UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The influence of ration level and swimming speed on sensory attributes, gas chromatographic properties,… Siemens, Beverly Ruth 1997

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1997-0600.pdf [ 4.51MB ]
Metadata
JSON: 831-1.0087998.json
JSON-LD: 831-1.0087998-ld.json
RDF/XML (Pretty): 831-1.0087998-rdf.xml
RDF/JSON: 831-1.0087998-rdf.json
Turtle: 831-1.0087998-turtle.txt
N-Triples: 831-1.0087998-rdf-ntriples.txt
Original Record: 831-1.0087998-source.json
Full Text
831-1.0087998-fulltext.txt
Citation
831-1.0087998.ris

Full Text

T H E I N F L U E N C E OF R A T I O N L E V E L A N D S W I M M I N G SPEED O N S E N S O R Y A T T R I B U T E S , G A S C H R O M A T O G R A P H I C PROPERTIES, I N S T R O N T E X T U R E P R O F I L E A N A L Y S I S A N D P H OF C O O K E D M U S C L E F R O M F A R M E D C H I N O O K S A L M O N (Oncorhynchus tshawytscha) C U L T U R E D I N S E A W A T E R by Beverly Ruth Siemens B.Sc. The University o f Manitoba, 1990 A THESIS S U B M I T T E D I N P A R T I A L F U L F I L L M E N T O F T H E R E Q U I R E M E N T S F O R T H E D E G R E E OF M A S T E R OF SCIENCE in T H E F A C U L T Y OF G R A D U A T E STUDIES (Department o f Food Science) We accept this thesis as conforming to the required standard T H E U N I V E R S I T Y OF B R I T I S H C O L U M B I A December 1996 © Beverly Siemens, 1996 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. ... : Department .of PcT)c\ $)Cl<?riC<C The University of British Columbia Vancouver, Canada : . D a t e £3 Se^esnhor fflf. DE-6 (2/88) Abstract Representative samples o f post-juvenile Ch inook salmon were obtained from the Department o f Fisheries and Oceans Canada - West Vancouver Laboratory. The fish were part o f a study directed to assessing the influence o f t w o ration levels (75% and 100% o f maximum ration) and three swimming speeds (0.5, 1.0, and 1.5 body lengths/second) on growth, body composit ion and thyroid function o f chinook salmon in seawater. A l l analyses in the present study were conducted using cooked muscle. Sensory analysis, conducted in 19 sessions (10 days), was performed by 6 trained panellists. The treated and reference samples, composed o f randomly mixed slices o f muscle from farmed chinook salmon obtained at a local fishmonger, were graded for 28 sensory attributes; 9 aroma, 10 flavour, 8 texture as wel l as "overall acceptability". Af ter completing preliminary analyses o f the sensory data, data from panellist 6 and the first panel session were eliminated due to excessive inconsistencies in the results. A N O V A revealed that 8 attributes were significantly affected by rat ion level. Af ter standardising the significantly affected attributes' data, using a z-transformation to remove the panellist effect, one aroma term was no longer significant. Principal Component (PC) Similarity graphs using the standardised data clearly illustrated the effect o f ration level on these sensory attributes. The effect o f using a replacement panellist for panellist 5 o n t w o occasions became apparent from a PC 1 vs. PC 2 graph o f that panellist's data. Purge and trap extracts were used for gas liquid chromatographic analysis o f volatile compounds from cooked salmon. A n A N O V A o f consistently appearing peaks revealed that 27 o f these were significantly affected by either SS or R L . Principal Component Similarity graphs o f data from these peaks showed a clear separation o n the basis o f R L but not SS. The Instron Texture Profile Analysis statistics differed sharply from the other results since they indicated that SS and not R L significantly affected the texture o f the cooked salmon. The p H values ii for cooked fish were significantly affected by R L . The results o f this study, w i th the exception o f those from Instron TP A , agreed w i t h those o f Kiessling et al. (1994a,b) who generally found that R L and not SS significantly affected the g rowth and whole-body and muscle proximate composit ion o f chinook salmon in seawater. iii Table of Contents Abstract ii Tables viii Figures x Acknowledgements xi 1. Introduction 1 2. Literature Review 2 2.1 Exercise Leve l 2 2.1.1 Types o f exercise experiments 2 2.1.2 Effect o f exercise training on fish 3 2.1.3 Effect o f water current on fish behaviour 3 2.1.4 Fuel use for fish locomotion 5 2.1.5 Effects o f exercise training on fish 6 2.1.5.1 Increased stamina and maximum swimming speed 6 2.1.5.2 Hypertrophy o f fish muscle fibre 7 2.2 Physiological factors in fish affected by ra t ion level 9 2.2.1 The effect o f rat ion level o n fish size 9 2.2.2 The effect o f rat ion level on the fat content o f fish 10 2.2.3 The effect o f rat ion level o n the protein content o f f i s h 10 2.2.4 The effect o f rat ion level on fish muscle fibre size 10 2.2.5 The effect o f ration level on the muscle glycogen content o f fish 11 2.2.6 The effect o f rat ion level on fish g rowth hormone levels 11 2.3 T h e combined effect o f ra t ion level and s w i m m i n g speed on fish 12 2.4 Sensory test ing 12 2.4.1 Quantitative Descriptive Analysis ( Q D A ) 12 2.4.2 Unstructured line scale 14 2.4.3 The diff iculty in the analysis o f sensory data 14 2.5 T h e measurement o f food texture 14 2.5.1 Instron Universal Testing Machine 17 iv 2.5.2 Instrumental Texture Profile Analysis (TPA) 18 2.5.2.1 T P A on cooked salmon 18 2.6 T h e factors af fect ing the texture o f cooked fish 19 2.6.1 The effect o f p H o n the texture o f cooked fish 19 2.6.2 The effect o f muscle fibre size o n the texture o f cooked fish 20 2.6.3 The effect o f the level o f connective tissue on the texture o f cooked fish 20 2.6.4 The effect o f the fat content o n the texture o f cooked fish 21 2.7 Gas C h r o m a t o g r a p h i c flavour volat i le analysis 21 2.7.1 Purge and trap analysis 21 2.7.1.1 The effect o f fat content on purge and trap extractions 22 2.7.1.2 The choice o f Tenax GC, a porous polymer, for use in purge and trap extraction 23 2.7.1.3 The elution o f adsorbed volatiles from porous polymers w i t h ethyl ether 23 2.7.2 The relationship between Gas Chromatography (GC) data and quantity and intensity judgements from trained sensory panellists 24 2.8 Stat ist ical analysis 25 2.8.1 The box plot 25 2.8.2 Principal component similarity (PCS) 25 3. M a t e r i a l s a n d M e t h o d s 27 3.1 Exper imen ta l condi t ions used in the rear ing o f salmon used in th is s tudy 27 3.2 Sensory Analys is 29 3.2.1 The selection o f sensory panellists for Q D A analysis 29 3.2.2 Sensory panel training 30 3.2.3 The selection o f sensory attribute terms 30 3.2.4 Ballot familiarisation by sensory panellists 31 3.2.5 The use o f composite samples 32 3.2.6 Sensory panel set-up 32 3.2.7 Sensory panel session scheduling 33 3.2.8 The preparation o f samples for sensory panels 33 3.2.9 The reference samples used during sensory panels 35 3.2.10 The sampling procedure employed by panellists 35 3.2.11 Generation o f numerical scores from the sensory ballot judgements 36 3.3 I n s t r u m e n t a l analysis o f cooked salmon samples 36 3.3.1 Preparation o f salmon samples for instrumental analysis 36 3.3.2 G C headspace analysis o f cooked salmon samples 40 3.3.3 Instron T P A analysis o f cooked salmon samples 43 3.3.4 p H measurement o f cooked salmon samples 45 v 2.5.2 Instrumental Texture Profile Analysis ( T P A ) 19 2.5.2.1 T P A on cooked salmon 19 2.6 T h e factors af fect ing the tex ture o f cooked f ish 20 2.6.1 The effect o f p H on the texture o f cooked fish 20 2.6.2 The effect o f muscle fibre size on the texture o f cooked fish 21 2.6.3 The effect o f the level o f connective tissue o n the texture o f cooked fish 21 2.6.4 The effect o f the fat content on the texture o f cooked fish 22 2.7 Gas C h r o m a t o g r a p h i c flavour volat i le analysis 22 2.7.1 Purge and trap analysis 22 2.7.1.1 The effect o f fat content on purge and trap extractions 23 2.7.1.2 The choice o f Tenax G C , a porous polymer, for use in purge and trap extraction 24 2.7.1.3 The elution o f adsorbed volatiles from porous polymers w i th ethyl ether 25 2.7.2 The relationship between Gas Chromatography (GC) data and quantity and intensity judgements from trained sensory panellists 25 2.8 Stat ist ical analysis 26 2.8.1 The box plot 26 2.8.2 Principal component similarity (PCS) 26 3. Materials and Methods 28 3.1 Exper imen ta l condi t ions used in the rear ing o f salmon used in th is s tudy 28 3.2 Sensory Analys is 31 3.2.1 The selection o f sensory panellists for Q D A analysis 31 3.2.2 Sensory panel training 32 3.2.3 The selection o f sensory attribute terms 32 3.2.4 Ballot familiarisation by sensory panellists 33 3.2.5 The use o f composite samples 34 3.2.6 Sensory panel set-up 34 3.2.7 Sensory panel session scheduling 35 3.2.8 The preparation o f samples for sensory panels 35 3.2.9 The reference samples used during sensory panels 37 3.2.10 The sampling procedure employed by panellists 37 3.2.11 Generation o f numerical scores from the sensory ballot judgements 38 3.3 I n s t r u m e n t a l analysis o f cooked salmon samples 38 3.3.1 Preparation o f salmon samples for instrumental analysis 38 3.3.2 G C headspace analysis o f cooked salmon samples 43 3.3.3 Instron T P A analysis o f cooked salmon samples 47 3.3.4 p H measurement o f cooked salmon samples 49 v i 3.4 Da ta Analys is 51 3.4.1 Analysis o f sensory data 51 3.4.1.1 Exploratory analysis 51 3.4.1.2 Principal component analysis (PCA) 51 3.4.1.3 A N O V A ..52 3.4.1.4 Z-transformation o f significant sensory attribute scores 52 3.4.2 Calculation o f Instron T P A parameters 53 3.4.2.1 Calibration o f results 53 3.4.3 Calculation o f T P A 'Tirmness" 53 3.4.4 Calculation o f peak area 54 3.4.5 A N O V A o f Instron T P A and p H data 54 3.4.6 Principal Component Similarity (PCS) analysis o f Sensory, and G C headspace volatile data 54 4. Results and Discussion 57 4.1 Sensory analysis o f cooked salmon samples 57 4.1.1 Sensory panel reference samples 57 4.1.1.1 Purpose o f reference sample 57 4.1.1.2 Reference sample observations 59 4.1.2 Treated samples 59 4.1.2.1 Exploratory analysis 59 4.1.2.1.1 Boxplots 60 4.1.2.1.2 Delet ion o f unacceptable data 60 4.1.2.2 The use o f replacement panellists 63 4.1.2.3 Summary statistics o f treated samples 78 4.1.2.4 Three factor A N O V A o f sensory attribute data 79 4.1.2.5 Z-transformation o f sensory attributes significantly affected by the treatment 84 4.1.2.6 P C A and PCS o f sensory data 84 4.1.2.7 Effect o f SS on the sensory attributes 92 4.1.2.8 Effect o f R L on the sensory attributes 92 4.1.3 Instrumental analysis 93 4.1.4 G C headspace analysis 93 4.1.5 Instron T P A analysis 102 4.1.6 p H analysis 110 4.1.6.1 Comparison o f p H and sensory analysis results 113 5. Conclusions 115 5.1 Sensory Analyses 115 5.2 G C headspace analyses 116 vi i 5.3 Ins t ron T P A 116 5.4 p H 117 5.5 Overa l l Conclusions 117 References 119 Appendix A: Samples of sensory exploratory analysis boxplots 126 viii Tables Table 1 Comparative behaviour o f instruments and human subjects (Pangborn, 1987) 16 Tab le 2 Experimental design used by Kiessling et al (1994 a, b) to assess the influence o f sustained exercise and t w o ration levels on g rowth o f chinook salmon in seawater. A 2 X 3 factorial design was used w i t h t w o ration levels and three swimming speeds and their treatment numbers have been used as identifiers in statistical analyses. 30 Tab le 3 Conditions used in the extraction o f cooked salmon using a purge and trap procedure 46 Tab le 4 G C conditions used in the analysis o f purge and trap extracts from cooked chinook salmon samples 48 Tab le 5 Conditions used for Instron measurements o f minced cooked chinook salmon samples — 50 Tab le 6 Calculation o f Instron T P A parameters 55 Tab le 7 Range, mean, and standard deviation (St. Dev.) o f sensory attributes o f the cooked, farmed chinook salmon reference samples (5 panellists, 9 panel days) 58 Tab le 8 A N O V A results o f judge and panel day effect on reference samples for 28 sensory attributes o f cultured chinook salmon (5 panellists, 9 panel days) 62 Tab le 9 Range, mean, and standard deviation (St. Dev.) o f cooked, cultured chinook salmon sensory attributes (all treatments combined; 5 panellists, 9 panel days) 65 Tab le 10 Mean sensory scores and standard deviation o f cultured chinook salmon aroma attributes for each ration level X swimming speed treatment ( 5 panellists; 9 panel days) 66 Tab le 11 Range o f sensory scores o f cultured chinook salmon aroma attributes for each rat ion level X swimming speed treatment (5 panellists; 9 panel days) 67 Tab le 12 Mean sensory scores and standard deviation o f cultured chinook salmon flavour attributes for each ration level X swimming speed treatment (5 panellists; 9 panel days) 68 Tab le 13 Range o f sensory scores o f cultured chinook salmon flavour attributes for each rat ion level X swimming speed treatment (5 panellists; 9 panel days) 69 Table 14 Mean sensory scores and standard deviations o f cultured chinook salmon texture attributes for each ration level X swimming speed treatment (5 panellists; 9 panel days) 70 Tab le 15 Range o f sensory scores o f cultured chinook salmon texture attributes for each rat ion level X swimming speed treatment (5 panellists; 9 panel days) 71 Tab le 16 Summarised A N O V A results o f rat ion level, swimming speed and panellist effect o n sensory aroma attributes o f cultured chinook salmon (5 panellists, 9 panel days) — 81 Tab le 17 Summarised A N O V A results o f ration level, swimming speed and panellist effect o n sensory flavour attributes o f cultured chinook salmon (5 panellists, 9 panel days) 82 Table 18 Summarised A N O V A results o f ration level, swimming speed and panellist effect on sensory texture attributes o f cultured chinook salmon (5 panellists, 9 panel days) 83 Tab le 19 Summarised A N O V A results o f ration level and swimming speed effect on standardised data from significant sensory attributes o f cultured chinook salmon 85 Tab le 20 Factor score coefficients o f the first 6 principal components o f the z-transformed sensory attribute scores o f the R L X SS treated chinook salmon samples 86 Tab le 21 Peak labels, retention times and level o f significance for ration level and swimming speed o f G C peaks that consistently appeared in purge and trap G C headspace analysis o f cultured, cooked chinook salmon 94 ix Table 22 Factor score coefficients o f the first 7 principal components from G C headspace peaks significantly affected by either ration level or swimming speed 97 Table 23 Summarised A N O V A results o f ration level and swimming speed effect on Instron T P A parametersof cooked cultured chinook salmon 109 Tab le 24 A N O V A table o f ration level and swimming speed effect o n the p H o f cultured cooked chinook salmon 112 x Figures Figure 1 Sensory ballot that was used to evaluate samples o f cooked, farmed chinook salmon 40 F igure 2 PC 1 versus PC 2 using sensory aroma attribute scores from panellist 5, data points labelled w i t h treatment numbers 72 F igure 3 PC 1 versus PC 2 o f panellist 5 sensory flavour attribute scores, data points labelled w i t h treatment numbers 73 F igure 4 PC 1 versus PC 2 o f panellist 5 sensory texture attribute scores, data points labelled w i th treatment numbers 74 F igure 5 PC 1 versus PC 2 o f panellist 5 sensory aroma attribute scores, data points labelled w i t h panel day number 75 F igure 6 PC 1 versus PC 2 o f panellist 5 sensory flavour aroma attribute scores, data points labelled w i t h panel day numbers 76 F igure 7 PC 1 versus PC 2 o f panellist 5 sensory texture attribute scores, data points labelled w i t h panel day number 77 F igure 8 PCS graph o f significant sensory attributes, data points labelled w i t h treatment numbers— 88 F igure 9 PCS graph o f significant sensory attributes, data points labelled w i t h ration level numbers (1 = 75%; 2 = 100% ration level) 89 F igure 10 PCS graph o f significant sensory attributes, data points labelled w i t h swimming speed numbers (1=0.5,2=1.0, 3=1.5 bl/s) 90 F igure 11 PCS graph o f significant sensory attributes, data points labelled w i t h panel day number — 91 F igure 12 Sample gas chromatogram from a purge and trap extract o f cultured chinook salmon — 96 F igure 13 PCS graph o f significant G C peaks produced from a purge and trap headspace extraction o f cooked, cultured chinook salmon, data points labelled w i th treatment numbers 99 F igure 14 PCS graph o f significant G C peaks produced from a purge and trap headspace extraction o f cooked, cultured chinook salmon, data points labelled w i t h ration level number (1 = 75%, 2 = 100% ration level) 100 F igure 15 PCS graph o f significant G C peaks produced from a purge and trap headspace extraction o f cooked, cultured chinook salmon, data points labelled w i t h swimming speed number (1 = 0 .5 ,2 = 1.0,3 = 1.5 bl/s) 101 F igure 16 Boxplot o f the Instron T P A parameter Hardness 1 for cooked, cultured chinook salmon samples, results by treatment number 103 F igure 17 Boxplot o f the Instron T P A parameter Hardness 2 for cooked, cultured chinook salmon samples, results by treatment number 104 F igure 18 Boxplot o f the Instron T P A parameter Firmness 1 for cooked, cultured chinook salmon samples, results by treatment number 105 F igure 19 Boxp lo t o f the Instron T P A parameter Firmness 2 for cooked, cultured chinook salmon samples, results by treatment number 106 F igure 20 Boxplot o f the Instron T P A parameter Cohesiveness for cooked, cultured chinook salmon samples, results by treatment number 107 F igure 21 Boxplot o f the Instron T P A parameter Gumminess for cooked, cultured chinook salmon samples, results by treatment number 108 F igure 22 Boxplot by treatment number o f the p H o f t h e cooked chinook salmon samples 111 xi Acknowledgements I wish to thank the Department of Fisheries and Oceans for providing the fish for these experiments, my advisory committee for their patience and advice, and my parents and friends for their love and support. xii 1. I n t r o d u c t i o n Changes in fish that result from constant high levels o f swimming have long been observed. Recently, several researchers work ing on this topic, talking informally, agreed that fish kept in flowing water look subtly healthier, for instance their skin appeared shinier, flesh seemed firmer, also their eyes seemed to be brighter (Love, 1988). This is not a recent revelation. I n 1650, Venner observed this phenomenon and wrote ". . .of sea-fish the best swimmeth in a pure sea, and is tossed and hoist w i th the wind and surges; for by reason o f continual agitation it becometh o f purer and less slimey sic substances" (Love, 1988). Certainly, a clear-cut demonstration o f a favourable effect o f exercise (swimming speed) on one or more quality attributes o f the flesh o f farmed salmon wou ld be o f interest to the salmon farming industry. This is because the market values o f farmed salmon can approach their cost o f product ion and consequently any approach that raises market value or reduces product ion costs is o f importance for economic viability o f the industry. This thesis examines the effect o n the eating quality o f the cooked muscle o f cultured chinook salmon o f rearing these fish w i t h different combinations o f swimming speed and rat ion levels. Sensory analysis was used to quantitate the changes in aroma, flavour and texture attributes. Various forms o f instrumental analysis were employed to obtain objective measurements o f the treatment effects on the fish. Textural changes were quantified using the Instron Texture Profile Analysis and p H . The treatment effect on the flavour volatiles was quantified by gas liquid chromatographic analysis o f purge and trap extracts. 