Está en la página 1de 5

EXPANDED STATISTICAL

IN PAIRED-PREFERENCE,

E. B. ROESSLER,

TABLES FOR ESTIMATING SIGNIFICANCE


PAIRED-DIFFERENCE,
DUO-TRIO
AND
TRIANGLE TESTS

R. M. PANGBORN,

ABSTRACT
Two setsof expandedtableshavebeen compiled for usein determining
significance in paireddifference and triangle tests (one-tailed) and in
paired-preferencetests (two-tailed). One set of tables lists the number
of correct responses(or agreeingjudgments) for trials ranging from
7-100, at p < 0.05, 0.04, 0.03, 0.02, 0.01, 0.005 and 0.001. These
tables are convenient for a quick estimate of significanceof laboratory
sensory data as well as consumer responses.The secondset of tables
gives the probabilities of obtaining a given number of correct (or
agreeing)judgments in trials ranging from 5-50. Theseprobability tables provide a more precise estimate of significance,which may be
neededin more critical researchor in making decisionsof considerable
importance. Some examplesare given, with guidelines for the proper
use of these tables and the interpretation of significancebasedupon
them.
TABLES used to determine significance in discrimination and
preference tests usually indicate the number of judgments required at only three levels of significance, i.e., the 5%, 1% and
0.1% levels (Roessler et al., 1948, 1956). Stone and Side1
(1978) have pointed out the inconsistencies in the entries in
subsequent tables of this type (Amerine et al., 1965; Larmond,
1970; Stahl and Einstein, 1973). Since these tables appear to
have been constructed using different criteria, it is recommended that exact probabilities be obtained from tables of the
cumulative binomial probability distribution, or in the event
that such tables are not available, that approximate probabilities be computed using the normal curve. This procedure
leaves no doubt as to the true probability. To be almost significant at a certain probability level is not the same as being
significant at that level. Assurance of significance of the
occurrence of an event at a particular level of cr requires that
the probability of the event occurring is equal to or less than
(Y. We cannot concur with Basker (1976) that his contrived
tables of probability for triangle testing by individual judges
are a satisfactory approximation which can be used instead of
unwieldy tables of exact significance levels.
Since many investigators prefer, and will continue to use,
tables giving the number of judgments required for significance
at various levels, it is desirable to have tables with more than
three probability levels. Otherwise, conclusions may disregard
valuable information. For example, in 46 trials of a triangle
test, 22 correct judgments are required for significance at p <
0.05. However, 22 correct judgments are also significant at p <
0.03, which represents considerably better performance and
will be overlooked with tables giving only the usual three levels
of p < 0.05, 0.01, and 0.001.

Tablesfor correctjudgments
Table 1 gives the numbers of correct judgments required
for significance in the paired-difference and duo-trio tests.

Author
Roessler is with the Dept. of Mathematics,
Univ. of California,
Davis, CA 95616. Author
Pangborn
is with the Dept. of Food Science
& Technology,
Univ. of California,
Davis, CA 95676. Authors
Side1 and
Stone are with Tragon Corp., Palo Alto, CA 94302.

-JOURNAL

OF FOOD SCIENCE-Volume

Table l-Minimum
nificance
at various
trio tests (one-tailed,

numbers
of correct judgments
to establish sigprobability
levels for paired-difference
and duop = %)a
Probability

No. of
trials (n)

0.05

0.04

7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
60
70
80
90
100

7
7
8
9
9
10
10
11
12
12
13
13
14
15
15
16
16
17
18
18
19
19
20
20
21
22
22
23
23
24
24
25
26
26
27
27
28
28
29
30
30
31
31
32
37
43
48
54
59

