Ž .
analyses of variance ANOVA were carried out in each year, based on individual fish means for the various attributes. The effects of sessions were removed in the analyses
providing a total of 105 and 63 residual degrees of freedom, respectively, for the 2 Ž
. years. When significant P - 0.05 differences were found between the sources, Dun-
can’s multiple range test at a probability level of P s 0.05 was used to establish Ž
. significant differences O’Mahoney, 1986 . Data for hedonic scaling were also analysed
Ž by ANOVA and Duncan’s Test. While data obtained by hedonic scaling or, for that
. matter, profiling
may not be strictly linear and therefore may not fulfil all the assumptions for these parametric tests, they are still useful methods for the interpretation
Ž .
of such data O’Mahoney, 1986 . This should be less of a problem for the profiling data as reference to the Central Limit Theorem implies that the assumption of a Normal
Distribution for means over panellists may be reasonable. However, this is not the case for the hedonic scaling data and, accordingly, these were also analysed by the nonpara-
metric Kruskal–Wallis rank sum test which does not assume normality of the observa-
Ž .
tions. This showed significant differences P - 0.05
for appearance, flavour and aftertaste in agreement with the parametric tests. Hence, this paper concentrates on the
results of the parametric tests. Correlation coefficients were calculated to determine whether there was any relation-
ship between the mean scores obtained by hedonic scaling for acceptability and the mean scores for each profiling attribute for the six salmon sources assessed in 1993.
Regression analysis using quadratic terms was also conducted to check for possible curvilinear relationships.
Ž .
Principal components analysis PCA was conducted on the mean profiling scores for each individual salmon assessed, for both years of study, using the Genstat 5 statistical
package and based on the sums of squares matrix between the attributes as the observations were on a common scale.
3. Results
3.1. Comparison of fish from different sources Analysis of the sensory profiling results by ANOVA shows that 25 out of 32
attributes in the 1993 trial, and 32 out of 37 attributes in 1994, gave significant Ž
. differences between two or more sources Tables 3 and 4 . The attributes which did not
differ significantly between sources included oily odour, salty odour, beige colour, fishy flavour, smoky flavour, springy texture and clingy texture in the 1993 trial and salty
odour, peach colour, paleness, crumbly texture and clingy texture in the 1994 trial. Inspection of the data for the above nonsignificant attributes indicates that, for salty
odour, fishy and smoky flavour, and crumbly and clingy texture, the source means were similar; in addition, scores for the first three of these attributes were less than 16. For
beige and peach colour, and paleness and springy texture there were large differences between the source means, but the variability between individual fish was such that
these differences were not significant.
Table 3 Effect of salmon source on mean scores for eating quality attributes for salmon harvested during 1993. Scores
Ž .
are on a scale of 0 to 100 increasing intensity
†
Attributes Farm 1 Farm 2 Farm 3 Sea 2
River 1 Rivers 2r3 Significance SEM
Odour
UUU
‡ a
a a
a b
a
Salmon-like odour 27.3
26.1 28.1
25.4 19.5
26.0 0.87
UUU
a a
a a
c b
Earthy odour 20.4
19.0 19.4
18.3 32.7
25.5 1.53
UUU
ab ab
a ab
c b
Stagnant odour 17.5
16.7 16.2
19.9 34.1
22.1 1.75
Appearance
U
a bc
c abc
ab c
Pink 11.3
14.1 14.8
13.3 11.6
14.5 0.81
UUU
c a
b a
bc ab
Peach 41.6
28.2 34.9
27.2 36.5
31.4 1.71
UUU
a bc
a c
ab abc
White 10.4
14.8 10.9
15.1 12.2
13.0 0.85
UUU
d a
bcd ab
cd abc
Orange 19.6
10.8 16.0
11.6 17.6
13.5 1.38
UUU
c bc
a c
ab ab
Coagulation 25.5
23.0 18.5
25.3 19.9
21.3 1.23
UU
ab a
b a
a a
Juicy appearance 27.7
26.0 31.7
24.3 25.0
25.4 1.28
UU
a a
b a
a a
Separation 23.3
23.3 26.9
23.0 21.4
21.1 1.03
UU
ab bc
c ab
ab a
Moistness 26.2
28.6 31.0
25.1 25.8
24.4 1.24
U
a ab
b a
ab ab
Flaky 29.1
30.7 33.4
29.8 31.0
30.5 0.93
FlaÕour
UUU
ab b
b b
c a
Salmon-like flavour 28.6
31.3 30.9
30.8 17.3
27.4 0.93
U
ab ab
b ab
ab a
Oily flavour 17.8
17.0 19.9
17.8 16.7
14.8 1.04
UUU
bc ab
ab a
d c
Earthy flavour 30.0
24.5 28.5
23.3 49.5
35.4 1.73
Texture
UUU
a b
c ab
a ab
Moisture 30.0
35.4 40.5
31.9 28.3
31.3 1.58
UUU
ab b
c ab
a ab
Tenderness 44.5
50.4 57.7
44.3 39.9
44.6 1.90
UUU
a b
b a
a a
Lightness 39.7
45.4 49.1
40.4 38.3
40.4 1.31
UUU
bc ab
a bc
c b
Chewiness 36.3
31.7 27.3
35.7 38.4
33.6 1.46
UU
ab bc
c a
ab bc
Flakiness 31.9
33.7 36.0
30.2 32.3
35.1 1.02
Aftertaste
UUU
ab ab
ab b
c a
Chickeny aftertaste 12.6
14.5 12.7
15.9 8.7
12.4 1.05
UUU
a a
a a
c b
Earthy aftertaste 24.6
20.1 23.5
20.7 43.0
31.1 1.55
UUU
a a
a a
b a
Metallic aftertaste 15.9
14.2 14.0
15.3 23.3
17.6 1.16
UU
b b
b b
a ab
Time when aftertaste starts 18.2 21.6
19.2 18.0
13.1 17.2
1.49
UUU
a a
a a
c b
Strength of aftertaste 35.5
32.7 34.5
36.4 52.1
43.3 1.52
†
Significance of results calculated by ANOVA.