1 2. Literature Review 2.1 Exercise Leve l 2.1.1 Types o f exercise exper iments Three types o f exercise tests have been performed on fish. These include sprint, sustained svWmming and training. Sprint swimming, otherwise known as "Burst" swimming tests, are conducted by forcing the fish to sw im against very h igh water velocities for a short t ime; usually measured in seconds. Sustained swimming tests, which have been the subject o f the greatest amount o f research, involve forcing the fish to swim for several hours (Davison and Goldspink, 1978). The fish obtained from Fisheries and Oceans Canada, West Vancouver laboratory for my study had undergone sustained exercise training. I n experiments o f this type, the fish first go through a period o f training where they swim against increasing currents unti l reaching the desired water velocity. Then fish maintain this swimming velocity for the duration o f the experiment. The experimenter notes any adaptive change(s) that occur (Davison and Goldspink, 1978). Most o f the research involving this type o f exercise training in fish has involved salmonids i.e. salmon, t rout and charr. There are several reasons for this. First salmonids are reared commercially, they are easy to obtain and generally respond wel l to captivity. Second, their behaviour is predictable unlike many other species (Davison, 1989). Third, salmonids respond readily to training under artificial conditions (Davison and Goldspink, 1977; Greer Walker and Emerson, 1978). Finally, since salmonids are a commercially important group o f fish, research funding is more readily available than for other types o f fish (Davison, 1989). 2 2.1.2 Effect o f exercise t r a i n i n g on fish Fish do not respond to exercise training in the same way as mammals. Changes in fish do occur as a result o f training, but they are comparatively modest. Training has been shown to affect fish g rowth rates as wel l as their ability to swim (reviewed by Davison, 1989). Unfortunately, as noted by both Davison (1989) and Broughton and Goldspink (1978), due to their l imited number, comparisons between studies are difficult. There are several factors that contribute to this problem. Some o f the diff iculty is due to differences in the studies w i t h regards to the species, size, sex and life history o f the fish (Broughton and Goldspink, 1978). There is also a diverse array o f apparatus that has been employed. Other areas o f disparity include the dissimilar durations and intensities o f exercise between studies and in many studies there have been differences in the types o f tests that have been used to detect changes. A s a result, there is a great deal o f variability in the results, and conflicting data are often presented (Davison, 1989). 2.1.3 Effect o f w a t e r cu r ren t on fish behav iour I n many previous studies (Davison and Goldspink, 1977; Davison and Goldspink, 1978; Greer Walker and Emerson, 1978; East and Magnan, 1987; Houlihan and Laurent, 1987; Christiansen et al., 1989) the control fish have been held in calm water in an attempt to reduce the amount o f energy required for l ocomot ioa This protocol , however, has led to behavioural problems, such as increased aggressive responses as manifested by biting and fin-nipping. Christiansen and Jobling (1990) found that fish wi thout bite marks grew significantly better than those w i t h evidence o f bite marks. Elevated plasma Cortisol levels which are k n o w n to suppress g rowth (Pickering, 1990) and feed efficiency (Vijayan and Leatherland 1989), have also been found. I n contrast, salmonids reared in water w i t h a 3 current, school, swim less randomly, and they also exhibit less aggressive behaviour (Christiansen et al. 1989; Christiansen and Jobling 1990) and have lower levels o f plasma Cortisol adrenaline, and noradrenaline (Woodward and Smith, 1985). This leads to questions whether the improved growth, noted in the studies w i t h the still water controls, was a result o f the exercise per se or was due to the increased energy demands and stress that accompany aggression in the control fish (Kiessling et al., 1994b). Davison and Goldspink (1977, 1978) conducted t w o experiments that address this issue. These researchers studied the effect o f different levels o f exercise training on the growth o f b rown trout, a member o f the salmonid family, and goldfish, a fish normally found in still water. I n both o f these experiments the control fish were held in still water. The results o f these two experiments were dramatically different. I n the trout experiment, fish at the lowest swimming speed grew much more rapidly than the control fish. Moreover, they had large stores o f glycogen and lipids and other physiological changes. The fish at the medium swimming speed had decreased food utilisation due to increased energy demands. Many fish at the highest swimming speed were unable to survive. B y contrast, in the goldfish study the fish subjected to exercise often grew less than the controls and they also consumed substantially more food. Surprisingly, most o f the goldfish survived at the highest swimming speed. These preceding results led the researchers to speculate that once the fish acclimates to a higher exercise rate, elevated levels o f anabolic hormones such as thyroxine are produced. They theorised that the goldfish were unable to acclimate to the flowing water which resulted in the fish having elevated levels o f stress-related hormones such as adrenaline, noradrenaline, and Cortisol. 4 However, they never considered the possibility that the trout control may have also been under stress. Hence, they drew some incorrect conclusions from their work . 2.1.4 Fue l use fo r f ish locomot ion The fish's use o f fuels for locomotion can be quite complicated; protein, lipids as wel l as glycogen may be used as sources o f energy (reviewed by Davison, 1989). I n fish, red muscle has an aerobic fo rm o f metabolism and is used for normal locomotion while the white muscle is used for burst (sprint) swimming, or when the fish is involved in strenuous exercise. Whi te muscle uses glycolysis or the anaerobic degradation o f glucose to yield lactic acid and energy. The overall rise in lactate is accompanied by a fall in glycogen concentration fo l lowing exercise (reviewed by Broughton and Goldspink, 1978). The findings o f Johnston and M o o n (1980a, b) suggest that training produces a shift towards fat utilisation rather than use o f glucose as an energy source. Davison and Goldspink (1977) had similar findings. They observed that b rown t rout when exercised at a slower swimming speed (1.5 body lengths / second (bl/s)) had elevated levels o f both glycogen and lipids. The trout at the intermediate speed (3 bl/s) still had elevated glycogen levels, but the lipid levels had fallen. This suggests that lipids were the major source o f fuel for the fish at that swimming speed. White and L i (1985) also found that lipids were the primary source o f fuel during training. They found that chinook salmon at all speed by ration level combinations, except the slowest swimming speed by highest rat ion leve l experienced a net decrease in body fat. The fish exercised at 2, 3 and 4 bl/s, at both the 2.5 and 6 % ration levels (% dry body weight/day), exhibited greater decreases in their percentages o f fat than noted for the unfed fish ( 0 % dry body weight/day). Alternatively, 5 Kiessling et al. (1994b) found that the fat contents in the fillet and whole bodies o f chinook salmon were not significantly affected by changing the swimming speed over the range o f 0.5-1.5 bl/s. Several studies indicate that protein may also serve as a source o f fuel for fish in training experiments. I n the Davison and Goldspink (1977) study on trout, the protein content o f the fish decreased as swimming speed increased. This suggested that protein might have been utilised as a fuel source. East and Magnan (1987) obtained similar findings w i t h brook charr. I n another study, White and L i (1985) observed that chinook salmon forced to swim for 10 days wi thout food, had decreased protein levels while their l ipid levels were unaffected. 2.1.5 Effects of exercise training on fish 2.1.5.1 Increased stamina and maximum swimming speed Fish, like athletes, must be subjected to a period o f conditioning before the ful l expression o f their capacity for swimming is realised (Farlinger and Beamish, 1978). I t has been shown that training generally increases stamina and aerobic capacity (Hochachka, 1961; Farlinger and Beamish, 1978; Broughton et al. 1980) as wel l as maximum swimming speed (Davison and Goldspink, 1977). Hammond and Hickman (1966) showed that conditioned fish can not only tolerate higher levels o f blood lactate, but they can also remove lactate from the blood more quickly than unconditioned fish. B o t h total accumulations o f muscle lactate during exercise and its subsequent rate o f removal during recovery were found to vary directly w i th the degree o f physical conditioning. Hochachka (1961) demonstrated that trained fish could acquire an oxygen debt that was three times higher than that o f untrained fish before becoming fatigued. He postulated that the increased ability o f trained fish to resist fatigue was due to increased buffering capacity. This hypotheses was 6 based on the observation that the trained fish had higher haemoglobin levels. W i th respect to this, haemoglobin has two functions, first as a carrier for oxygen, and second as a buffer. Hochachka theorised that in t rout, the primary function o f the haemoglobin may be to act as a buffer, and the respiratory function might be secondary in nature during resting metabol ism However, during more extreme conditions, e.g., higher exercise levels, the respiratory function may become more important. Under these conditions the increased oxygen carrying capacity wou ld be invaluable. Love (1988) also postulated that other buffers may be present in larger quantities in trained fish. I n addition, Love (1988) speculated that i f buffers, such as anserine, were present in greater than normal quantities, there could also be an effect on the flavour o f the fish. 2.1.5.2 Hypertrophy of fish muscle fibre When a fish swims for long periods o f time, there are often significant changes in the number and diameter o f red and white muscle fibres, which together result in increased muscle size (reviewed by Davison, 1989). Unl ike mammals where the fibre number in the body remains constant, the fibre number in many species o f fish increases during their lifetime, paralleling the increases in body size (Weatherly and Gil l , 1981, 1984). Davison and Goldspink (1977) noted that the initial size o f the muscle fibre in the fish w i l l determine whether the fibre splits or enlarges. For instance, when they exercised fish w i t h small fibres, there was an increase in fibre size. Conversely, when fish w i t h large muscle fibres were exercised, there were concomitant increases in muscle mass that were due to an increase in fibre number. Similarly, Patterson and Goldspink (1976), work ing w i t h saithe, observed that the muscle fibres split longitudinally once they reached a diameter o f 1.2 micrometers. 7 Greer Walker and Pull (1973) observed that the extent o f fibre hypertrophy in coalfish varied w i th swimming speed. Also, they noted that different muscle types became active as the swimming speed was changed. I n this regard, Johnston et al. (1977), work ing w i t h carp, noted that red muscle fibres were used predominately at lower swimming speeds. However, as the speed was increased, the red fibres became progressively less important, whereas the pink, and then the white muscle fibres became increasingly more active. Collectively, the studies on fish suggest that white muscle is used primarily when the speed is near or above the threshold for sustained speed. A t lower speeds both the red and white muscles are used (Greer Walker and Pull, 1973). Greer Walker and Emerson (1978), in an experiment involving rainbow trout found that oxidative metabolism occurred predominantly in the red muscle at swimming speeds up to 1.4 bl/s. Beyond this, the white fibres became increasingly active. I n another experiment, Johnston and M o o n (1980a) subjected brook trout to a water current o f 1 bl/s and noted that electrical activity occurred only in the red muscle. A t water speeds above 1.8 bl/s, however, electrical activity was exclusively observed in the white muscle . Changes in muscle fibre number and hypertrophy provide good indications o f the different muscle types that are active at a given swimming speed (Greer Walker and Pull, 1973). Generally, red muscle fibres hypertroph to a greater extent than white fibres at any given swimming speed (Greer Walker and PulL 1973; Davison and Goldspink, 1977). Kiessling et al. (1994b) found that the red muscle o f chinook salmon trained at 1.5 bl/s showed significant fibre hypertrophy (25%) in the rostral region o f the fish. Also, they observed in another experiment that fish that swam at their critical swimming velocity every other day for 120 days had doubled the total red muscle area in their caudal region. Kiessling et al. (1994b) speculated that the dissimilarity in red muscle area between rostral and caudal regions was probably due to the differences in swimming patterns between the t w o experiments. 8 I n situations where swimming speed is excessive, muscle fibre hypertrophy may only be evident to a small extent on not at all. For example Greer Walker and Emerson (1978) observed that red muscle fibre hypertrophy decreased when t rout were forced to sw im between 2 and 3 bl/s. They speculated that the reason for this response was due to the fibres contracting above their opt imal frequency. Other experiments have not shown any significant increase in muscle mass in relation to swimming speed. Davie et al. (1986), for example, did not find any alteration in the ratio o f total muscle to tota l body weight o f t rout that had been forced to sw im continuously at a l ow svsdmming speed for 200 days. They did, however, observe an increase in the proport ion o f red fibre muscle. 2.2 Physiological factors in fish affected by ration level 2.2.1 The effect of ration level on fish size Level o f dietary intake can have a profound effect o n numerous aspects o f fish physiology. N o t unexpectedly, an increase or decrease in ration level can significantly affect fish size (weight and length). Kiessling et al. (1991a) found significant differences in the weights o f 1 - 2 year o ld rainbow trout maintained on rat ion levels o f 50, 75 and 100% (100% ration level defined as the ration level required for opt imum growth). Numerous other studies on various fish species have also shown that g rowth is positively correlated w i t h increased ration level unt i l feed intake becomes excessive (Kiessling et al., 1989a; Storebakken et al., 1991; L i and Lovel l 1992). Kiessling et al. (1994b) also found that chinook salmon weight increased significantly as rat ion level was increased from 7 5 % to 100% o f maximum. 9 2.2.2 The effect of ration level on the fat content of fish Fish maintained on higher ration levels often have increased body fat content. Kiessling et al. (1994b) found that the fat content o f chinook salmon on maximum ration was significantly higher than noted for those given 7 5 % o f maximum ration. Storebakken et al. (1991) and Kiessling et al. (1991a) also showed a trend towards increased fat accretion in trout as ration level was increased. I n contrast, Kiessling et al.(1989a) did not find any significant effect o f fixed rations (25%,50% and 100%) on total fat levels in either the white or the red muscle tissue o f trout. They did find, however, that the fat content in the dorsal muscle fat depot rose in direct relation to ration level. 2.2.3 The effect of ration level on the protein content of fish Changes in ration level generally alter the absolute but not the relative (%) amounts o f protein in fish (Kiessling et al. 1991b; Storebakken et al., 1991). For instance, Kiessling et al. (1991b), noted that the increase in protein content o f rainbow trout epaxial muscle was independent o f ration level once the fish had reached 50g. They also found that a profound drop in muscle protein content only took place in mature fish, and that this was only in response to decreased feed availability in combination w i t h prolonged physical activity. 2.2.4 The effect of ration level on fish muscle fibre size Fibre size has been found to be highly correlated w i t h fish size (Kiessling et al. 1989b; 1991a) which, in turn, is significantly affected by rat ion level. I n the study performed by Kiessling et al. 10 (1989b), significant changes were seen in both white and red muscle fibre areas. The white muscle fibre area was positively correlated w i t h slaughter weight. Kiessling et al. (1989b) concluded from this observation that the pattern o f g rowth was based mainly on enlargement o f single fibres rather than on increases in fibre number. There was also a shift toward dominance o f fibres w i t h larger areas as fish weight increased. They went on to speculate that fibres also grew in length as rat ion level was increased. 2.2.5 The effect of ration level on the muscle glycogen content offish Glycogen content appears to be affected by ration level, but this relationship is less clear. Kiessling et al. (1989b), for example found that the glycogen levels in both red and white muscle o f rainbow trout increased as ration level was varied between 50 and 100%. I n the same experiment, however, the glycogen content o f the white muscle from trout on the 2 5 % ration level was observed to be significantly higher than noted for trout on the 100% ration. I n an experiment performed by Storebakken et al. (1991), b lood glucose levels were found to increase between the 0 - 1 % rat ion levels but then dropped at the 2 % ration level. Finally, the data o f Kiessling's et al. (1991b) did not reveal any effect o f rat ion level on the glycogen content o f trout. 2.2.6 The effect of ration level on fish growth hormone levels G r o w t h hormone levels tend to decrease as ration levels increase. Storebakken et al. (1991) found a 3.5-4.0 fo ld reduction in the circulating concentrations o f g rowth hormone in t rout maintained on a ration o f 2 % o f body weight relative to those deprived o f feed. Elevations in g rowth hormone levels in fish maintained on reduced levels o f dietary intake have been linked to depressed fat content. 11 Indeed, it is generally accepted that there is an inverse relationship between growth hormone level and fat deposition in fish (Storebakken et al. 1991). 2.3 T h e combined effect o f ra t ion level and s w i m m i n g speed on fish Several studies have evaluated the combined effect of ration level and swimming speed on fish growth and other aspects of performance. One of the most recent of these was conducted by Kiessling et al. (1994b), and fish from this study were used for this thesis. White and Li (1985) and Leon (1986), also conducted similar projects on this theme using juvenile chinook salmon and brook trout, respectively. White and Li (1985) found that nearly all of the variation associated with growth could be accounted for by differences in level of dietary intake. In addition to the energy required for standard metabolism, which is constant at all swimming speeds, the amount of energy required for activity (swimming) increased steadily as swimming speed was raised. As a result, the fish had to ingest more feed (energy) at the higher swimming speeds to maintain their body weight. Kiessling et al. (1994b) arrived at the same conclusion. 2.4 Sensory test ing 2.4.1 Quan t i t a t i ve Descr ipt ive Analys is ( Q D A ) Traditionally, each food manufacturing company employed one or a group of experts to perform sensory analysis. These expert(s) judgements were relied upon in all facets of production, from the choice of ingredients, to the choice of which products were ready for release into the 12 marketplace. I n recent years, our society has become increasingly multicultural. Consequently, the markets have become much more complex and competitive. The judgements o f the expert are still useful, but there are limits to one person's ability (Stone et al. 1974). A round 1949, the Ar thur D. Li t t le Co. proposed the Flavour Profile Method ( F P M ) as a means o f dealing w i th the complex wor ld o f food flavours (Anon. 1963). I n this method, a small group o f judges evaluates the product together in a conference style meeting. Before testing, the judges undergo some descriptive training. T o accomplish this, a broad selection o f references is presented to the judges to prepare them for subsequent evaluation o f the intensity o f f lavour and aroma attributes in one or more test samples. F P M allows, in many cases, the successful replacement o f the individual expert by the expertise o f the group (Stone et al., 1974). Quantitative Descriptive Analysis ( Q D A ) , developed by Tragon Corp., has brought additional improvements to sensory evaluation (Stone et al., 1974). I n Q D A , trained individuals identify and quantify the sensory properties o f a product or an ingredient (Stone et al., 1974). This method uses an interval scale w i t h anchor points located one hal f inch from each end. The panellist places a vertical mark at the point that she/he feels best represents the magnitude o f the intensity o f the attribute (Stone et al., 1974). Use o f Q D A makes it possible to statistically analyse data. For example, a researcher can perform a one or t w o way analysis o f variance to analyse individual and group performance. He/She could also use principal component analysis (PCA) to determine which are the primary sensory variables and then redundant terms could be identified and removed (Rutledge and Hudson, 1990; Stone eta l . , 1974). 