7
7
8
9
9
10
11
11
12
12
13
14
14
15
15
16
17
17
18
18
19
20
20
21
21
22
23
23
24
24
25
25
26
27
27
28
28
29
29
30
30
31
32
32
38
43
49
54
60

levels

0.03

0.02

0.01

0.005

7
8
8
9
10
10
11
11
12
13
13
14
15
15
16
16
17
18
18
19
19
20
21
21
22
22
23
23
24
25
25
26
26
27
27
28
29
29
30
30
31
31
32
33
38
44
49
55
60

7
8
8
9
10
10
11
12
12
13
14
14
15
16
16
17
17
18
19
19
20
20
21
22
22
23
23
24
25
25
26
26
27
27
28
29
29
30
30
31
31
32
33
33
39
45
50
56
61

7
8
9
10
10
11
12
12
13
14
14
15
15
16
17
17
18
19
19
20
20
21
22
22
23
24
24
25
25
26
26
27
28
28
29
29
30
31
31
32
32
33
34
34
40
46
51
57
63

8
9
10
11
11
12
13
13
14
15
15
16
17
17
18
19
19
20
20
21
22
22
23
24
24
25
25
26
27
27
28
28
29
30
30
31
31
32
33
33
34
34
35
41
47
52
58
64

a Values
(X)
not
appearing
in
X = (2 J;; + n + 1112. See text.

table

may

be

derived

0.001

10
11
12
13
13
14
15
16
16
17
18
18
19
20
20
21
22
22
23
24
24
25
26
26
27
27
28
29
29
30
30
31
32
32
33
34
34
35
36
36
37
43
49
55
61
66
from:

These are one-tailed tests as only one response is correct, and p


= l/z (Amerine et al., 1965). Table 2 gives the same information
for the triangle test which is one-tailed with p = l/3. Table 3 is
for use in paired presentation for preference which is two-

0022-1147/78/0003-0940$02.25/O
0 1978 Institute of Food Technologists
940

J. L. SIDEL and H. STONE

43 (1978)

EXPANDED
Table 2-Minimum
numbers of correct
nificance
at various probability
levels
tailed, p = l/3) a

judgments
to establish sigfor the triangle
test (one-

Probability
No.of
trials (n)

0.05

0.04

4
5
5
6
6
7
7
8
8
9

5
5
6
6
7
7

5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
60
70
80
90
100
aValues

t
10
10
11
11
12
12
12
13
13
14
14
15
15
15
16
16
17
17
17
18
18
19
19
19
20
20
20
21
21
22
22
22
23
23
27
31
35
38
42
(XI

not

X = 0.4714

z fi+

ii
8
9
9
10
10
11
11
11
12
12
13
13
14
14
14
15
15
16
16
16
17
17
18
18
18
19
19
20
20
20
21
21
22
22
22
23
23
24
27
31
35
39
43
appearing

0.03
5
5
6
6
7
7
8
8
9
9
10
10
10
11
11
12
12
13
13
13
14
14
15
15
16
16
16
17
17
18
18
18
19
19
20
20
20
21
21
22
22
22
23
23
24
24
28
32
36
40
43

TABLES FOR ESTIMATING

Table 3-Minimum
numbers of agreeing
tablish significance
at various probability
ence test (two-tailed I p = Xla

levels

Probability

0.02

0.01

5
5
6
6
7
7
8
8
9
9
10
10
11
11
12
12
13
13
13
14
14
15
15
16
16
16
17
17
18
18
19
19
19
20
20
21
21
21
22
22
23
23
23
24
24
25
29
33
36
40
44

5
6
6
7
7
8
8
9
9
10
10
11
11
12
12
13
13
14
14
15
15
15
16
16
17
17
18
18
18
19
19
20
20
21
21
21
22
22
23

in table

STATISTICAL

0.005

23
24
24
24
25
25
26
30
34
38
42
45

rhay

be derived

5
6
7
7
8
8
9
9
10
10
11
11
12
12
13
13
14
14
15
15
16
16
17
17
17
18
18
19
19
20
20
20
21
21
22
22
23
23
24
24
24
25
25
26
26
26
31
35
39
43
47