‡
Ž .
Scores without common superscripts within a horizontal line differ significantly P - 0.01 according to Duncan’s test.
U
P - 0.05.
UU
P - 0.01.
UUU
P - 0.001.
The results of PCA show which attributes gave the greatest discrimination between Ž
. the mean scores for individual salmon Figs. 1 and 2; results for PC3 are not illustrated .
For the 1993 salmon, PC1, PC2 and PC3 account for 34.6, 16.4 and 9.5 of the variability, respectively; while, for the 1994 salmon, PC1, PC2 and PC3 account for
34.7, 14.2 and 13.2 of the variability.
The additional inclusion of age, weight, length, weightrlength and storage time in the PCA did not show any relationships between these parameters and the sensory scores
Table 4 Effect of salmon source on mean scores for eating quality attributes for salmon harvested during 1994. Scores
Ž .
are on a scale of 0 to 100 increasing intensity
†
Attributes Farm 1 Farm 2 Farm 3 Sea 1
Sea 2 River 1 River 2 River 3 Significance SEM
Odour
U
‡ b
b ab
ab b
a ab
a
Salmon-like odour 24.9
25.1 24.7
22.9 25.2
21.0 22.6
20.9 1.17
UUU
ab a
ab ab
ab c
bc c
Earthy odour 27.8
25.8 26.8
27.2 27.7
32.0 30.0
33.8 1.24
UUU
a a
a a
a b
a b
Stagnant odour 25.4
23.5 25.8
23.6 24.4
30.2 26.0
30.3 1.10
UUU
a a
a a
a a
a b
Farmyard odour 20.8
20.2 19.2
20.5 23.4
23.5 22.7
29.8 1.48
UUU
c c
c bc
c ab
bc a
Oily odour 27.1
26.6 27.1
25.7 26.4
23.1 25.1
21.6 0.92
Appearance
UUU
bc c
c bc
bc ab
ab a
Juicy appearance 30.5
34.7 33.4
30.7 30.9
26.4 27.5
25.4 1.51
U
abc bc
c abc
abc ab
ab a
Separation 28.0
28.4 29.8
27.8 26.4
25.6 25.9
24.8 1.05
UUU
a a
a ab
b c
a ab
Coagulation 25.8
26.4 26.3
27.6 32.0
21.0 25.9
28.3 1.39
UU
a bcd
bcd abc
bcd ab
cd d
Pink 13.4
16.9 16.9
16.3 17.6
14.6 18.4
19.7 1.00
U
b a
ab ab
a a
ab a
Orange 21.5
15.6 18.6
18.2 16.6
15.1 17.9
16.8 1.31
UUU
abcd bcde
abc cde
de e
ab a
Beige 22.1
24.2 22.0
26.0 26.1
26.5 21.4
19.1 1.24
UUU
bc c
c bc
bc bc
ab a
Moist appearance 32.9
34.9 35.1
33.3 32.1
31.0 28.4
24.3 1.75
UU
a a
a b
a ab
a a
Flaky 32.5
32.8 33.1
40.0 33.3
36.5 34.2
33.8 1.27
UU
c abc
c a
bc ab
abc abc
Crumbly 30.3
29.0 30.5
24.8 29.2
24.9 26.5
28.5 1.31
FlaÕour
UUU
def ef
f cde
bcd a
bc ab
Salmon-like flavour 28.7
30.6 31.1
26.6 26.0
18.6 24.4
22.3 1.22
UUU
bcd ab
abc a
ab d
cd d
Earthy flavour 33.1
29.4 32.0
27.5 28.8
38.2 36.7
38.0 1.61
UUU
bc ab
ab a
ab d
cd d
Farmyard flavour 25.7
21.8 23.7
20.0 22.6
32.3 29.9
30.7 1.48
U
bc abc
abc a
bc abc
c ab
Salty flavour 20.5
19.5 19.2
17.6 20.5
20.0 20.7
18.1 0.74
U
b b
b ab
ab ab
a a
Oily flavour 27.9
27.99 27.7
25.8 26.3
24.4 23.4
23.5 1.18
UUU
d cd
cd abc
bcd a
ab a
Fishy flavour 25.0
23.5 24.3
21.4 22.6
18.6 20.2
19.5 0.93
UUU
abc ab
c ab
bc d
a ab
Ž .