13 2.4.2 U n s t r u c t u r e d l ine scale Using an interval scale can lead to problems due to the panellist's diff iculty in attributing the same psychological w id th to the various intervals on the scale. This problem can be alleviated to a large extent by using unstructured graphic scales that consist o f a 10 c m horizontal line anchored at the ends. The panellist response is indicated by a vertical mark o n the line (Giovanni and Pangborn, 1983). Unfortunately, as Gacula (1987) has pointed out, panellists have a tendency to underestimate the score at the lower end and overestimate the score at the higher end o f the unstructured scale. 2.4.3 T h e d i f f i cu l ty in the analysis o f sensory da ta There are many differences between sensory and instrumental analyses (Table 1) that can create problems in data analysis. Gacula (1987) found that sensory data were among the most difficult scientific data to statistically analyse and interpret since there are often untested assumptions about the data and the analysis procedure. Some o f the difficulties are as fol lows. First, panellists tend to use the scale differently. The data are relative, easily skewed, and very difficult to replicate (Gacula, 1987). Third, panellists are prone to fatigue, time-order effects, and subject to drifts (Pangborn, 1987). Even when panellists have been screened and wel l trained, there is still the chance that a panellist by treatment interaction w i l l stem from differences in motivation, sensitivity or psychophysical response behaviour (Lundahl and McDaniel , 1990). 2.5 T h e measurement o f food tex ture Szczesniak (1963) defined texture as "the sensory manifestation o f the structure o f the food and the manner in which this structure reacts to the applied forces and specific senses involved being 14 Table 1 Comparative behavior (Pangborn, 1987) Instrumental Sensory Separator Univariate Absolute Fast Calibratable Precise Doesn't Fatigue No time-order effects Equal-interval Units Expensive to purchase and maintain Cannot measure hedonics Cannot mimic sensory Integrator Multivariate Relative Slow Difficult to calibrate Subject to drift Fatigues, Adapts Time-order effects Unequal-interval units Expensive to hire judges Biased by hedonics Artificial to mimic instrumental 15 vision, Anaesthetics and hearing." I n 1990 she simplified this definition to "how the food feels in the mouth on manipulation and mastication, and h o w it handles during transport, preparation, and o n the plate" (Szczesniak, 1990). Texture for many years has been considered by some to be an overlooked food attribute. There are several reasons for this. First, there has been a lack o f government funding for research into food texture. A second problem has been that off-texture is not a signal that a food is unsafe, unlike attributes such as smell, colour and taste. Finally, changes in texture are often more difficult to accomplish and often affect other quality parameters such as taste; these changes can not just be "added from a bottle" whereas those related to aroma and flavour can (Szczesniak, 1990). Attempts have been made to measure texture quantitatively since the 1860's. According to Szczesniak (1990), the first texture measuring device was developed in Germany by L ipowi tz in 1861 which was an instrument, designed to quantify the consistency o f jelly. Since then, other instruments, designed to measure textural qualities o f various types o f food, have been developed and these have evolved into the instruments used today. Some examples o f texture measuring devices currently in use include, the Instron Universal Testing Machine, the General Foods Texturometer, and the Brabender Farinograph. According to Szczesniak (1963), textural measurements can be grouped into three types o f characteristics: (1) mechanical, (2) geometrical and (3) other characteristics. Mechanical characteristics result from pressure exerted on the teeth, tongue and r o o f o f the mouth during eating. These characteristics include the hardness, cohesiveness, viscosity, elasticity and adhesiveness, etc. o f the food. Geometrical characteristics are related to the size, shape, and arrangement o f the particles 16 with in a food (Brandt et al., 1963). The last group is 'other' characteristics which includes mainly moisture and fat; qualities concerned w i t h lubricating properties o f the food product. Brandt et al. (1963) developed the Texture Profile Method (TPA) , patterning it after the flavour profi le method developed by Cairncross and Sjostrom (1950). They used the standard rating scales developed by Szczesniak (1963), and systematically examined various textural attributes, breaking them d o w n into initial (textural attributes perceived on the first bite), masticatory (perceived during chewing), and residual characteristics (changes that occur during mastication). I n T P A , additional scales can be added to enable the judgement o f moisture and fat content (Brandt et al. 1963). T P A requires that the panellists be trained thoroughly w i t h respect to the texture classification system and the use o f standard evaluation procedures for assessment o f the product. Panellists must also become reliable in recognising and identifying the degrees o f each characteristic (Brandt et al., 1963). 2.5.1 I ns t ron Un iversa l Test ing M a c h i n e The Instron is an instrument designed to study stress-strain properties o f materials (Bourne, 1982). I n addition to food, it can also be used to study texture in other materials such as fabric, metals, w o o d , rubber, plastics, etc. (Bourne, 1982). W i t h an assortment o f accessories available (Bourne, 1982), this machine can perform various types o f tests such as penetration, shear, bending, compression, and extension (Segars and Kapsalis, 1987). The Instron generates both force-time and force-distance curves, al lowing w o r k function to be calculated in pounds, ki logram, or Newton 's (Bourne, 1978). The curve(s) can become the basis for calculating various mechanical properties o f the material. These values may be used to correlate or predict sensory response to texture (Segars and Kapsalis, 1987). 17 2.5.2 I n s t r u m e n t a l Tex tu re Prof i le Analys is ( T P A ) T P A was a major breakthrough in the quest to produce a machine that could imitate mastication. The General Foods Texturometer attempts to imitate mastication by twice compressing a bite sized piece o f food to 2 5 % o f its original height; mimicking a person taking t w o bites. F rom this, a force-time curve is produced which captures the entire force history o f this simulated masticatory action. Several textural parameters can be determined from the force-time curve. Bourne (1978) described seven parameters; five measured and t w o calculated. The measured parameters include fracturability, hardness, cohesiveness, adhesiveness, and springiness. The t w o calculated parameters are gumminess and chewiness (Bourne, 1978). Firmness may also be calculated from the curve by measuring the maximum slope on each compression cycle (Durance and Collins, 1991). 2.5.2.1 T P A o n cooked salmon I t is difficult to obtain meaningful, reproducible instrumental texture measurements on cooked fish. Mos t o f the devices that are commonly used in the rheological testing o f foods, even those that are used for red meat, are generally unsuitable for fish. Dur ing eating, almost all o f the energy required to prepare the fish for swallowing is used for mastication. As a result, instrumental testing o f fish samples needs to measure resistance o f the muscle fibres to mechanical disintegration (Dunajski, 1979). Durance and Collins (1991), and Reid and Durance (1992) examined textural changes o f canned late run salmon by using Bourne's (1978) T P A and an Instron Universal Testing Machine (Mode l 1122, Instron Corp. Canton M A ) . I n their experiments, a modif ied syringe was used to f o r m cylinders o f flaked fish o f un i form size. Using samples composed o f thoroughly flaked fish, they managed to gain a greater degree o f homogeneity between replicates. Borderias et al. (1983) tested 18 both minced fish and intact fillets using various Instron attachments and found a higher coefficient o f variation for the fillets relative to the minced fish. They speculated that this occurred because when the force o f compression was applied to the fish fillets, the myotome layers slid away from the force. Hence, it was more difficult to obtain reproducible results in separate detenninations. 2.6 T h e factors af fect ing the tex ture o f cooked fish 2.6.1 T h e effect o f p H on the tex ture o f cooked fish Love (1988) and Dunajski (1979) both stated that the p H o f fish muscle is probably the most important factor affecting the rheological properties o f a given muscle. Love (1988) postulated that muscle from exercised fish wou ld have a lower post-mortem p H due to an elevated glycogen content. He went o n to theorise this wou ld lead to firmer muscle texture. Feinstein and B u c k (1984) found a linear relationship between p H and the texture o f flounder but only in the head section o f the fish. They also looked for a relationship between p H and texture in cusk wi thout success. As w i t h most animals, after the death o f a fish, glycogen is degraded to lactic acid via the Embden-Meyerhof-Parnas glycolytic cycle. This causes the p H in the muscle to fall dramatically wi th in the first several hours post-mortem. I n most species o f fish, the final p H is usually around 6.5 - 6.2, but is also can be as low as 5.4. As the p H approaches the isoelectric point o f the myofibril lar proteins, there is a change in the ionisation o f the polar groups o f the protein molecules. Originally these were negatively charged but after death they become neutral. This causes a decrease in the repelling forces between proteins, resulting in a tightening in the protein structure. A s the myofibril lar proteins become more concentrated, the muscle becomes increasingly tougher and drier. Dunajski (unpublished), for 19 example, noted that there was a 2.5 fold increase in the toughness o f the fish as the p H changed from 6.7-5.7(Dunajski, 1979). 2.6.2 T h e effect o f muscle f ib re size on the texture o f cooked fish Fibre size has also been found to affect the texture o f fish muscle. Kanoh et al. (1988), in an experiment using yel lowfin tuna, found that the fish had a firmer texture when the fibre diameter was less than that o f ordinary muscle. Hatae et al. (1990), drew the same conclusion after examining the role o f muscle fibre contribution to firmness in the cooked flesh o f five species o f fish. Dunajski (1979) reported an increase in the coarseness o f the muscles when the diameter and length o f fibres increased. 2.6.3 T h e effect o f the level o f connective tissue on the tex ture o f cooked fish Unl ike red meat, connective tissue in fish muscle is present in l ow quantities and hence does not play an important role in the texture o f fish. Collagen is thermally denatured during cooking and as a result generally has very little influence on fish texture. The texture o f muscle after cooking is more a consequence o f the state o f the myofibril lar proteins (Dunajski, 1979). Hatae et al. (1990) did, however, report an effect o f muscle collagen content on texture. I n this regard, they observed that when cooked fish tissue was masticated the coagulated proteins tended to impede the sliding o f the muscle fibres over each other. F r o m this they concluded that, in fish w i t h a lower muscle collagen content, the muscle fibres slide more easily over one another, resulting in a softer texture. 20 2.6.4 T h e effect o f the fat content on the texture o f cooked fish Fat content can also affect the texture o f fish samples. Samples o f fish muscle w i th a higher fat content are often perceived as being more tender (Dunajski, 1979). Dunajski (1979) explained that, among other post-mortem changes in fish, the l iquid neutral lipids are immobilised by the physical structure o f the muscle. This tends to dilute the structural elements and decrease the overall mechanical strength o f the fish meat. A higher fat content w i l l also impart an oi ly mouthfeel (Szczesniak, 1963). 2.7 Gas C h r o m a t o g r a p h i c f l avour volat i le analysis 2.7.1 Purge a n d t r a p analysis The principle behind purge and trap extraction is quite simple. First, the sample is placed in a sealed container that is flushed w i t h an inert gas. This gas then passes through a trap containing a small amount o f adsorbent, such as tenax, wh ich retains the volatiles. Fol lowing the extraction, the trap is removed and adsorbed compounds are eluted w i t h a small amount o f solvent, and a sample o f this is then analysed by gas liquid chromatography (GC) (Gilbert, 1990). Heikes and Hopper (1986) outlined several advantages o f this method. First, it is non labour-intensive and may be carried out unattended. Furthermore, it does not require highly specialised equipment ( though n o w available); a suitable apparatus can be constructed quite simply w i t h materials already available in most analytical laboratories. The extract is concentrated and relatively clean. The limits o f detection that can be achieved are much lower relative to solvent extraction or static headspace analysis; both G C quantitation at l o w parts per bil l ion and sub-parts per bi l l ion levels as wel l 21 as G C / M S corifirmation are possible. Generally, this technique provides an inexpensive alternative to other methods (Olafsdottir et al., 1985). There are conflicting opinions as to who originally developed purge and trap extraction. According to Gilbert (1990), this method was developed by Heikes in 1985 for the analysis o f ethylene dibromide in grains. However, it is noteworthy that Josephson and co-workers published an experiment in 1983 that described the use o f purge and trap extraction to identify aroma compounds from fresh white fish. In recent years, this extraction method has been employed in several studies that have examined volatiles produced by several types o f seafood. For example, Josephson et al. (1983, 1985, 1991) used purge and trap extraction to study the volatiles produced by fresh seafood such as fresh Whitefish, Great Lakes salmon and Atlantic and Pacific oysters. Shamaila et al. (1995) also used this technique to evaluate the volatiles in Pacific ocean perch (Sebastes alutus). 2.7.1.1 The effect of fat content on purge and trap extractions Some studies have shown that the fat content of a food product undergoing purge and trap extraction affects the amount and types of compounds found, while others have not. Heikes (1985), in a study on the determination of ethylene dibromide (EDB) in table-ready foods, where various foods ranging from boiled cabbage to chocolate cake icing were spiked with E D B , did not find that fat content had an effect on the efficiency o f the extraction. Persson and von Sydow (1973), on the other hand, in their study of the aroma o f canned beef did find that fat content affected the amount and types o f compounds found. It was also noted that when frit was added to some samples, some volatile compounds were more lipid soluble than others and consequently they were detected to a lesser 22 degree. Also, in those samples, some compounds such as straight chain aldehydes, furan and 2-methyl furan were detected in higher concentrations; possibly because fat is a precursor for those compounds. 2.7.1.2 The choice o f Tenax G C , a porous po lymer , f o r use in purge and t r a p ext ract ion Tenax G C (2,6-diphenyl-p-phenylene oxide polymer) is one o f a group o f porous polymers that are often used in research because o f their ability to trap organic compounds. When the sample o f gas is passed through the porous polymer, the organic compounds are retained and concentrated. Once these compounds are collected, they can either be thermally desorbed or eluted w i t h a solvent (Butler and Burke, 1976; Olafsdottir et al. 1985). Butler and Burke (1976), looking at the capacities and efficiencies o f several porous polymers, concluded that no single porous polymer was universally suitable. One needs to examine the pros and cons o f each and then choose the polymer that is most suitable for the application. Tenax G C has emerged as a widely used porous polymer for food, beverage, and environmental applications (Olafsdottir et al. 1985). I t is particularly advantageous for samples consisting o f only high boil ing point components (Butler and Burke, 1976; Jennings and Fisloof 1977). This is due to this polymer's high temperature limit and relatively l o w retention volumes, which al low the trapped compounds to be desorbed more rapidly than from other adsorbents (Butler and Burke, 1976). I n addition, it also has the advantage o f having shorter recovery times (Jennings and Fi lsoof 1977). 2.7.1.3 T h e elut ion o f adsorbed volati les f r o m porous polymers w i t h e thy l ether Olafsdottir et al. (1985) examined the reproducibility and absolute recoveries o f volatiles from Tenax G C desorbed w i t h ethyl ether. They found that this method resulted in a variability in analysis o f 23 less than 2 0 % which is comparable to other adsorbents and solvents. Thus, this procedure was found to be suitable for many objectives and applications. 2.7.2 T h e re lat ionship between Gas C h r o m a t o g r a p h y ( G C ) da ta a n d q u a n t i t y a n d intensi ty j udgements f r o m t ra ined sensory panell ists Persson et al. (1973 a, b) and von Sydow et al. (1970) were among the first researchers to present a clear-cut relationship between quality and intensity judgements from a trained panel and the G C / M S output for the product. Comparisons o f these t w o sources o f data are, however, not wi thout their pitfalls. V a n Gemert et al. (1987) found that the relationship between sensory analysis and chemical, physical and instrumental parameters was complex. The characteristic aroma o f a food is often not the result o f one compound alone, but rather, results from an interaction between several compounds. Consequently, an increase or decrease in only one odour compound might result in both increases and decreases in several o f the sensory odour qualities (von Sydow et al., 1970). There is a second diff iculty in comparing these t w o types o f data. Sometimes compounds that are highly correlated w i t h flavour w i l l not be flavour substances, although, more commonly they w i l l be (Powers and Keith, 1968). A t times this can occur i f more than one compound co-elutes, or i f the compounds responsible for the aroma are being adsorbed by an active site near the exit port and wash o f f when a major peak elutes (Will iams and Tucknott , 1977). 24 2.8 Stat ist ical analysis 2.8.1 T h e box p lot The box plot, first introduced in 1977, has proven to be an effective means o f producing a visual summary o f data. In this type o f plot, there is a box that is divided w i t h a horizontal line, and there are also t w o protruding "whiskers" that extend vertically from the top and bo t tom o f the box. The box port ion is comprised o f the t w o middle quartiles o f data, that is, the data that falls between the 25 t h and 75 t h percentiles. This interquartile range is computed as IQR= Q0.75 - Qo 25, which serves to measure the amount o f variation in the data. A horizontal line splits the box at the median (50 t h percentile). The lower whisker is defined as the smallest observation that is greater than or equal to the lower quartile minus 1.5 X IQR. Similarly, the upper "whisker" is defined as the largest observation that is less than or equal to the upper quartile plus 1.5 X . IQR, which may be the upper extreme o f the data. A n y values that fall outside this range are considered to be outliers and are plotted as individual points (Ma , 1992). 