0.001

7
8
8
9
10
10
11
11
12
12
13
13
14
14
15
15
16
16
17
17
18
18
19
19
20
20
21
21
22
22
22
23
23
24
24
25
25
26
26
27
27
27
28
28
33
37
41
45
49

from:

levels

7
8
8
9
10
10
11
12
12
13
13
14
15
15
16
17
17
18
18
19
20
20
21
21
22
23
23
24
24
25
25
26
27
27
28
28
29
29
30
31
31
32
32
33
39
44
50
55
61

7
8
8
9
10
10
11
12
12
15
14
14
15
16
16
17
17
18
19
19
20
20
21
22
22
23
23
24
25
25
26
26
27
27
28
29
29
30
30
31
31
32
33
33
39
45
50
56
61

7
8
9
9
10
11
11
12
13
13
14
15
15
16
16
17
18
18
19
19
20
21
21
22
22
23
24
24
25
25
26
27
27
28
28
29
30
30
31
31
32
32
33
34
39
45
51
56
62

7
8
9
10
10
11
12
12
13
14
14
15
15
16
17
17
18
19
19
20
20
21
22
22
23
23
24
25
25
26
26
27
28
28
29
29
30
30
31
32
32
33
34
34
40
46
51
57
63

aValues
(XI
not appearing
in
X = (2 Ji+
n + 1112.See text.

table

judgments
necessary t0 eslevels for the paired-prefer-

No.of
trials in)
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
60
70
80
90
100

TESTS..

8
9
10
11
11
12
13
13
14
15
15
16
17
17
18
19
19
20
20
21
22
22
23
24
24
25
25
26
27
27
28
z;
30
30
31
31
32
33
33
34
34
35
41
47
52
58
64
may

be

9
10
11
12
12
13
14
14
15
16
16
17
18
18
19
20
20
21
22
22
23
24
24
25
25
26
27
27
28
29
29
30
30
31
32
32
33
33
34
35
35
36
42
48
53
59
65

21
21
22
23
23
24
25
25
26
27
27
28
29
29
30
31
31
32
32
33
34
34
35
36
36
37
37
44
50
56
61
67

derived

from:

11
12
13
14
14
15
16
17
17
18
19
19

20

[(2n +3)/6].Seetext.

tailed (as either sample may be preferred) and p = % (Amerine


et al., 1965). For numbers of trials not shown in the tables,
numbers of required judgments may be obtained from tables
of the cumulative binomial probabilities or, excellent approximations may be computed from the following formulas:
Tables 1 and 3: X = (2 fi+

Table 2: X= 0.4717 z fi+

where n is the number of trials, and the minimum number of


correct (agreeing) judgments is X, if X is a whole number, or
the next higher integer if X is not a whole number, and where
z is taken from the following table.

n + 1112

[(2n + 3)/6]

Values of z
Probability ((u)
Tables 1 and 2
Table 3

0.04 0.03 0.02 0.01 0.005 0.001


1.64 1.75 1.88 2.05 2.33 2.58
3.10
1.96 2.05 2.17 2.33 2.58 2.81
3.30

0.05

Volume 43 (1978kJOURNAL

OF FOOD SCIENCE-

941

Table I-Probability

of X or more correct

judgments

in n trials

(one-tailed,

p = XJa

As an example of the use of the formulas for values exceeding tabular entries, consider the case of preferences between two canned meats by 150 consumers. Since Table 3
extends only to n = 100, one would apply the formula for the
paired-preference test (two-tailed, p = 1/):
x=zJ150+150+1
?L
Inserting the z values corresponding to probabilities of 0.05,
0.04, 0.03, 0.02, 0.01, 0.005 and 0.001, the minimum numbers of agreeing judgments (X) are 88, 88, 89, 90, 92, 93 and
96, respectively.