Fishy flavour skin 30.9
30.3 33.2
30.1 31.6
25.0 28.8
29.7 0.82
Texture
UUU
a a
a b
b b
b b
Firmness 35.1
30.5 30.7
44.1 44.9
42.7 43.7
43.3 1.80
U
a ab
ab b
ab b
ab ab
Flakiness 29.3
31.1 31.7
33.9 31.6
33.7 31.3
32.0 0.92
UUU
b b
b a
a a
a a
Light texture 44.6
48.9 48.8
39.3 36.6
36.9 36.5
36.1 1.53
UUU
b b
b a
a a
a a
Moist texture 35.4
38.7 37.1
27.9 26.3
25.7 25.0
26.2 1.77
UU
ab a
a abc
c bc
bc bc
Chewiness 28.3
26.0 26.2
29.8 33.9
31.5 31.7
31.7 1.49
Aftertaste
UUU
bc bc
bc d
c a
ab ab
Time when starts 14.4
15.7 15.2
22.6 18.4
10.1 12.6
12.1 1.27
UUU
ab a
a a
ab c
bc c
Overall aftertaste 40.4
34.4 37.0
33.6 40.5
52.8 47.5
50.9 2.39
UUU
bc ab
ab a
ab d
cd d
Earthy aftertaste 32.3
26.2 28.2
25.0 28.5
42.4 37.1
40.8 2.12
UUU
cde ab
abc a
bcd f
ef def
Metallic aftertaste 27.6
22.3 24.5
21.4 26.2
32.5 30.4
29.8 1.24
UUU
b bc
b c
bc a
a a
Chicken-like aftertaste 12.7 14.1
12.5 15.0
13.9 8.9
10.0 10.1
0.61
UUU
bcd d
cd abc
abc a
ab a
Oily aftertaste 27.6
29.0 28.3
24.6 24.5
20.6 23.5
22.8 1.28
†
Significance of results calculated by ANOVA:
‡
Scores without common superscripts within a horizontal line differ significantly according to Duncan’s test; P - 0.05 for salmon-like aroma, P - 0.01 for all other attributes.
U
P - 0.05.
UU
P - 0.01
UUU
P - 0.001.
Ž .
Fig. 1. Principal components PC analysis of the mean profiling scores for salmon harvested during 1993 Ž
. Ž . Ž
. Ž . showing PC1 and PC2 PC3 not illustrated . a Principal component scores for individual salmon scaled . b
Ž Correlations of each attribute with principal component axes
A sappearance, T s texture, Os odour, .
F s flavour, AT saftertaste .
Ž .
and did not account for any more of the variance not shown . Neither were any Ž
significant correlations detected between these attributes and profiling attributes not .
shown .
Ž .
Fig. 2. Principal components PC analysis of the mean profiling scores for salmon harvested during 1994 Ž
. Ž . Ž
. Ž . showing PC1 and PC2 PC3 not illustrated . a Principal component scores for individual salmon scaled . b
Ž Correlations of each attribute with principal component axes
A sappearance, T s texture, Os odour, .
F s flavour, AT saftertaste .
The use of hedonic scaling methods to assess the acceptability of the salmon harvested during 1993 shows that, in general, the farmed salmon were at least as
Table 5 Effect of salmon source on the acceptability of salmon harvested during 1993. Scores are on a scale of 0 to 8
Ž .