2.8.2 P r inc ipa l component s imi la r i t y (PCS) PCS is a technique that was developed by combining principal component analysis (PCA) , a data compression method which is based on identifying the most important directions o f variability in a multivariate data space, w i t h pattern similarity. I n PCS, principal component (PC) scores are used for computing pattern similarity constants instead o f using the original data directly (Vodovotz et al., 1993). Furtula et al. (1994b) considered PCS to be an extended version o f PCA. PCS can utilise the information o n variation o f the P C A principal components for classification purposes (Vodovotz et al., 1993). 25 When Vodovotz et al. (1993) compared PCS w i t h P C A and other types o f multivariate analyses they found that they compared favourably. Also, they found that PCS had better resolution than PCA. PCS can also graphically illustrate a larger number o f computed PC outputs than P C A (Furtula et al. 1994b). I n P C A it is customary to use only 2 PC scores for a 2-dimensional (2-D) PC plot or three PC scores for a 3-dimensional (3-D) plot. However, portions o f the original data that may have been important for classification can be ignored. This is particularly true in flavour analysis where it is not uncommon for seemingly minor compounds to play an important role in creating characteristic f lavour notes. W i t h PCS, results from numerous PC scores can be displayed graphically in a 2 D figure, minimising this problem (Vodovotz et al., 1993). PCS is most useful when the size o f the data matrix is large (Furtula et al. 1994a). The advantages offered by PCS diminish as the number o f PC scores for computation decreases (Furtula et al. 1994b). 26 3. Materials and Methods 3.1 Exper imen ta l condi t ions used in the rear ing o f salmon used in th is s tudy The fish for this research were obtained from the Department o f Fisheries and Oceans, West Vancouver Laboratory, West Vancouver, B.C. A total o f 660 salmon were used in the Fisheries and Oceans experiment. The all-female seawater-adapted hatchery-raised one year o ld Qual icum chinook salmon had been selected for uni form size and they originated from Sea Spring Salmon Farm L t d . (Chemainus B C , Canada). Before commencement o f the study, the fish were divided equally into 12 groups o f 55 fish. Each group was placed into a separate outdoor fibreglass tank. Each o f the 4 m 3 circular tanks was fitted w i t h a 1.5 m diameter inner fibreglass hoop to create a circular swimming channel that was 45 c m wide and 55 c m deep. One hal f o f the tank was covered w i t h plastic netting while the other hal f was covered w i t h black nylon cloth. This allowed the fish to choose between dark and light areas. T o create a current, the seawater was pumped into each tank through a vertically placed pipe that was equipped w i t h three horizontally oriented pipes. A l l the tanks were designed w i t h a f low through system and there was no recirculation o f water (Kiessling, et al. 1994b). A two-by-three factorial design w i t h two ration levels (maximum ration = RL100, 7 5 % o f maximum ration = RL75) and three swimming speeds (SS), (0.5, 1.0, and 1.5 body lengths bl/s) was used (Table 2). Duplicate groups o f f i sh were assigned to each o f the six treatments. The actual amounts o f feed that the fish ingested varied between swimming speeds. A s the SS level was increased, the fish o n the RL100 protocol, which were fed to satiety, consumed more feed.. A t each SS, the RL75 fish were given 7 5 % o f the ration that was given to their RL100 counterparts (Kiessling, et al. 1994b). 27 Table 2 Experimental design used by Kiessling et al (1994 a, b) to assess the influence of sustained exercise and two ration levels on growth of chinook salmon in seawater. A 2 X 3 factorial design was used with two ration levels and three swimming speeds and their treatment numbers have been used as identifiers in statistical analyses. Treatment No. Ration Level3 Swimming Speed (bl/s) 1 75 0.5 2 75 1.0 3 75 1.5 4 100 0.5 5 100 1.0 6 100 1.5 a Ration levels: RL100 (100% ration level) is a ration sufficient for satiation, RL75 is 75% of the RL100 at a given swimming speed 28 Careful records o f daily feed waste were maintained in each case and these allowed accurate estimations o f the actual rations consumed. A l l fish, irrespective o f treatment, were fed 4 to 6 m m Biodry 2500 pellets (Bioproducts Inc., Warrenton, Oregon, U S A ) . The mean levels (% o f dry matter) o f protein, l ipid and ash in the Biodry pellets were 52.0, 20.2 and 12.6, respectively. For the purposes o f this thesis, five representative fish were removed from each tank at the end o f the 212-day study. Subsequently, the fish were kil led by a b low to the head. The salmon were then filleted, labelled, and vacuum packaged in mylar film. Thereafter, the packaged fillets were placed into a - 35°C freezer pending analysis months later. Fillets from the left side o f the fish were used for sensory analysis, whereas those from the right side were reserved for instrumental analysis. 3.2 Sensory Analys is 3.2.1 T h e selection o f sensory panell ists fo r Q D A analysis Seven panellists, three men and four women, were recruited from the staff and students o f the U B C Food Science Department. A l though seven people were trained, only six people could be accommodated in any one sensory test due to the limited amount o f sample. The extra trained person was available in the event that a regular panellist was unable to attend a session. Interest in the experiment and the availability o f personnel were the main criteria used for panel selectioa I t was also important that the selected panellists generally l iked to eat salmon. The individual's experience on sensory panels was also considered to be an asset. I t was fortunate that most o f the panel members had some previous sensory panel experience. 29 3.2.2 Sensory panel t r a i n i n g Panel training was carried out over a three week period, during which the panellists met five times. Training consisted o f first gathering descriptive terms from the panellists in a round table format fol lowed by eliminating the less distinguishable or redundant terms. That procedure produced 27 terms to describe the aroma, flavour, and texture, as wel l as the overall acceptability o f the salmon. Agreement was then reached amongst the panellists regarding the scoring o f those attributes. Bo th w i ld and farmed Spring (chinook) salmon were used in the training sessions. These fish were purchased at a local fish market and they were o f similar size to the experimental fish. The purchased fish were thought to be the best sources o f the characteristic flavour, texture, and aroma extremes needed for training. The w i ld salmon were assumed to have been much more active and less wel l fed than their pen-reared counterparts. The market fish were filleted. Subsequently, the fillets were vacuum packaged in a barrier film, and then frozen at -35 °C unti l the day before they were needed for analysis. A t that t ime they were placed in a -4 °C freezer to partially thaw overnight. Samples o f approximately the same size (approximately 1 c m by 3 cm) were wrapped in foi l , and baked at 190 °C for approximately 15 minutes, or unt i l they were cooked before being presented to the panellists. 3.2.3 T h e selection o f sensory a t t r i bu te terms The panellists were asked to list as many aroma, taste and texture notes as they could detect in the cooked salmon samples. When a panellist observed a flavour note, all the panellists wou ld retaste the samples, looking for that sensory note. They then discussed their individual observations. This list 30 o f attributes that was compiled was subsequently reviewed by the panel for the purpose o f eliminating the terms that were either ambiguous, or redundant. The terms were defined by the panellists to ensure that all the panellists were measuring the same sensatioa Whenever it was possible, reference samples for the attributes were provided to aid in clarifying the terms for the panellists. For example, boiled milk, boiled potato, and seaweed samples were provided to the sensory panel as a reference for the corresponding aroma terms. 3.2.4 Ballot familiarisation by sensory panellists Addit ional training sessions were necessary to al low the panellists to become familiar w i t h using the ballot. Af ter selection o f the terms for the study, a sensory ballot was produced and copies were presented to the panellists. They were then given fish samples and asked to rate them using this ballot and then discuss their scores. This procedure enabled assessment o f whether each panel member was using the same intensity scale in the prescribed manner. One source o f disagreement in the panellists responses occurred when one or more o f the panellists did not have the same conceptualisation o f a particular descriptive t e r m I n this situation, every effort was made to clarify the te rm in question. This was sometimes accomplished by producing a reference, or by having the panellists discuss the sensation amongst themselves. I f it became apparent that a term was ambiguous, or that there wou ld never be any real agreement among panellists, the term was discarded. 3 1 3 .2 .5 T h e use o f composite samples The training sessions also provided a fo rum for the panellists to express any ideas that they felt could improve the panel's performance. One suggestion was to construct composite samples from each fish, by combining slices from the anterior, middle and posterior sections o f the fillet before cooking. Dur ing the course o f the first few training sessions it had become apparent that the intensities and profiles o f the sensory attributes changed substantially between the different portions o f the fillet; the anterior port ion being much more flavourful then the posterior port ion. This finding is in agreement w i t h Johnsen and Kel ly (1990) who found that anterior and posterior portions o f fish could have quite disparate flavour profiles. 3.2 .6 Sensory panel set-up Steps were taken to eliminate as many sources o f error as possible that may have influenced the panellists perceptions. For instance the panellists were asked to refrain from various activities such as drinking coffee, wearing after-shave, or perfume (Rutledge and Hudson, 1990). Since the appearance o f the fish was not being tested, the sensory testing was conducted under red lights to mask any variation in appearance and, thus reduce the risk o f panellist bias. T o avoid any carry over o f flavours from one sample to another, the panellists were given distilled water and unsalted crackers to help cleanse their palate between samples. As much as possible, background sound was kept to a min imum to prevent this from disturbing the panellists. 32 3.2.7 Sensory panel session schedul ing Sensory tests were performed on ten different days. A l l six treatments were represented o n each o f these days. Originally it was planned to have one session per day during which all the treatments and a reference wou ld be rated. Af ter the first session, however, it was found that having the task o f assessing samples from six treatments and a reference for 28 attributes at one sitting resulted in some o f the panellists becoming fatigued and making errors. Steps were then taken to adapt the procedure to reduce panellist error. I n this regard, it was decided that three o f the treatments wou ld be selected at random for the morning session, and the remaining three were set aside for the afternoon. A reference was evaluated by the panellists at both the morning and afternoon sessions. Sometimes, a panellist wou ld be unable to attend one o f the t w o sessions on a given day. When this situation arose, the panellist(s) wou ld rate both sets o f samples at the session they attended. T o accomplish this the panellist was given a short break fo l lowing the scheduled sensory panel session and then he/she was presented w i th the samples from the session that could not be attended. A l l the sensory panels were carried out between June and August, 1992. Seven sensory panel days were carried out in a t w o and a hal f week period. This was fol lowed by a three week break. Af ter the break, the panellists participated in a training session to ensure consistency in judgements between the t w o periods. The three remaining sessions were then held over a one week period. 3.2.8 T h e p repara t ion o f samples fo r sensory panels O n the day before a panel session was to take place, one fillet from each o f the six treatments was selected at random and transferred along w i t h packages o f reference fish, from the -35°C freezer 33 to one set at -4°C. This allowed the fillets to partially thaw overnight so that they could be sliced the next day on a meat slicer without becoming too mushy. The fillets samples were removed from the -4°C freezer just prior to being sliced using a Hobart meat slicer to a thickness o f approximately 3 mm. Dur ing the slicing o f each fillet, at least 6 slices were taken from each o f the three sections referred to above i.e., anterior, middle and posterior. Slices from each section were then randomly distributed into six piles o f slices on pieces o f aluminium foiL and efforts were made to ensure that the six samples were as similar as possible. The six aluminium foi l sheets were labelled previously w i th a three digit random code and samples from the same fillet had the same number inscribed, using a permanent ink felt marker. The dul l side o f the foi l was always on the outside. Care was taken to pile the slices from the various sections so that the skin faced the same direct ioa The samples would , once cooked, have the appearance o f a solid piece o f fish. Samples for both the morning and afternoon sessions were prepared in the morning before the first panel sitting. The afternoon samples were stored in a 4°C cooler pr ior to being cooked and presented to the panel. The fo i l wrapped samples were placed on a fo i l pan and placed in a 190°C oven for 15-20 minutes. Then the cooked samples were served to the panellists as promptly as possible. Frequently, it was difficult to have all six panellists assembled when the samples were ready, even when they were notif ied just prior to the fish being served. When it was k n o w n that a panellist was going to be a few minutes late, the samples were left over the vent from the oven to keep them warm. 34 3.2.9 T h e reference samples used d u r i n g sensory panels The reference consisted o f a composite sample from the fillets o f numerous farmed Spring (chinook) salmon that had been obtained commercially from a local seafood market. These fillets were sliced to uni form thickness using a meat slicer. Af ter this, the slices were divided into four groups: the anterior, anterior and back midsections, and posterior o f the fillet. The slices from the corresponding sections o f all the fillet were pooled together and mixed thoroughly. Fol lowing this they were vacuum packaged w i th each bag containing at least 12 slices o f fish. The packages were numbered either 1, 2, 3 or 4 depending on which section o f fillets had been enclosed. A l l o f these packages were placed in a -35°C freezer. One package from each section was removed the day before a sensory panel session and.subsequently these were placed in a -4°C freezer over night. O n the morning o f the sensory panel day, composite samples were prepared from the fish slices. Slices from each section were evenly distributed to twelve samples i.e., six reference samples for the morning and six for the afternoon session. The reference samples were wrapped in fo i l and labeUedwi than"R" . 3.2.10 T h e sampl ing procedure employed by panellists The sensory tests were performed in the sensory panel r o o m located in the U B C Food Science Building. The panellists first rated the reference and then the treatment samples in random order for the aroma attributes. The reference sample, and then the treatment samples were judged o n the remaining attributes. Testing all the samples first for aroma helped to insure that all the samples wou ld be close 35 to the same temperature. This sequence was important because as the fish cools the amount o f volatiles given o f f decreases. The panellists were asked to take a fork ful l o f fish, including portions from all the slices in the composite sample. This forkfu l o f fish was then placed in the mouth and chewed. The panellists then evaluated the sample for the various taste and texture attributes. I f necessary, the panellist could take a second or third forkful . The panellists then either expectorated or swallowed the fish. Disti l led water was provided to the panellists to rinse their mouths, and unsalted crackers were provided to help cleanse the palate between samples. 3.2.11 Genera t ion o f numer ica l scores f r o m the sensory bal lot judgements The ballot (Fig. 1) used by each o f the panellists consisted o f a 10 c m unstructured scale anchored w i th a term at both ends for each o f the attributes being tested. The panellists were asked to indicate their score by placing a vertical line through the scale at the appropriate spot. A numerical score could then be generated by measuring w i th a metric ruler from the left side o f the scale to the point where the panellist's line crossed the line. 3.3 I n s t r u m e n t a l analysis o f cooked salmon samples 3.3.1 P repara t ion o f salmon samples fo r i ns t rumenta l analysis Due to t ime and equipment constraints, it was only possible to perform instrument analysis o n a maximum o f t w o samples per day. Late in the afternoon o n the day prior to analysis, the salmon fillets were placed in a freezer that was set at -4°C. This allowed them to thaw slightly overnight which 36 Figure 1 Sensory ballot that was used to evaluate samples o f cooked, farmed chinook salmon Aroma Profi l ing Score Sheet Name: Date: Sample Number: Instructions 1. P E R F O R M A R O M A P R O F I L I N G O N A L L S A M P L E S B E F O R E M O V I N G O N T O F L A V O U R A N D T E X T U R E P R O F I L I N G 2. Open the fo i l wrapper and smell contents, flake the fish i f necessary to release more o f the aroma 3. M a r k the horizontal lines w i t h a vertical Erie to indicate the intensity o f each o f the odours listed. 4. Record any meaningful observations either on the bot tom or the back o f the page. 4. D O U B L E C H E C K T O M A K E S U R E T H A T T H E S A M P L E N U M B E R R E C O R D E D O N T H E SCORE S H E E T IS CORRECT. Seaweedy Boi led mi lk Boi led potato Lemony Sour Fishy Chickeny Oi ly Fresh 1 none 1 " 1 very seaweedy 1 1 none 1 1 strong 1 1 none 1 1 - strong 1 1 none 1 1 very lemony 1 1 none 1 1 very sour 1 1 none 1 1 very fishy 1 1 none 1 1 very chickeny I 1 not oily 1 1 very oily 1 old fresh 37 Flavour Profi l ing Score Sheet Instructions 1. Taste the sample (note: texture analysis can be performed simultaneously) 2. M a r k the horizontal lines w i t h a vertical line to indicate the intensity o f each o f the flavour terms listed below. 3. Wri te any additional comments on the bot tom or back o f the page. Flavour Fishy Earthy Papery Bitter Sour Lemony Salty Spicy Brothy 1 weak 1 1 intense 1 1 none 1 1 very fishy 1 1 none 1 1 very earthy 1 1 none 1 1 very papery I 1 not bitter 1 1 very bitter 1 1 not sour 1 1 very sour I 1 not lemony 1 1 very lemony 1 1 not salty 1 1 very salty 1 1 not spicy 1 1 very spicy not brothy very brothy 38 Texture Profi l ing Score Sheet Instructions 1. Chew sample. 2. M a r k horizontal line w i t h a vertical line to indicate the intensity o f each o f the texture terms listed. 3. Wr i te any additional comments o n the bo t tom or back o f the page. 4. P L E A S E R E C H E C K A N D M A K E SURE S A M P L E N U M B E R IS CORRECT. Moistness dry very moist Powderiness not powdery very powdery Flakiness Firmness not flaky firm very flaky 1 soft Chewiness not chewy very chewy Cohesiveness loose mass compact mass Adhesiveness not sticky very sticky Mushiness not mushy very mushy Overall Acceptance extreme dislike extreme like 39 facilitated easier handling. O n the day o f the test, the fillets were skinned and chopped into 1 c m 3 cubes. T o ensure that there wou ld be sufficient sample for a min imum o f three replicates o f each treatment for GC, Instron, and p H analysis, all the fish cubes from the same treatment were pooled. These cubes were mixed thoroughly to ensure that the samples were uniform. T o avoid having any refreezing o f the samples, the instrumental analyses were performed on the same day that the fillets were removed from the freezer. The fillets were divided by weight into the sample sizes that were required for the various experiments. These samples were wrapped in aluminium foi l , dul l side out, and stored in a 4 °C cooler unti l required. A t that time the foi l packages were placed in a preheated 190 °C oven and baked unti l the fish muscle was no longer translucent (15-20 minutes). 