Tablesof probabilities

OF FOOD SCIENCE-Volume

40-

$30-

Tables l-3 are convenient to consult for quick estimates of


significance. For much greater accuracy, exact probabilities
should be calculated as recommended by Stone and Side1
(1978). Tables 4, 5 and 6 list the probabilities of obtaining X
or more correct responses (one-tailed) or agreeing judgments
(two-tailed) in n trials. Note that the initial decimal point has
been omitted in the body of the tables to save space, i.e., the
value of 969 should be 0.969.
If an experimenter collected 50 paired comparisons to
determine if an expert panel could detect the added vanilla in
a pudding mix, and obtained 31 correct responses, consultation of Table 1 would lead him to conclude there was no
significant difference, i.e., not significant at p < 0.05. However, consultation of Table 4 shows that the probability of
obtaining 3 1 out of 50 judgments in a paired-difference test by
chance alone is 0.059. The experimenter would not only be
more accurate in reporting that there were significant differences between the pudding formulas at p = 0.059, but
might change his mind regarding a go, no-go decision on the
product. If he is willing to take a chance at the 5% level, i.e.,
risk being wrong five times out of a hundred trials, might he
not also be willing to take a chance at the equally-arbitrary
cut-off of 5.9%?
An understanding of probability is essential for the correct
use of these tables in order to avoid misinterpretation of the
significance of results from sensory testing. Furthermore, discretion is needed in the interpretation of n in these tables. For
942 -JOURNAL

43 (1978)

20 -

Chance probability = +

------

Chance probability = +

IO0
0

I
I
IO 20

I
I
30 40

I
I
50 60

t
70

I
80

I
I
90 100

Number of observations
Fig. l-Plot
of the percent correct responses necessary for significance at p < 0.01 for one-tailed
tests when chance probabilities
are
l/3 (triangle tests) or l/Z (paired-difference
tests). Reprinted
from
Amerine
et al. (1965) with permission
from Academic
Press, New
York.

example, for n = 20, the same values apply for 1 judge X


20 sessionsas for 20 judges X 1 session,but the interpretation
of the result, obviously, is different.
Tables l-3 list the number of correct decisions for a maximum n of 100. However, these values should not be misconstrued as recommendations for the maximum or minimum
number of trials. In most situations, in fact, discrimination
testing requires less than 50 trials in order to reach a reliable
and valid decision. It is important to recognize that as n increases,the percent correct observations needed for statistical
significance decreases,but in a nonlinear fashion. Figure 1 is a
graphic representation of correct responsesneeded for significance at p < 0.01 for chance probabilities of % (paired tests)

EXPANDED
Table 6-probability
n x o
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50

868
912
941
961
974
983
988
992
995
997
998
998
999
999

539
649
737
805
857
896
925
946
961
973
981
986
990
993
995
997
998
998
999
999
999

210
320
429
532
623
701
766
819
861
895
921
941
956
967
976
982
987
991
993
995
996
997
998
999
999
999

045
100
173
259
350
441
527
607
678
739
791
834
870
898
921
940
954
965
974
980
985
989
992
994
996
997
998
998
9.99
999
999

004
018
045
088
145
213
289
368
448
524
596
661
719
769
812
848
879
904
924
941
954
964
972
979
984
988
991
993
995
996
997
998
998
999
999
999

a NOTE: initial decimal punt

6
001
007
020
042
077
122
178
241
310
382
453
522
588
648
703
751
794
831
862
888
910
928
943
955
965
972
978
983
987
990
992
994
996
997
997
998
999
999
999
999

003
008
020
039
066
104
149
203
263
326
391
457
521
581
638
690
737
778
815
847
874
897
916
932
946
957
965
973
978
963
987
990
992
994
995
996
997
998
998
999
999
999
999

of X or more correct
8

001
003
009
019
035
058
088
126
172
223
279
339
399
460
519
576
630
679
725
765
801
833
861
885
905
922
937
959
967
973
979
983
987
990
992
994
995
996
997
996
998

judgments

STATISTICAL
in n trials

10

11

12

13

14

15

001
004
009
017
031
050
075
108
146
191
240
293
349
406
462
518
572
623
670
714
754
789
821
849
873
895
913
928
941
952
961
968
974
980
984
987
990
992
994
995

