


x

r = +1

x

x x x x x x x xx x x x xx x x x x x x

(a)

r=0

o (b)

x

o o

x x

x

x

o ox x o

x

x

x

x

o

r = −1

o o

x

x

(c)

x

x x

xxxx xx x

x

x x

r≅0

x

x x

(d)

x x

x

x

x

x

x

r

Sensory Analysis for Food and Beverage Quality Control: A Practical Guide (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Food and Beverage Stability and Shelf Life (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Functional and Speciality Beverage Technology (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Processed Meats: Improving Safety, Nutrition and Quality (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Protective Cultures, Antimicrobial Metabolites and Bacteriophages for Food and Beverage Biopreservation (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Frying - Improving Quality (Woodhead Publishing in Food Science and Technology)

Food Preservation Techniques (Woodhead Publishing in Food Science and Technology)

Maillard Reactions in Chemistry, Food and Health (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Yeasts in Food: Beneficial and Detrimental Aspects (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Food Preservation Techniques (Woodhead Publishing in Food Science and Technology)

Food Colloids: Proteins, Lipids and Polysaccharides (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Food Emulsions and Foams (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Food, diet and obesity (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Endocrine Disrupting Chemicals in Food (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Meat Products Handbook: Practical Science and Technology (Woodhead Publishing in Food Science, Technology and Nutrition)

Determination of Veterinary Residues in Food (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Tracing Pathogens in The Food Chain (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Rice Quality: A guide to rice properties and analysis (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Environmentally-Compatible Food Packaging (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Handbook of Food Proteins (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Developing Children's Food Products (Woodhead Publishing Series in Food Science, Technology and Nutrition - Volume 204)

Cereals Processing Technology (Woodhead Publishing in Food Science and Technology)

Functional Foods: Principles and Technology (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Handbook of Indices of Food Quality and Authenticity (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Dairy-Derived Ingredients: Food and Nutraceutical Uses (Woodhead Publishing in Food Science, Technology and Nutrition)

The Maillard Reaction in Foods and Medicine (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Food Safety Control in the Poultry Industry (Woodhead Publishing in Food Science and Technology)

Handbook of Hygiene Control in the Food Industry (Woodhead Publishing in Food Science and Technology)

Phytochemical Functional Foods (Woodhead Publishing in Food Science and Technology)

Food Intolerance and the Food Industry (Woodhead Publishing in Food Science and Technology)

Sensory Analysis for Food and Beverage Quality Control: A Practical Guide (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Food and Beverage Stability and Shelf Life (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Functional and Speciality Beverage Technology (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Processed Meats: Improving Safety, Nutrition and Quality (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Protective Cultures, Antimicrobial Metabolites and Bacteriophages for Food and Beverage Biopreservation (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Frying - Improving Quality (Woodhead Publishing in Food Science and Technology)

Food Preservation Techniques (Woodhead Publishing in Food Science and Technology)

Maillard Reactions in Chemistry, Food and Health (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Yeasts in Food: Beneficial and Detrimental Aspects (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Food Preservation Techniques (Woodhead Publishing in Food Science and Technology)

Food Colloids: Proteins, Lipids and Polysaccharides (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Food Emulsions and Foams (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Food, diet and obesity (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Endocrine Disrupting Chemicals in Food (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Meat Products Handbook: Practical Science and Technology (Woodhead Publishing in Food Science, Technology and Nutrition)

Determination of Veterinary Residues in Food (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Tracing Pathogens in The Food Chain (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Rice Quality: A guide to rice properties and analysis (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Environmentally-Compatible Food Packaging (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Handbook of Food Proteins (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Developing Children's Food Products (Woodhead Publishing Series in Food Science, Technology and Nutrition - Volume 204)

Cereals Processing Technology (Woodhead Publishing in Food Science and Technology)

Functional Foods: Principles and Technology (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Handbook of Indices of Food Quality and Authenticity (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Dairy-Derived Ingredients: Food and Nutraceutical Uses (Woodhead Publishing in Food Science, Technology and Nutrition)

The Maillard Reaction in Foods and Medicine (Woodhead Publishing Series in Food Science, Technology and Nutrition)

Food Safety Control in the Poultry Industry (Woodhead Publishing in Food Science and Technology)

Handbook of Hygiene Control in the Food Industry (Woodhead Publishing in Food Science and Technology)

Phytochemical Functional Foods (Woodhead Publishing in Food Science and Technology)

Food Intolerance and the Food Industry (Woodhead Publishing in Food Science and Technology)

Recommend Documents