3.3.2 G C headspace analysis o f cooked salmon samples G C Headspace analysis, after using a purge and trap extraction, was performed on the cooked samples. Due to limitations in the amount o f fish available from each treatment, it was only possible to perform this analysis in triplicate. A series o f experiments that were designed to select the most appropriate conditions for this purge and trap extraction were performed before the test samples were analysed. The five variables examined were: sample size, extraction temperature, extraction time, nitrogen flow rate, and the amount o f Tenax packed into the traps. T o determine the right leve l samples were extracted at various levels wi th in a realistic range e.g. extraction times o f 1, 2, 3, and 4 hours were used. The four variables not being examined in a particular experiment were kept at a moderate level (which, coincidentally, turned out to be the levels chosen for this study). Once the results from these 40 experiments were graphed, the best level was determined based on the level o f volatiles extracted and the size o f the increase in recovery between levels. For choice o f temperature and the extraction time, the risk o f sample loss due to moisture build-up in the Tenax G C also was taken into account. I t was concluded that the best conditions for conducting this extraction were: a sample size o f 200 g, an extraction temperature o f 70 °C, an extraction time o f 3 hours, a f l ow rate o f 50 ml/min, and 120 m g Tenax GC. (conditions used are summarised in Table 3). Three 200 g foi l packages o f the chopped fillets were prepared from each treatment and these were placed in a 190 °C oven unti l cooked (approximately 20 minutes). Once cooked, the fillets were gently flaked w i t h a fork and then promptly deposited, along w i t h any liquid that was released during cooking, into prewarmed 1 L extraction vessels (Wheaton, Mil lv i l le, NJ) . The six extraction vessels were maintained at 70 °C w i th a waterbath (Haake FS). The water passed from the waterbath through plastic tubing into a copper pipe that connected all the extraction vessels in parallel. The water exited from the vessels in a similar fashion and was recirculated back to the waterbath. This was designed to ensure that all the vessels wou ld be maintained at the same temperature. The internal standard used in this experiment was tetradecane (purity: 9 9 + %, Aldr ich) that was dissolved in diethyl ether (spectranalyzed grade, Fisher Scienti f ie)( l : 10 v/v) . Af ter removing the tenax trap assembly, 20 microlitres o f this standard was injected into the extraction vessel through the open side arm. The trap assembly was promptly replaced and the extraction vessel was closed tightly. The vessels were left to equilibrate for 30 minutes. Af ter this time had expired, the volatile compounds that were produced by the cooked fillet were purged from the flask w i t h prepurif ied N2 gas ( U H P grade, Linde Un ion Carbide) into the tenax G C trap. 41 Table 3 Conditions used in the extraction o f cooked salmon using a purge and trap procedure sample size 200 g no. o f replicates 3 extraction vessel temperature 70 °C equuibrium time 3 0 m i n . extraction time 3 hours N 2 gas f l ow rate 50 ml/min. adsorbent used Tenax G C amount o f adsorbent 120 mg internal standard tetradecane solvent used to elute traps diethyl ether 42 Approximately 120 m g o f Tenax G C (60/80 mesh, Al l tech), a porous polymer, was packed into a glass tube (11.5 cm, 6 m m O.D., 4 m m I.D.), that was held in place between t w o plugs o f glass woo l . The Tenax G C had been conditioned pr ior to the extractions to remove contaminants. This involved holding the Tenax G C at 200°C w i th N 2 flowing through the tube at 30 ml/min for a min imum o f 4 hours (Jennings and Filsoof, 1977). The narrow end o f the trap was wrapped several times w i t h tef lon tape before attaching it to the extraction vessel to help ensure an air-tight fit. Af ter the 3 hour extraction was completed, the Tenax traps were removed and eluted w i t h 2 m l diethyl ether (spectranalyzed grade, Fisher Scientific). The extracts were stored in 3.7 ml glass vials (screw top l id w i t h septa) (Supelco), wh ich were placed in a 4 °C cooler unt i l required for G C analysis. A t that t ime, the extract was concentrated by evaporating the ether using a gentle stream o f nitrogen, unt i l approximately 100-200 microlitres remained. One microlitre o f this extract was then injected into a Varian 3700 gas chromatograph (Varian and Associates, Inc. Palo A l to , C A ) set up and operated according to the specifications given in Table 4. Relative amounts o f each compound were then determined by taking a ratio o f each peak area to that o f the internal standard. These data were then subjected to an A N O V A (Systat 5.01, Systat Inc.), to determine i f there were any treatment effects. Only peaks that consistently appeared in all o f the chromatograms were examined. 3.3.3 Ins t ron TPA analysis o f cooked salmon samples Bourne's (1978) T P A method, based on the compression o f the sample w i t h the Instron Universal Testing Machine (Mode l 1122, Instron Corp., Canton, M A ) , was used to achieve an objective quantitative measurement o f the cooked salmon texture. Cylinders o f flaked salmon were 43 Table 4 G C conditions used in the analysis o f purge and trap extracts from cooked chinook salmon samples G C Var ian3700 Integrator 3390A Hewlett-Packard Detector Flame Ionisation Split injection 100:1 Column capillary SPB-1 nonpolar Column manufacturer Supelco Inc. Internal diameter 25 m m F i lm thickness 0.25 micro meters Column length 30 m Init ial temperature 50 °C Time initial temperature held 5 min. Rate o f heating 5 °C/min. Final temperature 220 °C Injector port temperature 250 °C Detector temperature 250 °C H e ( U H P grade) f low rate 30 ml/rnin A i r (Zero Gas) f l ow rate 300ml /m in H 2 ( U H P grade) f l ow rate 30 ml/min Volume o f sample injected l u l 44 formed using a 60 m l syringe (2.6 c m internal diameter) w i t h the end cut o f f at the zero line. Ten grams o f the cooked, deboned, flaked fish were poured into the top o f the syringe and then gently compressed w i t h a flat bottomed plunger to f o r m a cylinder 2 c m high. These fish samples were compressed twice w i t h the Instron, between t w o parallel plates (approximately 14.8 c m diameter), to a height o f 0.5 cm; 2 5 % o f their original height. This testing procedure creates a texture profile curve w i t h t w o peaks from which numerous textural parameters may be measured. These include hardness, firmness, cohesiveness, chewiness and gumminess. Test conditions (Table 5) were selected after preliminary trials. The Instron was interfaced w i th a personal computer, using JCL6000 software, the force/deformation curves at a rate o f 2 times per second were recorded. Prior to analysis, the Instron was calibrated by measuring the difference in load weight output w i t h no weight and w i t h a k n o w n weight. Quadruplet samples o f each treatment were used in this por t ion o f the experiment. 3.3.4 p H measurement o f cooked salmon samples The p H o f the salmon was measured using the procedure outlined by Feinstein and Buck (1984). Samples were prepared by adding 3 g o f f i s h muscle to 30 m l o f deionized, distilled water and homogenisedfor approximately 10 seconds using a Kinematica G m b H homogenizer (speed setting 6). p H was then measured using a Corning p H meter 220 (standardised to p H 7). These analyses were performed in quadruple. 45 Table 5 Conditions used for Instron measurements o f minced cooked chinook salmon samples Analysis T P A N o . o f replicates 4 Load cell 1001b. Crosshead speed 100 mm/min N o . o f cycles o f crosshead 2 Sampling rate 2 data points/sec. Temperature 20°C (approximately) 46 3.4 Da ta Analysis 3.4.1 Analys is o f sensory da ta 3.4.1.1 Exp lo ra to ry analysis Prior to performing complex statistical analyses o f the sensory data, some basic statistics were calculated. For example, for each panellist, the number o f observations, the averages and standard deviations were tallied for the six treatments for each attribute. I n addition, the data were also represented graphically in a series o f boxplots. Boxplots were also constructed for the reference in the same manner. Through this exploratory data analysis, portions o f the data were found to be unacceptable due to an excessive number o f inconsistencies. These included the first replicate (first sensory panel session) as wel l as all the data from panellist 6. These unreliable portions o f the sensory data were subsequently removed pr ior to further analysis, leaving 5 panellists and 9 taste panel days (replicates). 3.4.1.2 Pr inc ipa l component analysis ( P C A ) P C A was performed on the pooled data for the sensory aroma attributes, as wel l as the pooled flavour and texture data using Systat software. For each set o f P C A data, a series o f graphs o f P C I vs. P C 2 were produced. Separate graphs were also produced by labelling data points w i t h the treatment number, the panel day number, and the panellist number. 47 3.4.1.3 ANOVA Several sets o f A N O V A s , were performed on the sensory data. These included a t w o factor A N O V A o n reference data that examined the effect o f panellists, the day effect, and the interaction o f these t w o factors. For the treated samples, all the sensory attributes were subjected to a three factor A N O V A , in which the main effects o f ration level (RL ) , swimming speed (SS), and panellists ( P A N ) were assessed. 3.4.1.4 Z-transformation of significant sensory attribute scores When A N O V A s were performed on the raw sensory data (treated samples), the panellist effect was consistently found to be highly significant. Addit ionally, for several o f the attributes where a significant treatment effect had been uncovered, the panellist X treatment interactions were also significant. To remove the variation in the sensory data due to the panellist to panellist variation, a z-transformation was performed o n all the sensory attributes that had been significantly affected by either R L or SS. Wi th respect to this, z-scores for a given sensory attribute were calculated in several steps. First, the data were sorted, so that the responses o f each panellist could be identified. Fol lowing this, the averages and standard deviations o f the responses o f the individual judges were calculated from the raw data. The z-score for each response was then calculated by first subtracting the panellists average score from each response that she/he gave for that attribute, and then dividing the product by her/his standard deviation (Reid and Durance, 1992). 48 Subsequently, the transformed data were examined using A N O V A . Since the panellist to panellist variation had been eliminated, only a t w o way A N O V A examining the effects o f the R L and SS was necessary. 3.4.2 Calcu la t ion o f I ns t ron T P A parameters 3.4.2.1 Ca l ib ra t ion o f results Prior to using the Instron, the instrument was calibrated daily by using a k n o w n weight. This was accomplished by taking measurements w i t h the empty load cell for thirty seconds to establish a baseline, fo l lowed by placing a 1 k g weight o n the inverted load cell for 30 seconds. This procedure was repeated three times in successioa The calibration factor was then calculated by taking an average o f the scores recorded when the 1 k g weight was applied, and then subtracting the baseline score. The test sample's data were calibrated by first subtracting the baseline value from the data and then dividing it by the calibration factor. These measurements were then converted into Newtons. This was accomplished by multiplying the scores by 9.8 m/s 2. The data were then used to measure or calculate Instron T P A measurements as outlined in Table 6. 3.4.3 Calcu la t ion o f T P A " F i r m n e s s " Firmness, the maximum slope o f the compression cycle, was determined by measuring the maximum slope o f the force curve. The slope was determined by calculating the distance the curve had risen o n the y-axis divided by the distance it had covered o n the x-axis. The slope was calculated for every 1 second interval on the curve. 49 3.4.4 Calculation of peak area I n order to measure cohesiveness and gumminess, it was necessary to first calculate the area under the curve. T o accomplish this, the area o f the "bites" was measured between the start o f the curve and the peak force (the highest measured force). T o calculate the area under the curve, the distance travelled o n the chart for each reading was first determined. This was accomplished by first converting the chart speed to mm/s and then dividing it by the sampling rate, giving the distance travelled on the chart during each measurement. The area was then determined by the sum o f this value multiplied by each point on the curve. 3.4.5 A N O V A of Instron T P A and p H data A t w o way A N O V A looking at the effect o f swimming speed and ration level on various p H and T P A measurements was performed. When a significant result was found in swimming speed, having more then two levels, a Tukey test was also conducted to determine which levels were responsible for the significant differences. The data were analysed w i t h the aid o f Systat statistical software. 3.4.6 Principal Component Similarity (PCS) analysis of Sensory, and G C headspace volatile data P C A using Systat software, was performed separately on the sensory and G C headspace volatile data. The sensory data included only those sensory attributes that had been significantly affected by either R L or SS. These results had subsequently undergone a z-transformation to eliminate 5 0 Table 6 Calculation o f Instron T P A parameters Texture parameters Defini t ion Reference Hardness 1 Peak force during the first compression cycle (Bourne, 1978) Hardness 2 Peak force during the second compression cycle (Bourne, 1978) Firmness 1 M a x i m u m slope o f the first compression cycle (Durance and Collins, 1991) Firmness 2 M a x i m u m slope o f the second compression cycle (Durance and Collins, 1991) Cohesiveness Ratio o f positive force area during the second (Bourne, 1978) compression cycle to that o f the first ( A 2 / A i ) Gumminess Product o f hardness X cohesiveness (Bourne, 1978) 51 the panellist effect. B y averaging the responses o f all the judges for each replicate, a 5 fo ld reduction in the size o f the data set was achieved, leaving 9 replicates for each treatment combinat ioa For the analysis o f the G C data, only those peaks that consistently appeared and were significantly affected by either SS or R L were used. This resulted in 27 out o f a possible 71 peaks being included in this analysis. P C A was performed on these t w o sets o f data, and this resulted in a print out for each, and the scores were saved. Starting w i t h the first PC, the percent o f total variance explained by the PC were added together unt i l more than 90 % o f the variance was accounted for ; these are the PC that were used in the PCS. The scores from these PCs were copied into a data file to be imported into the PCS program. PCS produces the slope and coefficient o f determination for each case, which were subsequently saved in a data file. Af ter import ing this file into a computer spreadsheet, the slope was graphed against the coefficient o f determination. 52 4. Resul ts a n d Discuss ion 4.1 Sensory analysis o f cooked salmon samples 4.1.1 Sensory panel reference samples 4.1.1.1 Purpose o f reference sample A t both the morning and afternoon sensory panel sessions, a reference sample was presented to the panellists along w i t h the test samples. The reference samples were composite samples o f small, randomly selected slices o f fish from several 0.9-1.5 k g farmed chinook salmon. Using this method, a large number o f fairly uni form samples was produced. Since an individual fillet from each treatment was used along w i th a reference sample for each o f the nine panel days, it was essential to include reference samples. Without a reference it wou ld have been difficult to distinguish from the test samples whether a statistical difference in the day was due to a true difference between the sessions or simply stemmed from fish to fish variation. Consistency between reference samples was also important as panellists have a tendency to grade the test samples relative to the reference (Giovanni and Pangborn, 1983). A summary o f the sensory reference data is found in Table 7. The reference was also used to help the panellists calibrate their responses. These reference samples were prepared for the panellists o n several panel training sessions. A t these training sessions, the panellists were able to interact w i t h each other, discussing h o w and why they wou ld give these samples a particular score, eventually reaching a consensus o n the appropriate grade. Dur ing the sensory panels, these reference samples were rated prior to any o f the treated samples. This al lowed the panellists to calibrate their responses, between each other, as wel l as from session to session (Johnsen and Kelly, 1990). However, despite the use o f the reference samples and training, there 53 7 Range, mean, and standard deviation (St. Dev.) o f sensory attributes o f the cooked, farmed chinook salmon reference samples (5 panellists, 9 panel days) Attr ibute Range M e a n & S t . D e v A roma Seaweedy 0.7-7.0 3.75 ± 1 . 4 9 Boi led mi lk 0-4.0 1.29 ± 0 . 8 9 Boi led potato 0-5.8 2.51 ± 1.44 Lemony 0-5.8 1.62 ± 1 . 2 5 Sour 0-4.8 1.06 ± 0 . 9 3 Fishy 0.1-8.9 3.59 ± 1 . 7 7 Chickeny 0-7.0 1.76 ± 1 . 7 8 Oily 0-5.2 1.28 ± 0 . 9 6 Fresh " 1.7-9.6 5.72 ± 1 . 4 4 Flavour Flavour 1.0-9.1 5.43 ± 2 . 1 4 Fishy 0.2-8.8 3.49 ± 2 . 0 6 Earthy 0-5.5 1.30 ± 1 . 3 3 Papery 0-7.4 2.01 ± 1 . 4 8 Bitter 0-6.7 1 .22±1.16 Sour 0-4.4 1.06 ± 0 . 8 4 Lemony 0-7.7 1.65 ± 1 . 5 4 Salty 0-4.1 1.53 ± 1 . 0 6 Spicy 0-7.2 2.13 ± 1 . 5 4 Brothy 0-7.7 2 . 1 5 ± 2 . 1 7 Texture Moistness 0.5-9.0 4.89 ± 1 . 9 3 Powderiness 0-8.9 2.55 ± 1 . 9 3 Flakiness 0.1-8.3 3.55 ± 2 . 2 5 Firmness 0.6-8.6 5.30 ±1.71 Chewiness 0.1-8.5 4.13 ± 2 . 1 8 Cohesiveness 0.4-7.7 4.17 ± 1 . 8 3 Adhesiveness 0-8.0 2.91 ± 1.85 Mushiness 0-7.3 1.94 ± 1.58 Overall 2.1-8.7 5.58 ± 1 . 3 7 54 wou ld often be a wide range o f responses received from the panellists for a given sample. The panellists, each w i th their o w n personal style, were consistent from session to session. 4.1.1.2 Reference sample observat ions Significance for the day X panellist term interaction that appears in Table 8 is likely largely due to some o f the unavoidable differences between panel sessions. I t was not always possible to have the panellists start the panel at the same time. F r o m time to time, a panellist was unavoidably detained and came to the session late. Occasionally a panellist was absent for one o f the t w o sessions on a panel day. The panellist took a short break fol lowing the scheduled session he/she attended and then rated the samples from the missed session. The resulting differences in temperature and taste acuity may partially be responsible for the day X panellist interaction (Meilgaard et al., 1991; Larmond, 1977). I t was also noted that on occasion a panellist did not comply w i t h the request to abstain from eating lunch or consuming coffee prior to the panel session and this may have resulted in confusion or carry-over sensations w i t h the samples (Rutledge and Hudson, 1990). 4.1.2 Treated samples 4.1.2.1 E x p l o r a t o r y analysis The data set for this port ion o f the research was extremely large and required some exploratory data analysis. For each individual panellist, boxplots o f each attribute were constructed (examples in appendix A ) . Tables giving each panellist's average, standard deviation, range and number o f observations were also prepared (data not shown). 