001
002
004
008
016
027
043
065
092
125
163
206
264
304
357
411
464
517
56a
617
662
705
744
779
810
838
863
885
903
920
933
945
955
963
970
976
980
984
987

001
002
004
008
014
024
038
056
079
107
140
178
220
266
314
364
415
466
516
565
612
656
697
735
769
800
829
854
876
895
912
926
938
949
958
965
972

001
002
004
007
013
021
033
048
068
092
121
154
191
232
276
322
370
419
468
516
562
607
650
689
726
761
791
820
845
867
687
904
919
932
943

WI
002
004
007
012
019
028
042
058
079
104
133
166
203
243
285
330
376
422
469
515
560
603
644
683
719
753
783
811
836
859
879
896

001
002
003
CO6
010
016
025
036
050
068
WC
115
144
177
213
252
293
336
381
425
470
515
558
a00
639
677
713
745
776
803
829

001
GO2
w3
006
009
014
022
031
043
059
078
100
126
155
187
223
261
301
342
385
428
471
514
556
596
635
672
706
739

16

001
002
003
005

008
013
019
027
038
051
067
087
109
135
164
196
231
268
307
347
389
430
472
514
554
593
631

TABLES FOR ESTlMATlNG

(one-tails

(one-tailed,

TESTS..

p = 1/3Ja

17

18

19

20

21

22

23

24

25

26

27

28

001
002
003
005
007
011
016
023
033
044
058
075
095
118
144
173
205
239
275
313
352
392
433
473
513

001
001
002
004
006
010
014
020
028
038
051
066
083
104
127
153
182
213
246
282
318
356
395

WI
001
w2
004
006
009
012
018
025
033
044
057
073
091
111
135
161
189
220
253
287

001
001
w2
003
005
007
011
016
021
029
038
050
063
079
098
119
142
168
196

001
001
002
003
w4
007
010
014
019
025
033
043
055
070
086
105
126

001
001
002
003
004
006
008
012
016
022
029
038
048
061
076

001
001
001
002
003
cm5
007
010
014
019
025
033
042

Wl
001
002
003
004
006
009
012
017
022

001
001
002
003
004
006
008
011

001
CC1
002
002
003
005

wl
001
001
002

001

harbeenomitted.

and l/3 (triangle tests), showing that beyond an n of 30, the


percent correct responseschangesvery slightly.
In Tables 1 and 2, for n = 100, note that only 59% correct
responses are required in a paired-difference test, while only
42% correct responses are required in a triangle test to attain

significance at p < 0.05. Statistical significance is achieved, but


from a practical point of view, this may not be an important
difference, as in both cases, only nine correct decisions above
chance are necessary to be significant at p < 0.05. The behavioral implications and interpretation of statistical tests
-Continued

Table 6-Probabilitv

of X or more aoreeina

iudaments

in n trials

(two-tailed.

Volume 43 (1978kJOURNAL

on page

947

?4Ja

OF FOOD SCIENCE-

943

QUALITY

(3)

(aslaT), = a2 + a3t + 2a, tT

(4)

At 38C, the rate is equal to 0.548 units/Co and it is reduced


to 0.374 units/C at 1OC. The rate of increase in the score
of the same product with respect to increases in the storage
time becomes:
(as/at), = -0.132354 - 0.002781T+ 0.0002587T
We may easily calculate that the rates at 4.5, 21 and 38C are
-0.140 (units/ma), -0.076, and 0.131 respectively.