1sextremely acceptable, 8sextremely unacceptable
†
Acceptability of Farm 1
Farm 2 Farm 3
Sea 2 River 1
River 2r3 Significance
SEM
‡
Odour 3.25
3.35 3.10
3.13 3.35
3.53 NS
0.201
U
a b
ab ab
ab b
Appearance 2.58
3.38 2.83
3.13 3.05
3.35 0.187
UU
a a
a ab
b ab
Flavour 3.43
3.35 3.13
3.68 4.33
3.70 0.228
Texture 3.33
2.78 2.83
3.33 3.25
3.38 NS
0.204
UUU
a a
a ab
b b
Aftertaste 3.20
3.08 3.10
3.68 4.28
4.18 0.224
UU
ab ab
a abc
c bc
Overall 3.40
3.35 3.08
3.78 4.15
4.05 0.224
†
Significance of results calculated by ANOVA: The non-parametric Kruskal–Wallis rank sum test produced identical levels of significance to the ANOVA with the exception of acceptability of flavour
Ž .
P - 0.05 for Kruskal–Wallis, P - 0.01 for ANOVA .
‡
Scores without common superscripts within a horizontal line differ significantly according to Duncan’s test; P - 0.05 for salmon-like aroma, P - 0.01 for all other attributes.
U
P - 0.05.
UU
P - 0.01.
UUU
P - 0.001.
Ž .
acceptable as the wild fish Table 5 .The salmon from River 1 were significantly less acceptable than those from other sources in terms of flavour and overall acceptability.
Analyses of the correlation of the mean scores for acceptability of aroma, flavour and aftertaste with the appropriate mean profiling scores for each treatment gave the
Table 6 Significant correlations
a
between mean scores for acceptability and profiling attributes, and between overall acceptability and other acceptability scores
b c
a,d e
Attribute 1 Attribute 2
R Significance
U
Acceptability of flavour Salmon-like flavour
y0.898
U
Earthy flavour 0.812
U
Acceptability of texture Moisture
y0.818
U
Lightness y0.884
U
Acceptability of aftertaste Metallic aftertaste
0.833
U
Time when aftertaste starts y0.836
U
Strength of aftertaste 0.906
U
Acceptability of appearance Peach
y0.834
U
Orange y0.839
U
Overall acceptability Acceptability of flavour
0.908
UU
Acceptability of aftertaste 0.969
a
Positive correlation between attributes is represented, in this case, by negative values of R, due to the fact that hedonic scales were used in which low scores were most acceptable.
b
Attribute 1sa hedonic scale for acceptability scored by an untrained panel.
c
Attribute 2 sa profiling attribute, scored by a trained panel, except for the last two correlations, when further acceptability scales are used.
d
Pearson’s product moment coefficient.
e
Significance of R, using two-tailed table of critical values of Pearson’s product moments coefficient Ž
. O’Mahoney 1986 .
U
P - 0.05.
UU
P - 0.01.
Table 7 Mean eating quality attribute scores for fresh and frozen salmon analysed by ANOVA and Fisher’s LSD Test
a
Attribute Treatment
Significance SEM
Frozen salmon Fresh salmon
Appearance
UUU
Juicy appearance 31.5
41.2 1.29
UUU
Separation 28.4
35.4 0.89
UUU
Moist appearance 31.2
41.1 0.84
FlaÕour
UUU
Oily flavour 21.6
26.8 0.87
Texture
UUU
Moisture 33.1
45.9 1.16
UU
Lightness 44.5
50.1 1.09
U
Springiness 24.6
27.8 0.87
UU
Tenderness 49.9
55.9 1.42
a
Attributes for which a significant difference was detected by ANOVA.
U
Significance: P - 0.05.
UU
Significance: P - 0.01.
UUU
Significance: P - 0.001.
correlation coefficients listed in Table 6. No evidence was found for curvilinear relationships between acceptability and profiling scores.
3.2. Effect of freezing on eating quality For 24 of the 32 profiling attributes, there was no significant difference between
frozen and fresh salmon. However, significant dfferences were found for the remaining Ž
. 8 attributes Table 7 . Most of these described the appearance and texture of the salmon.
The only aspect of flavour altered by freezing was ‘oily flavour’, which was found to be reduced in the frozen fish.
Table 8 Mean eating quality attribute scores for salmon frozen for differing periods, analysed by ANOVA and Fisher’s
LSD Test
f
Attribute Treatment
Significance SEM
4–5 weeks 8–9 weeks
15–16 weeks 33–34 weeks
Appearance
U
a a
a,b b
Separation 26.6
27.4 24.8
18.3 2.02
U
a a
a,b b
Moist appearance 30.4
33.0 28.4
21.4 2.25
f
Attributes for which a significant difference was detected by ANOVA.
a,b
Ž Scores without common superscripts within a horizontal line differ significantly P - 0.01; Fisher’s
. LSD test .
U
P - 0.05
3.3. Effect of length of frozen storage on eating quality Thirty of the 32 eating quality attributes showed no significant change over 34 weeks
of frozen storage. The two attributes affected by length of frozen storage refer to the Ž
. appearance of the cooked salmon Table 8 . There were no significant effects of length
of frozen storage on odour, flavour, aftertaste or texture.
4. Discussion