55 4.1.2.1.1 Boxplots I t proved very difficult to ascertain from the boxplots whether the treatments were significantly different from each other. Even in attributes where a significant difference between treatments existed, the wide variation in panellist rating styles masked evidence o f the treatment differences. These boxplots, however, did clearly show that despite training there was a large amount o f judge to judge variation. The judges differed greatly in their style o f rating o f the attributes, varying widely in the range o f values that they used (data not shown). They, however, appeared quite consistent w i t h their individual styles o f rating the attributes, using the same range and psychological distances between grading levels i.e. what one panellist wou ld grade as a 0.3 cm, a second might score as 1.5 cm. Fortunately, according to Stone et al. (1974), it is not o f critical importance that the individual panellists used different segment o f the scale, as long as their individual performances were constant. 4.1.2.1.2 Delet ion o f unacceptable da ta U p o n examination o f these results, it was decided that some o f the data collected wou ld not be used in any further analysis. A l l the data collected on the first day, as wel l as the contribution o f panellist 6, were removed. The data set collected on the first day was eliminated because i t contained numerous panellist errors, mostly related to panellist fatigue. I n the first panel session, it was wrongly assumed that all the panellists wou ld be capable o f rating each o f the 7 samples (6 treatments and one reference) for the 28 sensory attributes wi thout becoming fatigued. I n subsequent sessions, the panel days were divided into morning and afternoon sessions. Three randomly chosen treatments and a reference were presented to 56 Table 8 A N O V A results o f judge and panel day effect on reference samples for 28 sensory attributes o f cultured chinook salmon (5 panellists, 9 panel days) Sensory attributes Day F ratio Panellist Day X Pan Mean square error A r o m a Seaweedy 1.748 3 0 . 9 9 2 * * * 1.006 0.848 Boi led mi lk 1.196 18 .450*** 3 .748** * 0.244 Boi led potato 0.482 18 .504*** 1.454 1.069 Lemony 0.682 2 7 . 4 1 6 * * * 0.941 0.668 Sour 0.261 3.494* 1.383 0.698 Fishy 3.048** 5 .303** 1.330 2.164 Chickeny 1.740 38 .885* * * 1.325 1.149 Oi ly 2.576* 11 .447*** 1.233 0.486 Fresh 2.109 18.584*** 1.233 0.486 Flavour Flavour 0.597 35 .396* * * 0.762 1.762 Fishy 0.501 19.857*** 1.474 2.315 Earthy 2.536* 4 7 . 9 2 5 * * * 2 .859** 0.418 Papery 1.321 5.834** 1.233 1.449 Bitter 0.978 5.266** 1.337 1.099 Sour 1.145 0.681 0.972 0.704 Lemony 1.209 12 .622*** 0.982 1.617 Salty 1.805 4 2 . 8 3 0 * * * 3 .383* * * 0.308 Spicy 0.853 2 0 . 3 2 1 * * * 1.005 1.092 Brothy 1.364 129.229*** 0.931 0.715 Texture Moistness 0.904 16.656*** 0.579 2.361 Powderiness 1.091 2.301 1.775* 2.469 Flakiness 0.713 39 .043* * * 1.100 1.817 Firmness 2 .331* 8 .238** * 1.206 1.842 Chewiness 2.500* 6 7 . 5 7 7 * * * 1.683* 1.232 Cohesiveness 1.593 13 .919*** 0.884 2.039 Adhesiveness 1.589 2 8 . 1 2 8 * * * 1.776* 1.232 Mushiness 1.857 3.148* 1.554 1.376 Overall 1.567 2 4 . 2 2 3 * * * 1.503 0.728 * p<.05, * * p< .01 , * * * p<.001 57 the panellists in the morning w i th the remaining three treatments and a second reference offered at an afternoon session. A large number o f missing data points, as wel l as excessive variation in replicates (data not shown), made it necessary to omit the contribution o f panellist 6 from the data set. The data from panellist 6 were actually a combination o f data contributed by three people. Each o f these three people had a personal style o f rating the samples that varied greatly from one other. Since consistency and accuracy are so very important for panellists in this type o f sensory analysis (Stone et al. 1974), it was decided not to include the data from panellist 6. U p o n elimination o f this data set a second set o f summary statistics was calculated (Tables 9 - 1 5 ) . 4.1.2.2 The use of replacement panellists As this panel took place over a period o f several weeks during the summer, it proved an impossible task to find six wi l l ing panellists who were able to commit to being present throughout the duration o f the experiment. I t was decided that back-up panellists, who had also completed the training sessions, wou ld substitute for absent panellists. Fortunately, back-up panellists were only necessary for panellists 5 and 6. Evidence o f the substitution o f a back-up panellist for panellist 5 on t w o panel days became very apparent during data analysis. I n Figures 2-4 a set o f data points, one from each treatment, was separated from the main cluster. This set was evident in the aroma and pooled texture variables (Fig. 2 and 4 ) ; no evidence o f this set was readily apparent in the corresponding flavour graph (Fig. 3). These irregularities in the data were further examined by producing a second set o f these figures where the treatment number was replaced w i t h the panel day number (Fig. 5-7). F r o m these 58 Table 9 Range, mean, and standard deviation (St. Dev.) o f cooked, cultured chinook salmon sensory attributes (all treatments combined; 5 panellists, 9 panel days) Attr ibute Range Mean & St. Dev. A roma Seaweedy 0-9.9 4.05 ± 2 . 1 7 Boi led mi lk 0-6.1 1.26 ± 1 . 0 0 Boi led potato 0-6.7 2.20 ± 1 . 4 7 Lemony 0-6.3 1.44 ± 1.21 Sour 0-6.3 1.37 ± 1 . 2 3 Fishy 0-8.9 3.58 ± 1 . 9 2 Chickeny 0-7.6 1.51 ± 1.47 Oily 0-5.6 1.36 ± 1 . 0 3 Fresh 0.6-9.6 5.26 ± 1 . 7 1 Flavour Flavour 0.3-9.4 5.63 ± 2 . 0 7 Fishy 0.1-8.8 3.42 ± 2.20 Earthy 0-7.5 1.23 ± 1 . 2 6 Papery 0-8.6 1.99 ± 1 . 6 4 Bitter 0-7.1 1.32 ± 1 . 2 5 Sour 0-7.1 1 .27±1.11 Lemony 0-7.7 1.30 ± 1 . 6 7 Salty 0-7.0 1.82 ± 1.28 Spicy 0-8.8 2.35 ± 1 . 7 2 Brothy 0-8.4 2.31 ± 2 . 1 3 Texture Moistness 0.5-9.2 5.34 ± 1 . 9 2 Powderiness 0-8.9 2.45 ± 1 . 9 7 Flakiness 0-8.4 3.14 ± 2 . 0 4 Firmness 0.5-8.8 4.63 ± 1 . 8 9 Chewiness 0-9.0 4.13 ± 2 . 1 5 Cohesiveness 0.2-9.0 4.14 ± 1 . 8 4 Adhesiveness 0-8.3 2.76 ± 1 . 8 8 Mushiness 0-8.6 2.43 ± 2 . 0 4 Overall 1.9-8.9 5.48 ± 1 . 4 4 59 Table 10 Mean sensory scores and standard deviation o f cultured chinook salmon aroma attributes for each ration level X swimming speed treatment ( 5 panellists; 9 panel days) Ration Level 7 5 % 100% Swimming Speed (bl/s) Attr ibute 0.5 1.0 1.5 0.5 1.0 1.5 Seaweedy 3.95 ±2.22 4.17 ±2.23 3.50 ±2.07 4.38 ±2.23 4.34 ± 2 . 4 6 4.52 ±2.63 Boiled milk 1.37 ±1.19 1.31 ±1.05 1.23 ±1.14 1.17 ±0.88 1.28 ±1.05 1.16 ±0.80 Boi led potato 2.28 ±1.56 2.19 ±1.61 2.36 ±1.50 2.06 ±1.50 1.84 ±1.26 1.89 ±1.29 Lemony 1.44 ±1.17 1.51 ± 1 . 3 3 1.26 ±1.14 1.34 ±1.07 1.56 ±1.33 1.17±1.00 Sour 1.56 ±1.29 1.36 ±1.23 1.06 ± 0 . 9 3 1.56 ±1.44 1.76 ±1.05 1.52 ±1.23 Fishy 3.44 ±1.99 1.46 ±1.71 3.46 ± 1 . 7 4 3.27 ±1.85 3.63 ± 2.20 4.09 ±2.18 Chickeny 1.49 ±1.28 1.30 ±1.12 1.75 ± 1 . 7 5 1.28 ±1.15 1.45 ±1.34 1.34 ±1.30 Oi ly 1.29 ±1.02 1.37 ± 0 . 9 7 1.37 ± 0 . 9 4 1.27 ± 0 . 9 9 1.51 ±1.29 1.51 ±1.04 Fresh 5.23 ±1.67 5.15 ±1.62 5.26 ± 1 . 3 0 5.24 ±1.98 5.00 ±1.98 4.78 ±1.88 60 Table 11 Range o f sensory scores o f cultured chinook salmon aroma attributes for each ration level X swimming speed treatment (5 panellists; 9 panel days) Ration Level 7 5 % 100% Swimming Speed (bl/s) Attr ibute 0.5 1.0 1.5 0.5 1.0 1.5 Seaweedy 0.7-9.3 0.2-9.6 0.2-9.7 0.7-9.2 0.2-9.9 0-9.8 Boi led mi lk 0-5.8 0-4.6 0-6.1 0-4.3 0-3.8 0-3.4 Boi led potato 0-5.7 0-6.0 0.1-6.7 0-6.0 0-5.3 0-5.3 Lemony 0-6.3 0-6.3 0-6.1 0-4.9 0-5.6 0-4.1 Sour 0-5.3 0-5.7 0-4.0 0.1-6.3 0-5.6 0-6.3 Fishy 0-8.1 0.1-7.3 .04-6.6 0.2-8.0 0.2-7.9 0.1-8.6 Chickeny 0-5.9 0-4.6 0-7.6 0-5.7 0-7.4 0-6.7 Oi ly 0-3.5 0-5.4 0-4.1 0-4.5 0-5.6 0-4.6 Fresh 1.9-8.6 1.5-8.9 2.3-9.6 1.2-9.1 1.2-9.5 0.6-8.9 61 Table 12 Mean sensory scores and standard deviation o f cultured chinook salmon flavour attributes for each ration level X sv\Tmming speed treatment (5 panellists; 9 panel days) Ration Level 7 5 % 100% Svvdmming Speed (bl/s) Attr ibute 0.5 1.0 1.5 0.5 1.0 1.5 Flavour 5.95 ±1.84 5.86 ±2.13 5.69 ±1.89 5.31 ±2.07 5.69 ±1.89 5.60 ±2.07 Fishy 3.41 ±2.08 3.31 ±2.25 3.33 ±2.40 3.49 ±2.15 3.33 ±2.40 3.50 ±2.26 Earthy 1.09 ±1.02 1.36 ±1.31 1.34 ±1.42 1.13 ±1.03 1.34 ±1.42 1.09 ±1.21 Papery 1.92 ±1.73 2.02 ±1.42 2.06 ±1.76 2.04 ±1.64 2.06 ±1.76 1.92 ±1.70 Bitter 1.46 ±1.39 1.52 ±0.94 1.52 ±1.19 1.13 ±1.16 1.52 ±1.19 1.49 ±1.66 Sour 1.47 ±1.54 1.16 ±1.00 1.25 ±0.96 1.25 ±1.00 1.25 ±0.96 1.46 ±1.22 Lemony 1.28 ±1.05 1.19 ±0.95 1.22 ±0.97 1.15 ±1.06 1.19 ±1.02 1.06 ±0.92 Salty 2.03 ±1.35 2.01 ±1.34 1.85 ±1.32 1.79 ±1.39 1.87 ±1.28 1.91 ±1.30 Spicy 2.48 ±1.72 2.38 ±1.69 2.64 ±1.68 2.42 ±1.88 2.22 ±1.81 2.40 ±1.78 Brothy 2.51 ±2.05 2.56 ±2.12 2.48 ±2.28 2.07 ±2.00 2.27 ±1.99 2.28 ±2.06 62 Table 13 Range o f sensory scores o f cultured chinook salmon flavour attributes for each ration level X swimming speed treatment (5 panellists; 9 panel days) Attr ibute Ration Level 7 5 % 100% Swimming Speed (bl/s) 0.5 1.0 1.5 0.5 1.0 1.5 Flavour 0.6-9.4 1.2-9.1 0.4-8.8 0.9-9.0 0.4-8.8 0.4-9.1 Fishy 0.1-8.0 0.1-8.0 0.1-8.2 0.2-8.2 0.1-8.2 0.1-8.1 Earthy 0-4.3 0-5.1 0-7.3 0-4.7 0-7.3 0-4.6 Papery 0-7.5 0-6.2 0-7.0 0-6.7 0-7.0 0.2-8.5 Bitter 0-5.3 0-4.4 0-6.3 0-6.2 0-6.3 0-7.1 Sour 0-7.1 0-4.2 0-4.0 0-3.7 0-4.0 0-5.0 Lemony 0-3.5 0-4.0 0-3.9 0-5.1 0-4.0 0-3.3 Salty 0-5.6 0-5.8 0.1-5.2 0-7.0 0-4.6 0.1-5.4 Spicy 0.1-5.9 0-7.8 0.1-8.0 0-7.8 0-6.9 0.1-8.8 Brothy 0.2-7.4 0-7.8 0.2-7.3 0-7.1 0.1-8.4 0.1-7.7 63 Table 14 Mean sensory scores and standard deviations o f cultured chinook salmon texture attributes for each ration level X swimming speed treatment (5 panellists; 9 panel days) Ration Level 7 5 % 100% Swimming Speed (bl/s) Attr ibute 0.5 1.0 1.5 0.5 1.0 1.5 Moistness 5.27 ±1.90 5.52 ±1.76 5.04 ±1.93 5.48 ±1.67 5.98 ±2.02 5.66 ±1.94 Powderiness 2.50 ±1.93 2.33 ±1.88 2.45 ±1.94 2.42 ±2.02 2.26 ±1.99 2.49 ±2.11 Flakiness 2.92 ±2.12 3.21 ±2.05 3.04 ±1.79 3.07±1.77 2.95 ±2.08 2.88 ±1.83 Firmness 4.78 ±1.68 4.49 ±1.98 5.02 ±1.30 4.21 ±1.97 3.80 ±2.03 4.22 ±2.07 Chewiness 4.27 ±2.26 4.03 ±2.12 4.33 ±1.92 3.83 ±2.18 3.99 ±2.05 4.30 ±2.23 Cohesiveness 4.34 ±1.58 4.17 ±1.75 4.62 ±1.72 3.74 ±1.78 4.03 ±2.07 3.91±2.02 Adhesiveness 2.83 ±1.89 2.77 ±1.89 2.67 ±1.85 5.58 ±1.64 2.94 ±1.98 5.52 ±2.05 Mushiness 2.35 ±1.70 2.17 ±1.83 1.95 ±1.69 2.82 ±2.34 3.19 ±2.50 3.02 ±2.30 Overall 5.65 ±1.34 5.70 ±1.32 5.49 ±1.43 5.23 ±1.32 5.20 ±1.62 5.39 ±1.60 64 Table 15 Range o f sensory scores o f cultured chinook salmon texture attributes for each ration level X svrimming speed treatment (5 panellists; 9 panel days) Ration Level 7 5 % 100% SvvTmming Speed (bl/s) Attr ibute 0.5 1.0 1.5 0.5 1.0 1.5 Moistness 1.1-9.1 1.5-9.1 0.8-8.8 1.9-9.1 0.6-9.2 0.7-9.0 Powderiness 0-6.8 0-7.6 0-8.5 0-8.7 0-8.4 0-7.9 Flakiness 0.2-8.3 0-7.2 0-7.9 0.1-8.2 0.1-8.4 0-7.2 Firmness 1.5-8.0 0.6-8.8 2.2-7.5 0.6-7.7 0.5-7.9 0.6-8.4 Chewiness 0.1-8.3 0-9.0 0.7-8.5 0.3-8.0 0.2-8.8 0.1-8.3 Cohesiveness 0.8-8.5 0.4-7.6 0.5-7.7 0.2-7.4 0.4-9.0 0.3-8.2 Adhesiveness 0-7.4 0.1-7.7 0.1-7.0 0.1-7.8 0-8.2 0.1-8.3 Mushiness 0.2-7.4 0-8.6 0.1-6.8 0-7.9 0-8.1 0.2-8.0 Overall 3.4-8.7 2.7-8.9 2.0-8.5 1.9-8.6 2.2-8.5 2.2-8.5 65 2.5 2-1.5 4 -0.5 -1 -1.5 -2 3'4 4 i l - J\J 2 L 0.5- 3 c 6 4 « n 2 Oh . Z 4 4 . 3 1 \l 6 ^ -4 -3 -2 -1 0 PC 1 ( 4 4 % ) Figure 2 PC 1 versus PC 2 using sensory aroma attribute scores from panellist 5, data points labelled w i t h treatment numbers 66 0 1 PC 1 (33 %) Figure 3 PC 1 versus PC 2 o f panellist 5 sensory flavour attribute scores, data points labelled w i t h treatment numbers 67 Figure 4 PC 1 versus PC 2 of panellist 5 sensory texture attribute scores, data points labelled with treatment numbers 68 Figure 5 PC 1 versus PC 2 o f panellist 5 sensory aroma attribute scores, data points labelled w i t h panel day number 69 3 -2 -3 9 9 7 o x OO w CM U 7 ? 1 1 13 5 6 ; 0 - 1 7 7 8 8 % 8 8 8 * 2 # 4 4 M 4 9 7 -4 -2 0 1 PC 1 (33 %) Figure 6 PC 1 versus PC 2 of panellist 5 sensory flavour aroma attribute scores, data points labelled with panel day numbers 70 3 cs O 0 -2 -2 8/7 8 4 2*4 0 PC 1 (35 %) Figure 7 PC 1 versus PC 2 o f panellist 5 sensory texture attribute scores, data points labelled w i t h panel day number 71 figures it is apparent that the anomalies in the data are due to events on panel day 9. With the panel day hi-lighted a second cluster, this time in day 7, was also identifiable. Unlike the first set of graphs (Fig. 2-4), these irregularities were also evident in the flavour graphs. When notes recorded during the course of the sensory panel sessions were reviewed, it became apparent that in these two sessions a substitute panellist was used in place of panellist 5 (the same person on both occasions). This irregularity in the judging attributed to one panellist, although unlikely to have a large effect on statistics such as the mean, standard deviation and range of responses, can lead to problems when testing for significant difference in ANOVA. For example, it could lead to panellist X day interactions. When attempting to standardize the data to remove panellist effect, the skewed average and standard deviation for that panellist will, in turn, skew the results. 4.1.2.3 Summary statistics of treated samples The sensory data scores, generally, were quite low. Upon examination of the means (Tables 10, 12 and 14), it was apparent that most were at the lower end of the 10 cm scale; only four had averages over 5 cm, while 11 out of the 28 attributes had an average under 2 cm In Tables 9, 11 and 13, the lower end of the range was often zero. On one or more occasions, this attribute was too faint to be discernible by at least one of the panellists. All the salmon samples had a very delicate flavour. In informal discussions held with the panellists, they would often comment that the flavour notes were quite faint and difficult to quantify. In addition, they commented that they could not discern much of a difference among the six treatments. Many of the sensory attributes proved impossible for individual panellists to even detect in some samples, resulting in a score of zero. 72 4.1.2.4 Three factor A N O V A o f sensory a t t r i bu te da ta A three factor A N O V A was performed on all the attributes individually, to evaluate the contribution o f SS, R L , and panellists as wel l as all the interactions (Tables 16-18). The panellist effect was highly significant (p< 0.001, or p<0.01) for all attributes. Even after training, it is quite common for the panellist effect to account for a large por t ion o f the variation in the data. This variation stems from the subjective nature o f this type o f sensory evaluation and the individual differences between panellists (Stone et al. 1974). There were a few attributes where the R L X P A N interaction was significant. I n overall acceptability, there was also a significant SS X P A N interaction (p<0.05). Even in cases where panellists are screened and wel l trained, there is always a possibility that panellist by treatment interaction w i l l occur due to differences in motivation, sensitivity or psychophysical response behaviour. This is especially true when the panel size is small (Lundahl and McDanieL 1990) as it was in this experiment. Some confusion in scoring is acceptable, particularly in cases such as this where there is not a large degree o f difference between samples (Stone et al. 1974). Out o f the 28 sensory attributes tested, no attributes were significantly affected by SS but eight were significantly affected by varying the R L . O f the aroma attributes tested, "Boi led potato," and "Sour" were significant at the p<0.05 level, and "Seaweedy" at p O . O l . Only one taste term, "Brothy" was found to be significant ( p O . O l ) . Four texture attributes were significantly influenced: "Moistness" (p<0.05), "Firmness" ( p O . O l ) , "Cohesiveness" (p<0.05) and "Mushiness" (pO.OOl ) . 73 Q PH X OO 00 X PH X oo oo PH X 00 00 X PH oo < N ^ H C O ( N O O t - - t - - < N 'C~-ooo\rooroooro co o o <—i co <—i o CN V O U ~ > O \ O \ O S C O 0 0 ( N V O ' - H ' O O O O O O O O •^-r~co*ocococor^ON O N r - i i n c N O O O O V O T j - C N c o ^ o o o o m m o i c o c o o ^ ' o o o O o o oo v~> oo r-~ r-~ M r - ON c-~ c o ^ T f T j - i o r ~ - < i - t — i o o o o o ^ ^ o o * * * * * * * * * * * * * * * * * * * * * * * * * * i - H 00 o CN CN co IT) VO CO CO r - 00 CN CN "ST © o OO CN CO o" oo" V—i in CN <—1 CN T - H CN CN co v o v o c o > o r - v o r t O \ ^ C N V O O N C — m in cs oo < o ^ « n c o T j - c N v o c o o © O © CN >—< CN >—< © ©" * * C N O C N V O V O T f r ^ - O N C N c O © r f - < ^ T a - r O ' - ^ C N v q K d iri d vo © co © © o ^ *3 «o o "3 74 2 PH X GO GO X PH X CO GO PL, x GO GO X PH PH CO GO 1/3 u I S O o »-H o "3- OO ,—1 »n CO o CN o CN VO CN r-NO CO o o ON CO CN in OO <n <—< CN © ~H o o .-H © vo VO CN «n o in oo o CO oo o ON r- 00 < ON ON VO m CO VO CN "3- CN * - H o o © © O © © o © o S 2 o \ t O O O N ^ O O O N £ J - H ' © ' © © " © © © • © © roVOVO^cocNcoON^vo O ^ ^ O O O V O C N t ^ c O t o O O ^ O O C J O ^ CO o VO CO ON 00 CN i — i o o NO VO CN CO VO CO » - H 1—< * - H r-; ON OO CN O © o o <—i © © o »—i * * * * * *• * * * * * * * * * * * * * * * * * * * ON VO in in in VO in CN co CN VO o CN o 00 CN ON »—< m vo ON CO in ON <n © CN* CN © CO CN in ON CO CN NO in 00 in CN ON o in CN t-» CO » - H vo f- r~ •<* in CN OO in >—' <—< CN "3- vo » - H o o © © o O © o o * * ON CN in 00 CO CN CN CO in o m r—i VO 00 VO VO in r - H ON «—1 co r—1 r»- ON O © © © © i - H © CN © © CO •ST >> .fl M "a "S ^ £ o S3 13 ix & P H ( X I W P - I P Q G O I - ) G O G O C Q •t-H O O § i v a &, * !C-.9 ^ i-, "3 in ° 1 in o i v o ^—' m u o ? V ^ & fl * 2 -g in o II V • i U bp B CH on II GO < GO PL, 75 2 X OO X PH X OO OO PH X OO 00 X PH OO OO c/3 I v o o o t - ~ i o r - ~ v o ^ H CN C N I O - « ^ » - H V O I O O J I ^ f~; C N c o C N c o c o c N C N C N <-* • ^ - r ^ ^ H T } - - H ^ f O o o TJ-m c N o o o o m v o ^ H N o r--O O O O O ^ ^ O O c»Nor~cNt~ON<oON 0 0 0 0 0 0 0 0 ^4 • ^ • r O N O c N N O C N O O o N • H ^ ^ t t ^ r H t ^ ^ f f f ) 000<—'OOOfNJ v o o N t - - O N o o < - H O o r ~ o - ^ - O N O O N O m o o i n O O O O* © © O CN ^ H r ~ < — l O N t ^ o o m c N O N O N O O C N © « O O N C N r - ~ H © © o o © m v o ^ O O C N O O O O * <^  fp. 2 <^  r-O NO m r?! CO g C— v vo  V  co O O ON O * CN m oo co O © * CO co CO * * * * * * * * * * •* * * * * * * * * * * •* * * f - CO V) O NO CO ON CN CN CN O VO CN VO VO ON C - <n >-H CN * - H CO ON O VO vo ON VO t - H 1—H CN CO 1 CN » — 1 VO VO oo CN =,•8 1 & > o O < c o ^ J H HO ° \% £ a § 9 § - 2 v * T3 * De © vel oo V • S CH d C * .2 V j H 8 * * H CH 2 oo II O II II V OO CH 3 OO PH * at o 76 4.1.2.5 Z - t rans fo rmat ion o f sensory a t t r ibu tes s igni f icant ly affected by the t rea tment A z-transformation to remove the panellist effect was performed o n the eight attributes that were significantly affected by R L . W i t h the panellist effect removed, analysis o f variance revealed that "Boi led Potato" was no longer significantly affected by R L (Table 19). The F scores from several o f the other attributes, including "Seaweedy," "Sour" (aroma), "Brothy," "Cohesiveness" and "Mushiness" also decreased while those o f "Firmness" and "Moistness" were largely unchanged. 4.1.2.6 P C A and PCS o f sensory data P C A and then PCS were performed on the transformed sensory attributes. The factor score coefficients o f the first 6 PC and the percentage o f the total variance that they explain are listed in Table 20. These six PCs, cumulatively accounting for approximately 9 2 % o f the variation in the data, were fed into the PCS program to calculate slope and coefficient o f determination ( r 2 ) . PCS proved to be an ideal analysis method. First, the use o f PCS is most effective when the individual PC do not account for a large percentage o f the variation (Vodvotz et al., 1993) or the size o f the data matrix is fairly large (Furtula et al. 1994a). The sensory data in this experiment fit both o f these criteria; the first PC only accounted for 40 percent o f the variation and the data set was extremely large. PCS also has the capacity to represent a larger number o f PC scores for graphic illustration than P C A alone (Turtula et al. 1994a) as wel l as giving better group resolution (Vodovotz et al., 1993). When graphs were produced from different pairs o f PCs (not included), no clumping or trends were observed. I n these types o f graphs portions o f the data that may be important are overlooked 77 Table 19 Summarised A N O V A results o f ration level and swimming speed effect on standardised data a from significant sensory attributes o f cultured chinook salmon (5 panellists; 9 panel days) Sensory F ratio Mean attributes Ration Swimming R L X S S square level" speed0 error Seaweedy d 7.840** 0.548 0.833 0.871 Boi led Potato d 3.681 0.684 0.071 1.013 Sour" 5.823* 1.057 0.200 0.992 Brothy 6 4.839* 0.187 0.836 0.998 Mois ture f 4.235* 2.344 0.329 0.987 Firmness f 9.964** 2.334 0.121 0.968 Cohesiveness f 4 . 7 3 1 * 0.113 0.140 1.003 Mushiness f 10.264** 0.728 2.077 0.965 * p<.05, * * p< .01 , * * * p<.001 a Standardised data = panellist effect removed by z transformation o f all scores wi th in each panellist b ration level = 7 5 % and 100% o f ful l ration 0 swimming speed = 0.5,1.0, and 1.5 bl/s d aroma term e f lavour term f texture term 78 Table 20 Factor score coefficients o f the first 6 principal components o f the z-transformed sensory attribute scores o f the R L X SS treated chinook salmon samples (5 panellists; 9 panel days) Sensory attributes Principal Component 1 2 3 4 5 6 Seaweedy 0.161 0.341 0.234 0.822 -0.252 0.524 Boi led potato -0.199 0.266 0.251 -0.347 -0.748 0.307 Sour 0.128 -0.222 0.851 0.049 0.146 -0.458 Brothy -0.069 0.551 0.306 -0.372 0.585 -0.016 Moistness 0.237 0.224 -0.270 -0.070 0.463 0.198 Firmness -0.273 0.099 0.014 0.152 0.239 -0.183 Cohesiveness -0.202 -0.311 0.176 0.121 0.474 1.018 Mushiness 0.236 -0.103 0.118 -0.639 -0.152 0.609 Variance explained Fraction 0.398 0.177 0.119 0.090 0.083 0.061 Cumulative 0.398 0.575 0.694 0.784 0.867 0.928 79 (Vodovotz et al., 1993). I n flavour analysis this is critical because seemingly minor compounds often play critical roles in constituting characteristic flavour notes (Vodovotz et al., 1993). The reference chosen for PCS was a slow SS, high R L (treatment 4) replicate from panellist 1. This treatment was singled out as the reference because, according to Kiessling et al. (1994b), this was the most cost effective o f the six treatments. The particular replicate used for Figures 8-11 gave the best group resolution o f those tested. The PCS results were imported into a spreadsheet, where a series o f graphs o f r 2 versus slope were constructed. The data points were labelled w i t h either the treatment number, ration level number, swimming speed number or the panel day that they represented (Figures 8-11 respectively). These results served to reconf irm that rat ion level, and not swimming speed, was largely responsible for the effect on the sensory properties o f the cooked salmoa I n Figure 9, although a small degree o f overlap is present, the t w o rat ion levels are largely in t w o separate groups. The lower R L was in the lower, left side o f the graph, while the higher R L was primarily in the upper right por t ion o f the graph. I n Figure 10, where the data points are labelled w i t h the SS level, the three swimming speeds appear randomly scattered indicating that SS did not affect the sensory properties o f the salmon. There was no appreciable evidence o f an interaction effect observable from the treatment number graphs (Figure 8). Within the groups o f treatments w i t h the same R L , data from the three SS appear randomly distributed. Some o f the overlapping o f the t w o R L in Figure 9 may have been due to fish to fish variability. Individual fillets were used in this experiment, and not samples compiled from several fish from the same treatment, as in all the other portions o f this research. Fish to fish variation may be partially 80 1.8|— 1.7-1.6-1.5-1.4-1.3-1.2-1.1-1-0.9-0.8-0.7-0.6-0.5-0.4-0.3-0.2-0.1-0L 3 * J 1 31 2 5 2 1 13 4 3 16 2 5 6 1 1 6 1 4 5 4 4 6 4 J I ! I I L i i 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Coefficient o f Determination Figure 8 PCS graph o f significant sensory attributes, data points labelled w i t h treatment numbers 81 0 0.1 0.2 0.3 0.4 0.5 0,6 0.7 0.8 0.9 1 Coefficient of Detennination Figure 9 PCS graph o f significant sensory attributes, data points labelled w i th ration level numbers (1 = 75%; 2 = 100% ration level) 82 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Coefficient of Determination Figure 10 PCS graph o f significant sensory attributes, data points labelled w i t h swimming speed numbers (1=0.5 ,2=1.0 ,3=1.5 bl/s) 83 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Coefficient of Determination Figure 11 PCS graph o f significant sensory attributes, data points labelled w i t h panel day number 84 responsible for this lack o f clear separation that is evident in the G C headspace data. Graphs using the panel day numbers (Fig. 11) showed no obvious day effect. 4.1.2.7 Effect o f SS on the sensory a t t r ibu tes F r o m these results, it appears that changing swimming speed did not result in significant changes in the aroma, taste and texture o f the cooked fish. This conclusion agrees w i t h the results o f Kiessling et al. (1994a,b). Kiessling et al.(1994b), measuring the proximate composit ion o f the fish from this study, did not find that SS had had any effect on the composit ion o f the chinook fillets (Kiessling et al. 1994b). Furthermore, Kiessling et al. (1994a) did not find any evidence o f hypertrophy in the white muscle o f the salmon due to exercise. Addit ionally, no evidence was found to indicate that SS affected the total muscle area. Also, SS did not influence the fibre size distribution. Kiessling et al. (1994a), however, d id find a significant increase in the amount o f fibre hypertrophy in the red muscle that occurred as a result o f the increase in SS. However, since the red muscle is only a fraction o f the size o f the white muscle tissue, it is unlikely that the hypertrophy o f red muscle wou ld have a large impact on the texture o f the fish. 4.1.2.8 Effect o f R L on the sensory a t t r ibu tes R L significantly affected some aspects o f the taste, aroma and texture o f the treated fish samples. This is likely attributable to the significant difference in fat content between the t w o R L noted by Kiessling et al. (1994b). Fat content can affect the mouthfeeL aroma and taste o f a food system. V a n Gemert et al. (1987), work ing w i th smoked sausages, found that there was a strong positive linear relationship between the percentage o f fat in the sausages and their odour intensity. 85 Fat can affect the mouthfeel qualities o f food systems in several ways. First, it lubricates the food while it is being chewed. Addit ionally, fat also imparts an oily sensation in the mouth; this affects the surface tension and can cause changes in the viscosity o f the product (Szczesniak, 1963). 4.1.3 I n s t r u m e n t a l analysis 4.1.4 G C headspace analysis As w i t h the sensory data analyses, A N O V A , P C A and PCS were used to determine what, i f any, effect the treatment had had on the headspace gases. O f the 71 G C peaks that consistently appeared, 27 were significant for either SS or R L (Table 2 1 , Figure 12). Significant peaks were found throughout the graph, and included both large and small peaks. P C A was performed using the areas o f the significant peaks. The factor score coefficients from the first seven PC and the percentage o f total variance explained by each o f these is presented in Table 22. The graphs o f the PCS results, once again using a slow SS, h igh R L treatment for the refernce and labelled w i t h treatment number, ration level number and swimming speed number, (Figures 13-15 respectively) were similar to those produced from the sensory data (Figures 8-11). Again, there is a very clear group resolution o f the data due to ration level (Figure 14). The rat ion level PCS graphs o f both the G C and sensory results showed very similar cluster patterns. I n both, the data points fell into the same pattern; the lower ration level having a coefficient o f determination between 0 and 0.8, w i t h a range in slopes between 0 and 1. The higher ration level fell in the range o f 0.7-1 coefficient o f determination w i t h a slope range o f 0.8-2.0. 86 Table 21 Peak labels, retention times and level o f significance for ration level and svvimming speed o f G C peaks that consistently appeared in purge and trap G C headspace analysis o f cultured, cooked chinook salmon Retention Peak Significance time (minutes) labels RT/ SS b R L X S S 6.69 a * 7.08 b 7.21 c * 7.82 d Me He 8.94 e ** 9.1 f 9.51 g 9.84 h 10.25 i 10.4 j 10.93 k * 10.97 1 12.19 m 12.52 n 12.74 o 13.04 P 13.19 q * 13.49 r 13.68 s * 13.87 t 14.1 u ** 14.5 V 15.04 w * 15.28 X * 15.29 y * 15.49 z * 15.57 aa ** 15.96 ab * 16.05 ac 16.21 ad * 16.68 ae 17.19 a f 17.4 ag * 17.81 ah 18.27 al 18.45 aj 19.18 ak 19.42 al 87 Retention Peak Significance time (minutes) labels R L a SS b R L X S S 20.12 am 20.51 an *** ** ** 20.88 ao * 21.27 ap 21.48 aq * 21.72 ar * 21.97 as 22.32 at 22.42 au 22.62 av 22.71 aw 22.81 ax 23.12 ay 23.27 az * 25.21 ba 26.41 bb *** 26.68 be * 27.27 bd * 27.61 be * * * 28.07 b f 29.16 bg 30.94 bh 31.42 bi 31.51 bj 31.75 bk 31.91 bl 33.14 b m * * 35.03 bn 35.37 bo 36.23 bp 36.58 bq 36.84 br 38.28 bs * a R L = ration level ( 7 5 % and 100% o f ful l ration) b SS = swimming speed (0.5,1.0, and 1.5 bl/s) c Level o f significance * p<.05, * * p< .01 , * * * p<.001 88 89 Table 22 Factor score coefficients o f the first 7 principal components from G C headspace peaks significantly affected by either ration level 3 or swimming speed b Peak Principal Components Label 1 2 3 4 5 6 7 A 0.741 - 0 . 5 4 7 - 0 . 1 0 2 0 .205 0 . 0 8 7 -0 .173 - 0 . 1 8 7 C 0 .735 0 . 5 8 2 - 0 . 1 4 7 0 .008 - 0 . 2 2 5 0 . 1 3 4 0 .131 D 0.841 0 . 3 6 4 -0 .011 0 . 1 0 7 0 . 0 5 4 - 0 . 2 3 8 0 . 2 4 0 E 0 . 9 3 2 0 . 0 9 0 - 0 . 2 8 6 - 0 . 0 4 5 0 . 0 3 8 - 0 . 0 2 9 0 . 0 1 6 K 0 . 7 3 4 0 .591 -0 .001 0 . 0 0 7 - 0 . 2 0 0 - 0 . 1 9 9 0 .175 Q 0 .468 0 . 5 8 7 - 0 . 0 5 7 0 . 3 8 6 0 . 4 1 8 -0 .181 0 .011 s 0 .847 - 0 . 3 2 9 0 . 0 6 7 - 0 . 1 1 7 0 . 1 1 9 - 0 . 0 3 9 -0 .253 u 0 . 7 7 4 0 .183 -0 .441 0 . 0 5 9 0 . 0 8 6 0 .223 - 0 . 1 4 0 w 0 .748 - 0 . 5 9 9 0 .075 - 0 . 1 5 7 0 .055 - 0 . 0 4 2 0 . 1 3 6 X 0 . 5 2 9 - 0 . 4 8 9 0 .115 - 0 . 5 4 7 0 . 2 5 0 0 . 0 2 0 0 . 2 8 6 Y 0 .682 - 0 . 4 2 7 0.301 0 .275 - 0 . 1 8 8 0 .033 0 .288 z 0 . 7 9 6 -0 .423 0 . 0 4 7 0 .215 0 .163 - 0 . 2 3 4 - 0 . 1 9 6 A A 0 . 6 7 6 -0 .543 0 . 1 3 9 0 . 4 1 6 0 .031 0 . 0 2 4 0 . 0 7 9 A B 0 . 5 9 6 0 . 2 9 2 -0 .681 0 . 0 1 6 0 .131 0 . 2 4 7 0 . 0 5 4 A D 0 .673 0 . 5 2 4 0 .135 0 . 1 8 4 0 . 2 9 7 0 . 1 5 0 - 0 . 0 3 0 A G 0 . 7 6 8 0 . 4 8 4 0 . 1 8 9 0 . 0 9 4 - 0 . 2 7 6 - 0 . 0 8 2 - 0 . 1 1 6 A N 0 .713 - 0 . 2 5 7 - 0 . 4 7 7 - 0 . 2 3 2 - 0 . 0 7 7 0 .141 - 0 . 0 8 7 A O 0 . 7 1 0 0 . 1 3 4 -0 .291 - 0 . 3 5 8 - 0 . 1 7 8 -0 .151 -0 .221 A Q 0 .845 -0 .183 0 .271 - 0 . 0 2 6 -0 .121 - 0 . 1 5 5 - 0 . 2 0 7 A R 0 . 9 2 2 0 . 0 4 2 0 .095 - 0 . 0 8 2 -0 .113 0 . 0 1 6 0 . 0 0 4 A Z 0 . 7 8 2 0 .421 0 .163 - 0 . 2 1 0 - 0 . 0 9 9 - 0 . 2 3 9 0 . 0 4 8 B B 0.861 0 .021 - 0 . 0 7 7 -0 .123 0 .125 0 .391 - 0 . 0 5 5 B C 0 . 6 4 4 - 0 . 3 4 4 0 . 4 4 8 0 . 1 6 0 0 . 0 5 9 0 .278 0 . 0 9 7 B D 0.841 - 0 . 0 5 6 0 . 1 2 7 - 0 . 2 3 8 0 .113 - 0 . 1 0 7 0 .161 B E 0 . 4 1 2 - 0 . 5 3 9 -0 .463 0 . 2 6 4 - 0 . 3 8 8 0 .003 0 . 0 6 8 B M 0.161 0 . 2 9 6 0 .863 - 0 . 2 2 8 0 . 0 7 0 0 .011 - 0 . 1 0 8 BS 0 .433 0 . 1 3 4 0 . 7 2 2 0 .081 - 0 . 2 0 2 0 .355 - 0 . 1 4 9 Variance explained Fraction 0 . 5 1 7 0 .155 0 . 1 1 4 0 . 0 5 0 0 . 0 3 4 0 .031 0 . 0 2 4 Cumulative 0 . 5 1 7 0 . 6 7 2 0 . 7 8 6 0 . 8 3 6 0 . 0 8 7 0 .901 0 . 9 2 5 3 rat ion level = 7 5 % and 1 0 0 % o f ful l rat ion b swimming speed = 0 . 5 , 1 . 0 , and 1.5 bl/s 9 0 2.30 1.90 1.50 CL O l . io 4-0.70 0.30 0.70 - t -0 • 76 0 .82 0.88 C o e f f i c i e n t of D e t e r m i n a t i o n — i 0.94 1.00 Figure 13 PCS graph o f significant G C peaks produced from a purge and trap headspace extraction o f cooked, cultured chinook salmon, data points labelled w i t h treatment numbers 91 2 . 3 0 1.90 1.50 1.10 0 . 7 0 0 . 3 0 0 . 7 0 — I 1 1 0 . 7 6 0 .B2 0 . 8 8 C o e f f i c i e n t of D e t e r m i n a t i o n — i 0 . 9 4 1.00 Figure 14 PCS graph o f significant G C peaks produced from a purge and trap headspace extraction o f cooked, cultured chinook salmon, data points labelled w i th ration level number (1 = 75%, 2 = 100% ration level) 92 2 . 3 0 1.90 1 .50 1 . 10 4-0 . 7 0 0 . 3 0 0 . 7 0 0 . 7 6 0 . 8 2 0 . 8 8 C o e f f i c i e n t of D e t e r m i n a t i o n — i 0 . 9 4 1.00 Figure 15 PCS graph o f significant G C peaks produced from a purge and trap headspace extraction o f cooked, cultured chinook salmon, data points labelled w i t h swimming speed number (1 = 0 .5 ,2 = 1.0,3 = 1.5 bl/s) 93 4.1.5 Instron TPA analysis Six T P A parameters were determined from the curves produced by the Instron Universal testing machine (Figures 1 6 - 2 1 ) . T w o factor A N O V A s , looking at the effect o f R L , SS and the interaction o f these treatments, were performed on all six parameters. The results o f these analyses are presented in Table 23. Unl ike the sensory texture results, SS and not R L was significant. Hardness 1, Firmness 1 and 2 were all significantly affected by SS (p<0.05). Dunajski (1979) did not feel that devices used for rheological testing o f food were suitable to measure the texture o f fish muscle. She felt that at best they may be applicable to raw, but not to cooked fish. She pointed out that when fish is eaten, the majority o f the energy is used for mastication As a result, the mechanical measurement should measure the resistance o f the fibres to mechanical disintegration (Dunajski, 1979). I n this experiment, only compression forces were measured. Since the force required to shear the fibres was not measured, this is likely not a true representation o f the texture experienced by the panellists. I t is also possible that the reason for the disagreement between the sensory and Instron results may be due to the way in which the samples were prepared. Samples for the sensory tests composed o f slices o f the fish placed side by side, gelled together into a composite sample upon cooking. For the Instron samples, once cooked and cooled to r o o m temperature, the fish was flaked and formed into a cylinder. I t is possible that these differences in sample temperature and structure contributed to the discrepancy in results. There were many difficulties producing consistent samples. The cylinders o f fish would often fall apart and/or distort, prior to , or in the process o f transportation to the Instron plat form for testing. Additionally, the product ion o f samples o f consistent height proved to be nearly impossible. Fol lowing 94 60 52 44 36 28 20 3 . 4 Trea tmen t N u m b e r Figure 16 Boxplot o f the Instron T P A parameter Hardness 1 for cooked, cultured chinook salmon samples, results by treatment number 95 35 30 « 25 c o (D 2 20 15 L 10 I 3 4 Treatment Number Figure 17 Boxplot o f the Instron T P A parameter Hardness 2 for cooked, cultured chinook salmon samples, results by treatment number 96 10.00 8.40 6.80 5.20 3.60 2.00 3 4 Treatment Number F igure 18 Boxplot o f the Instron T P A parameter Firmness 1 for cooked, cultured chinook salmon samples, results by treatment number 9 7 10.00 8.40 h «2 -S 6.80 f-Z f 5.20 H 3.60 -2.00 L- 1 J - : L I !_ 1 2 3 4 5 Trea tmen t N u m b e r Figure 19 Boxplot o f the Instron T P A parameter Firmness 2 for cooked, cultured chinook salmon samples, results by treatment number 98 0.60 0.48 h 0.37 h 0.25 h 0.14 h 0.02 1 2 3 4 5 6 T r e a t m e n t Number Figure 20 Boxplot o f the Instron T P A parameter Cohesiveness for cooked, cultured chinook salmon samples, results by treatment number 99 I 25 20 h s 1 5 h 10 h 5 h 1 2 3 4 5 6 Treatment Number Figure 21 Boxplot o f the Instron T P A parameter Gumminess for cooked, cultured chinook salmon samples, results by treatment number 100 Table 23 Summarised A N O V A results o f ration level and swimming speed effect o n Instron T P A parameters o f cooked cultured chinook salmon Sensory F ratio Mean attributes Ration Swimming Ration level X square level speed Swimming speed error Hardness 1 0.533 3.888* a 0.229 48.823 Hardness2 0.000 2.343 0.140 17.912 Firmness 1 0.173 5.077* 0.483 4.775 Firmness2 0.226 6.063* 0.512 3.571 Cohesiveness 0.434 2.115 1.222 0.005 Gumminess 0.010 0.507 0.615 0.120 * p<.05, * * p< .01 , * * * p<.001 a Tukey test results: swimming speeds 0.5 bl/s significantly different from 1.0 bl/s (p<0.05) 101 compression o f the sample to the desired height and release o f the plunger on the syringe, the sample wou ld spring back. 4.1.6 p H analysis The post-mortem p H o f fish, as w i th most other animals, is largely due to the degradation o f glycogen to lactic acid v ia the Emden-Meyerhof-Parnas pathway. The ultimate post-mortem p H o f most fish species usually falls in the range o f 6.5 - 6.2 (Dunajski, 1979). The p H o f most o f the fish samples in this experiment was above this; the majority ranging between 6.5 - 6.7 (Figure 22). A l though the p H did not vary widely between treatments, significant differences were found (Table 24). Highly significant statistical differences (p<.001) in the p H were found to have resulted from both R L and SS. The interaction between R L and SS also proved to be highly significant. F r o m the box plot showing the effect o f the 6 treatment combinations o n p H (Figure 22), some general trends become apparent. First, the fish that received the higher R L (treatments 4-6) generally had a lower p H when compared w i t h the salmon fed the lower rat ion level. A small increase in the p H also occurred when the swimming speed was increased. I t is unfortunate that no data is available as to the glycogen content o f the salmon used in this study. Kiessling et al. (1989b), in a study where fish were fed different ration levels, found an increase in the glycogen content o f fish muscle when the fish ration level was increased in the range between 5 0 % -100% R L . This appears to have also occurred in this experiment; there was a decrease in the post mor tem p H o f the fish at the higher ration levels that wou ld be consistent w i t h an increased glycogen content in the live fish pr ior to their being sacrificed. 102 32 CL 6.85 6.76 6.67 6.58 6.49 r -6.40 3 4 Treatment Number Figure 22 Boxplot by treatment number o f the p H o f t h e cooked chinook salmon samples 103 Table 24 A N O V A table o f ration level and swimming speed effect on the p H o f cultured cooked chinook salmon Source D F SS M S F P SS 2 0.054 0.027 49.189 0 . 0 0 0 a R L 1 0.015 0.015 27.273 0.000 S S * R L 2 0.097 0.048 87.977 0.000 Error 18 0.010 0.001 Tukey test results: swimming speeds 0.5 bl/s significantly different from 1.0 bl/s ( p O . 0 0 1 ) 104 A slight increase in the p H that corresponds to an increase in the SS o f the fish is also seen in Figure 22. There is no readily apparent explanation for this small increase in p H w i t h the corresponding increase in swimming speed. For an unknown reason, the replicates from treatment number 4 (SS = 0.5 bl/s, R L = 100%) were significantly higher than all but treatment number 3 (SS = 1.5 bl/s, R L = 75%). When the effects o f SS and R L were looked at individually, the data from this treatment combination skewed the results. I t is w o r t h noting that no similar deviations in the results are observed for this treatment in either the sensory, Instron or G C results. 4.1.6.1 Comparison of pH and sensory analysis results According to Dunajski (1979) and Love (1988) the p H o f the fish muscle is likely the most important factor affecting its rheological properties. W i t h a drop in the p H , theoretically there should be an increase in the toughness o f the fish. Thus, the drop in the p H o f the cooked fish muscle that resulted when the R L was increased should have caused these well-fed fish to have a slightly firmer eating texture. I n the sensory texture testing, however, although there was a significant difference in the texture o f the fish due to R L , the opposite trend was observed. The fish reared o n the 7 5 % R L were significantly more firm and less mushy, despite their higher p H , than the fish reared on the 100% R L . Similarly, w i t h a lower p H , it is expected that one wou ld find the fish to be drier (Dunajski, 1979). Here too, the results were contrary to theory; in the sensory testing, the higher ration level fish samples were significantly more moist than those from their lower ration counterparts. There are t w o factors that may have contributed to this contradiction o f the popular theory regarding the relationship between p H and the texture o f the fish. First, ignoring the treatment number 105 4 ( R L = 1 0 0 % X . SS=0.5 bl/s) data, the change in p H is small, approximately 0.2 p H units. Second, the fish reared w i th the higher R L had a significantly higher fat content than those at the lower R L (Kiessling et al. 1994b). This higher fat content wou ld have caused a higher degree o f lubrication and an oi ly sensation in the mouth (Szczesniak, 1963). This wou ld result in the sensory perception o f a more moist, less tough or firm sample. 106 5. Conclusions 5.1 Sensory Analyses R L and panellist effect significantly affected the sensory analysis port ion o f this experiment, while SS did not. Eight sensory attributes (three aroma, one flavour, and four texture) were significantly affected by ration level; no attributes were affected by swimming speed. The PCS graphs (Figures 8 and 9) , using the PC from the seven significant sensory attributes, graphically depicted this phenomenon. The panellist effect was highly significant for all the attributes. After completing a z-transformation to remove the panellist effect, and a second A N O V A was performed, one aroma term "Boi led Potato" was no longer significant. These results are consistent w i t h those o f Kiessling et. al. (1994b). Kiessling et al. (1994b) did not find that SS had any effect o n the proximate composit ion o f the fish. They, however, did observe that the fat content o f the RL100 samples were significantly higher than those o f the RL75 samples. This difference in fat content o f the fish samples was likely responsible for the significant R L effect observed in the sensory analyses results. I t could explain both changes in both the concentrations o f flavour volatiles, affecting the aroma attributes "Seaweedy" and "Sour," and flavour attribute "Bro thy , " and the mouthfeel o f the samples affecting the perception o f the texture attributes "Mois ture, " "Firmness," "Cohesiveness" and "Mushiness." The statistical analyses o f the sensory data also clearly demonstrated that the use o f replacement panellists in Q D A sensory analysis should be avoided. I n this type o f sensory analysis, despite training towards uniformity, each panellist develops their o w n unique style o f grading the samples. A s long as the panellist is consistent it is possible to remove the panellist effect during statistical analysis. Figures 2-7 illustrate that when the replacement panellist was used o n panel days 7 107 and 9, that irregularities due to the substitution were evident in the data. When a replacement panellist is used on one or more panel sessions, as it was in this experiment, the removal o f panellist effect is compromised. 5.2 G C headspace analyses Out o f the 71 G C peaks that consistently appeared, 27 were found to be significant for either SS and/or R L (Table 21). A n appreciable number o f peaks were significantly affected by R L , while comparatively fewer were significant for SS. The PCS graph o f these significant peaks clearly shows clear group resolution on the basis o f R L (Figure 14), while no trend was observable when the PCS graph data points were labelled w i th either SS levels (Figure 15) or treatment numbers (Figure 13). These results compare favourably w i t h both the sensory analysis por t ion o f this experiment and the findings o f Kiessling et. al (1994b). 5.3 Ins t ron TPA The Instron T P A results did not fo l low the trend established in the other areas o f this experiment where the R L significantly affected the sensory properties o f cooked, cultured chinook salmon, and SS did not. Rather, the Instron T P A results showed a significant difference due to SS, while R L had no effect. I t is possible that the compression method o f texture measurement, employed in this experiment, measured changes in the cooked muscle that were not discernible by the panellists, and was unable to account for the textural changes due to the differences in fat content between the t w o ration levels. I n the mastication o f cooked fish, the majority o f the energy is used for mastication. A s a result, to get an accurate instrumental measurement o f the texture o f the fish, the mechanical 108 measurement should measure the resistance o f the fibres to mechanical disintegration (Dunajski, 1979), which was not the case in this experiment. Other possible reasons for the discrepancy in results include: differences in method o f sample preparation and extreme difficulty in sample preparation. 5.4 pH The p H port ion o f this experiment was sensitive to both the changes in R L and SS, showing clear trends in both. There was a highly significant statistical difference found in SS, R L and the R L X . SS interaction. I n this experiment, the drop in muscle p H did not result in a firmer texture, but rather, the opposite trend was observed. I n the sensory analysis port ion o f this study, as the p H dropped, the fish became more mushy and they were rated lower in firmness. This trend may be attributable to a corresponding increase in fat content. The unexplainably high results o f treatment 4 badly skewed the p H data. F r o m the Tukey test results as wel l as the highly significant R L X . SS interaction, it became apparent that the significant difference due to SS was because o f this aberration in the data and was not evidence for a true SS effect. 5.5 Overa l l Conclusions This experiment demonstrated that although changing the rat ion level w i l l have an effect on the sensory attributes o f cooked chinook salmon muscle, the swimming speed o f the fish does not. As a result, increasing the swimming speed o f chinook salmon in a fish farming operation above that which is necessary for proper schooling, while decreasing the food conversion rate, is unlikely to result in any 109 appreciable difference in consumer acceptability. The significant effect due to ration level observed in the sensory and G C headspace analyses was likely attributable to the differences in fat content between the RL75 and R L 100 samples. 110 References Anonymous. 1963. The flavour profile. Ar thur D. Lit t le, Cambridge, M A . Bourne, M.C. 1968. Texture profile o f ripening pears. J. Food Sci. 33: 223-226. Bourne, M.C. 1978. Texture profi le analysis. Food Technol. 32 (7) : 62-66, 72. Bourne, M.C. 1982. Food Texture and Viscosity. Academic Press, N e w York . pp. 176-185. Borderias, A.J. , Lamura, M . , and Tejada, M . 1983. Texture analysis o f fish fillets and minced fish by both sensory and instrumental methods. J. Food Technol. 18: 85-95. Brandt, M.A. , Skinner, E.Z., and Coleman, J.A. 1963. Texture profi le method. J. Food Sci. 28: 404-409. Broughton, N . M . and Goldspink, G. 1978. Biochemical changes in the lateral muscles o f roach Rutilus rutilus from t w o habitats fol lowing exercise. J. Fish Bio l . 13: 613-618. Broughton, N . M . , Goldspink, G. and Jones N.V. 1980. The effect o f training on the lateral musculature o f o-group roach Rutilus rutilus (L . ) and their fatigue in subsequent exercise tests. J. Fish Bio l . 17: 209-217. Butler, L .D. and Burke, M.F. 1976. Chromatographic characterisation o f porous polymers for use as adsorbents in sampling columns. J. Chromatog. Sci. 14: 117-122. Cairncross, S.E. and Sjostrom, L.B. 1950. Flavour profiles-a new approach to flavour problems. J. Food Technol. 4: 308-311. Christiansen, J.S. and Jobling. M . 1990. The behaviour and the relationship between food intake and g rowth o f juvenile Arct ic charr, Salvelinus alpinus L., subjected to exercise. Can. J. Zoo l . 68 :2185-2191. Christiansen, J. S., Ringo, E. and Jobling, M . , 1989. Effects o f sustained exercise on g rowth and body composit ion o f first-feeding fry o f Arct ic charr, Salvelinus alpinus (L . ) Aquaculture, 79: 329-336. Davie, P.S., Wells, R.M.G. and Tetens, V . 1986. Effects o f sustained swimming o n rainbow trout muscle structure, blood oxygen transport, and lactate dehydrogenase isozymes: evidence for increased aerobic capacity o f white muscle. J .Exp . Zoo l . 237 :159-171 . Davison, W. 1989. M i n i review: Training and its effects on teleost fish. Comp. Biochem. Physiol. 94A: 1-10. I l l Davison, W. and Goldspink, G. 1978. The effect o f training on the swiinming muscles o f the goldfish (Carassius auratus). J. Exp. Bio l . 74:115-122. Davison, W. and Goldspink, G. 1977. The effect o f prolonged exercise on the lateral musculature o f the b rown trout (Salmo truttd). J .Exp . B io l . 70: 1-12. Dunajski, E. 1979. Texture o f fish muscle. J. Texture Studies 10: 301-318. Durance, T .D. , and Collins, L.S. 1991. Quality enhancement o f sexually mature chum salmon Oncorhynchus keta in retort pouches. J. Food Sci. 56: 1282-1286. East, P. and Magnan, P. 1987. The effect o f locomotor activity on the g rowth o f brook charr, Salvelinus fontinalis Mi tchi l l . Can. J. Zool . 65: 843-846. Farlinger, S. and Beamish, F.W.H. 1978. Changes in blood chemistry and critical swimming speed o f largemouth bass, Micropterus salmoides, w i th physical conditioning. Trans. A m . Fish. Soc. 107: 523-527. Feinstein, G.R. and Buck, E . M . 1984. Relationship o f texture to p H and collagen content o f yel lowtai l f lounder and cusk. J. Food Sci. 49: 298-299. Furtula, V . , Nakai , S., Amantea, G.F., and Laleye, L. 1994a. Reverse-phase FfPLC analysis o f cheese samples aged by a fast-ripening process. J. Food Sci. 59: 328-332, 567. Furtula, V . , Nakai , S., Amantea, G.F., and Laleye, L. 1994b. Reverse-phase H P L C analysis o f reference cheddar cheese samples for assessing accelerated cheese ripening. J. Food Sci. 59: 333-338. Gacula, M . Jr. 1987. Some issues in the design and analysis o f sensory data: revisited. J. Sensory Studies 2: 169-185. Gilbert, J. 1990. Gas chromatographic analysis o f contaminants in food. In . Principles and Applications o f Gas Chromatography in Food Analysis. Ell is H o r w o o d L td . , West Sussex England, pp. 220. Giovanni, M.E. and Pangborn, RJM. 1983. Measurement o f taste intensity and degree o f l iking o f beverages by graphic scales and magnitude estimation. J. Food Sci. 48:1175-1182. Greer Walker, M . and Pull, G. 1973. Skeletal muscle function and sustained swimming speeds in the coalfish Gadus virens L. Comp. Biochem. Physiol. 44A: 495-501. Greer Walker, M , and Emerson L. 1978. Sustained swimming speeds and myotomal muscle function in the trout, Salmo gairdneri. J. Fish BioL 13 :475-481. 112 Hammond, B.R. and Hickman, C P . Jr. 1966. The effect o f physical conditioning on the metabolism o f lactate phosphate and glucose in rainbow trout, Salmo gairdneri. J. Fish Res. B d Can. 23 :65-83. Hatae, K., Yoshimatsu, F., and Matsumoto, J.J. 1990. Role o f muscle fibres in contributing firmness o f cooked fish. J. Food Sci. 55: 693-696. Heikes, D.L. 1985. Purge and trap method for determination o f ethylene dibromide in table-ready foods. J. Assoc. Off. Anal. Chem. 68: 431-436. Heikes, D.L., and Hopper, M X . 1986. Purge and trap method for determination o f fumigants in whole grains, milled grain products and intermediate grain-based foods. J. Assoc. Off. Anal. Chem 69: 990-998. Hochachka P. W. 1961. The effect o f physical training on oxygen debt and glycogen reserves in trout. Can. J. Zool . 39: 767-776. Houlihan, D. F. and Laurent, P. 1987. Effects o f exercise training on the performance, growth, and protein turnover o f rainbow trout (Salmon gaidneri). Can. J. Fish. Aquat. Sci. 44: 1614-1621. Jennings, W.G. and Filsoof, M . 1977. Comparison o f sample preparation techniques for gas chromatographic analysis. J. Agr ic. Food C h e m 25: 440-445. Johnsen, P.B. and Kelly, C A . 1990. A technique for the quantitative sensory evaluation o f farm-raised catfish. Journal o f Sensory Studies. 4 : 1 8 9 - 199. Johnston L A . and M o o n T .W. 1980a. Exercise training in skeletal muscle o f brook trout (Salvelinusfontinalis). J. Exp. Bio l . 87:177-194. Johnston, L A . and M o o n , T .W. 1980b. Endurance exercise training in the fast and slow muscles o f ateleost fish (Pollachius virens). J. Comp. Physiol. 135B: 147-156. Johnston, I. A . , Davison W. and Goldspink G. 1977. Energy metabolism o f carp swimming muscles. J. Comp. physiol. 114B: 203-216. Josephson, D.B., Lindsay, R.C., and Stuiber, D.A. 1983. Identif ication o f compounds characterising the aroma o f fresh whitefish (Coregonus clupeaformis). J. Agr ic. Food C h e m 3 1 : 326-330. Josephson, D.B., Lindsay, R.C., and Stuiber, D.A. 1985. Volati le compounds characterising the aroma o f fresh Atlantic and Pacific oysters. J. Food Sci. 50: 5-9. Josephson, D.B., Lindsay, R.C., and Stuiber, D.A. 1991. Influence o f maturity on the volatile aroma compounds from fresh Pacific and great lakes salmon. J. Food Sci.. 56:1576-1585. 113 Kanoh, S., Polo, J .M. A . , Kariya, Y . , Kaneko, T., Watabe, S., and Hashimoto, K. 1988. Heat- induced textural and histological changes o f ordinary and dark muscles o f yel lowf in tuna. J. Food Sci. 53: 673-678. Kiessling A , Johansson, L. and Storebakken, T., T., 1989a. Effects o f reduced feed ration levels on fat content and fatty acid composit ion in whi te and red muscle from rainbow trout. Aquaculture 79: 169-175. Kiessling, A , Storebakken, T., Asgard, T., Andersson, I.L. and Kiessling, K.H. , 1989b. Physiological changes in muscle o f rainbow trout fed different ration levels. Aquaculture 79:293-301. Kiessling, A , Storebakken, T. , Asgard, T . and Kiessling K .H . , 1991a. Changes in the structure and function o f the epaxial muscle o f rainbow trout (Oncorhynchus mykiss) in relation to ration and age. I. G rowth dynamics. Aquaculture 93:335-356. Kiessling, A , Asgard, T., Storebakken, T., Johansson, L. and Kiessling, K . H . 1991b. Changes in the structure and function o f the epaxial muscle o f rainbow trout (Oncorhynchus mykiss) in relation to ration and age. JJI. Chemical composition. Aquaculture 93:373-387. Kiessling, A , Gallaugher, P., Thorarensen, H , Ko lok , A , Eales, J.G., Sweeting, R., Gong, B., M c K e o w n , B.A., Dosanjh, B., FarrelL A.P. , Higgs, D. 1994a. Influence o f sustained exercise and endurance training on growth, muscle physiology, cardiovascular parameters, and plasma levels o f metabolic hormones o f seawater adapted all-female chinook salmon. In , H i g h Performance Fish: Proceedings o f the International Fish Physiology Symposium. July 16-21, 1994. (Ed) D.D. Mackinlay. Fish Physiol. Assoc. Vancouver, B.C. pp.300-305. Kiessling, A , Higgs, D.A., Eales, J.G. and Dosanjh, B.S. 1994b. Influence o f sustained exercise at t w o ration levels on g rowth and thyroid function o f all-female chinook salmon (Oncorhynchus tshawytscha) in seawater. Can. J. Fish. Aquat. Sci. 5 1 : 1975-1984. Larmond, E. 1977. Laboratory methods for sensory evaluation o f food. Canadian Government Publishing Centre. Ottawa, Canada, pp 6-19. Leon, K.A. 1986. Effect o f exercise on feed consumption, growth, food conversion, and stamina o f brook trout. Prog. Fish Cult. 48 :43-46 . L i , M . and Lovett, R.T. 1992. Comparison o f satiate feeding and restricted feeding o f channel catfish w i t h various concentrations o f dietary protein in product ion ponds. Aquaculture 103: 165-175. Love, R.A. 1988. Individual influences. In , The Food Fishes - their intrinsic variation and practical implication. Farrand Press, London, pp. 183-212. 114 LundahL D.S. and McDanieL M.R. 1990. Use o f contrasts for the evaluation o f panel inconsistency. J. Sensory Studies 5: 265-277. M a , P. 1992. Personal communication. Statistical consulting and research laboratory. Dept. o f Statistics. Univ. o f Bri t ish Columbia, Vancouver, B.C. Meilgaard, M . , Civille, G.V., and Carr, B.T. 1991. Descriptive analysis techniques. In , Sensory Evaluation Techniques. C R C Press, Boca Ragon, F L . pp 1 8 8 - 2 0 0 . Olafsdottir, G., Steinke, J.A. and Lindsay, R.C., 1985. Qualitative performance o f a simple Tenax-G C adsorption method for use in the analysis o f aroma volatiles. J. Food Sci. 50:1431-1436. Pangborn, R . M . 1987. Sensory science in flavor research: achievements, needs and perspectives. In , Flavour Science and Technology: Proceedings o f the 5th Weurman Flavor Research Symposium, held at Sara H o t e l Volsenasen. Oslo. 23rd - 25th March. 1987. (Eds) M . Martens, G. A . Dalen, and H. Russwurm Jr. John Wiley & Sons L td . Chichenster, N e w York , pp. 275-300. Patterson R.H. and Goldspink, G. 1976. Mechanism o f myofibr i l g rowth and proliferation in fish muscle. J. Cell Sci. 22: 607-616. Persson, T. and v o n Sydow, E. 1973. The aroma o f canned beef: gas chromatographic and mass spectrometric analysis o f the volatiles. J. Food Sci. 38: 377-385. Persson, T., von Sydow, E. and Akesson, C. 1973a. A roma o f canned beef: sensory properties. J. Food Sci. 38: 386-392. Persson, T., von Sydow, E. and Akesson, C. 1973b. A r o m a o f canned beef: models for correlation o f instrumental and sensory data. J. Food Sci. 38: 682-689. Pickering, A D . 1990. Stress and the suppression o f somatic g rowth in teleost fish. Progress in Comparative Endocrinology, Wiley-Liss, Inc. N e w York , N e w York , pp 471-473. Powers, J.J., and Kei th, E.S. 1968. Stepwise discriminant analysis o f gas chromatographic data as an aid in classifying the flavor quality o f foods. J. Food Sci. 33: 208-213. Reid, R.A. and Durance, T .D. 1992. Textural changes o f canned chum salmon related to sexual changes. J. Food Sci. 57:1340-1342. Rutledge, K.P. and Hudson, J .M. 1990. Sensory evaluation: method for establishing and training a descriptive flavor analysis panel. Food Technol. 44(12): 78-84. 115 Segars, R.A. and Kapsalis, J.G. 1987. Chapter 7: Texture, rheology, psychophysics in objective methods. In , Food Quality Assessment. J.G. Kapsalis (Ed). CRC press, Boca Raton, FL . pp.155-181. Shamaila, M . , Skura, B.J., and Nakai , S. 1995. Moni tor ing quality o f packaged Pacific ocean perch (Sebastes alutus) in storage. J. Aquatic Food Prod. Technol. 4: 35-57. Stone, H. , SideL J., Oliver, S., Woolsey, A . , Singleton, R.C. 1974. Sensory evaluation by quantitative descriptive analysis. Food Technol. 28(11): 24-34. Storebakken, T., Hung, S.S., Calvert, C.C. and Plisetskaya, E .M. , 1991. Nutrient partit ioning in rainbow trout at different feeding rates. Aquaculture 96: 191-203. Szczesniak, A S . 1963. Classification o f textural characteristics. J. Food Sci. 28: 385-389. Szczesniak, A S . 1990. Texture: Is it still an overlooked food attribute? Food Technol. 44(9): 86-95. van Gemert, L.J., Maarse, H., Thissen, J .T .N.M. 1987. Relation between organoleptic attributes and chemical, physical and instrumental parameters o f smoked sausages. I n , Flavour Science and Technology: Proceedings o f the 5th Weurman Flavor Research Symposium held at Sara H o t e l Volsenasea Oslo. 23rd - 25th March. 1987. M . Martens, G.A. Dalen, and H . Russwurm Jr (Ed). John Wiley & Sons L t d . Chichenster, N e w York . pp. 439-452. Vijayan, M . M . , and Leatherland, J.F. 1989. Cortisol-induced changes in plasma glucose, protein, and thyroid hormone levels, and liver glycogen content o f coho salmon (Oncorhynchus kisutch Walbaum). Can. J. Zoo l . 67: 2746-2750. Vodovotz , Y. , Arteaga, G.E., and Nakai , S. 1993. Principal component similarity analysis for classification and its application to G C data o f mango. Food Res. Int. 26: 355-363. v o n Sydow, E., Andersson, J., An jou , BC, Karlsson, G., Land, D. and Griff iths, N . 1970. The aroma o f bilberries (Vaccinium myrtillus L.). 2. Evaluation o f the press juice by sensory methods and by gas chromatography and mass spectrometry. Lebensm-Wiss. u. Technol. 3: 11-17. Weatherley, A H . and G i l l H.S. 1981. Characteristics o f mosaic muscle g rowth in rainbow trout Salmo gairdneri. Experimentia 37: 1102-1103. Weatherley A H . and Gi l l H.S. 1984. Growth dynamics o f white myotomal muscle fibres in the bluenose minnow Pinephales notatus Rafinesque and comparison w i th rainbow trout Salmo gairdneri Richardson. J. Fish Bio l . 25: 13-24. White, J.R. and L I H.W. 1985. Determination o f the energetic cost o f swimming from the analysis o f g rowth rate and body composit ion in juvenile chinook salmon (Oncorhynchus tshawytschd). Comp. Biochem. Physio l 81 A : 25-33. 116 Williams, A . A . and Tucknot, O. 1977. Misleading information from G C effluents in aroma analysis. Chem. & I n d . 3:125-129. Woodward , J.J., and Smith, L.S. 1985. Exercise training and the stress response in rainbow trout, Salmo gairdneri Richardson. J. Fish Bio l . 26: 435-447. 117 Appendix A: Samples of sensory exploratory analysis boxplots 118 IT) OO OO © 00 00 © ' oo 00 ssaiunos %SL TH ssaumos %00l TM 1 « CO 4> VO >H OH OJ a £ •8.8 11 S OH OH © O 1) l i s o • tfa vo * J< •a ^ .S J$ *§ 3 •8 J f 8 O OO »H eg O PH co o o a CO 1 O o CO B OH EH CO o « ? • s OX) 119 1 oo oo t/3 o oo oo s © 00 00 %SZ. TH %00l TH r--i l g OH 1.8 O OJ 60 OH "53 -2 a ? co I s 1 OH 42 s i o § JH .a I 0 JH 1 O M £ & « " s »i I 5 £ a 5 g £ 6 II | e 8 2 o S3 120 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0087998/manifest

Comment

Related Items