Presented
at the 37th Annual
Meeting
of the Institute
of Food
Technologists,
Philadelphia.
PA, June 5-8.1978.
A paper of the Journal Series, New Jersey Agricultural
Experiment
Station. Rutgers. Contribution
No. 702 from the Engineerina
Research
Service; Canada Agriculture,
Ottawa, KlAOC6,
Canada.
This investigation
was supported
in part by the U.S. Public Health
Service Grant No. FD-00119
from the Food & Drug Administration.

CONCLUSION
THE CANNED PRODUCTS stored at 38OCdeteriorated at considerably greater rates as compared with those stored at the
lower temperatures. The storage at 3, 4.4 or 21C did not re-

TABLES

FOR

ESTIMATING

TESTS . . . From

extend well beyond the relatively straightforward computation


for estimation of statistical significance. Texts by Edwards
(1965), McCall (1970), Huff (1974) and Reichmann (1961),
among others, provide extensive statistical information and
recommendations for applications to the behavioral sciences.

REFERENCES
Amerine,
M.A., Pangborn. R.M. and Roessler. E.B. 1965. Principles
of
Sensory Evaluation
of Foods, Academic Press, New York.
Basker. D. 1976. Comparison
of discrimination
ability
between
taste
panel assessors. Chemical Senses & Flavor 2: 207.
Edwards, A.L. 1965. Statistical
Methods for the Behavioral
Sciences.
Rinehart,
New York
Huff, D. 1974. How
to Lie with Statistics,
Penguin Books. Philadelphia.
Larmond,
E. 1970. Methods
for Sensory Evaluation
of Food, Publ.
#1284, Canada Dept. of Agriculture,
Ottawa.

FOODS..

Ball, C.O., Joffe, F.M., Stier, E.F. and Hayakawa,


K. 1963. The role of
temperature
in retaining
quality in canned foods. ASHRAE
Journal
5(6): 93.
Barr. A., Goodnight,
J.H., Sall, J.P.. Helwig, J.T. 1976. A Users Guide
to SAS 76. SAS Institute,
Inc., P.O. Box 10522, Raleigh, NC.
Brady,
A.L. and Bedrosian.
K. 1961. Effect of room temperature
vs.
refrigerated
storage on aualitv of canned fruit and vegetable Drodu&s-Food
Tech&l.
15: 367. Cecil, SF. and Woodroof,
J.C. 1963. The stability
of canned foods in
long term storaae. Food Technol. 17: 131.
D&al, K.B. 1964.-Thermal
degradation
of pigments
and relating biochemical
changes in canned apricots
and cherries. Food Technol.
18: 1198.
Daoud.
H.N. and Luh, B.S. 1971. Effect
of partial
replacement
of
sucrose by corn syrup on quality
and stability
of canned apple
sauce. J. Food Sci. 36: 419.
Dryden.
B.C. and Hills, C.H. 1957. Consumer
preference
studies on
apple sauce: Sugar-acid relations. Food Technol. 11: 589.
Joffee. F.M.. Stier. E.F.. Bongolan.
D.C.. EDstein. A.I. and Ball. C.O.
1961. Low-temperature
handling
of ster%sed foods. 4. Color and
flavor of canned vegetables. Food Technol. 15: 507.
Luh. B.S. and Sioud. F.B. 1966. AseDtic carmine. of foods. 4. Stabilitv
of pear puree with essence recover;.
Food Teihnol.
20: 1590.
Steele, R.G.D.
and Torrie, J.H. 1960. Principles
and Procedures
of
Statistics,
p. 105. McGraw-Hill
Book Co., New York, NY.
MS received 6116177; revised 10129177: accepted 1115177.

(as/an,=, 1 = 0.311998+ 0.006208x T

STATISTICAL

CANNED

REFERENCES

For example, the rate of change in the flavor scores of the peas
with respect to increase in the storage temperature at 1 yr of
storage may be estimated by Eq (4):

EXPANDED

STORED

sult in severe loss of quality during the entire test periods.


Changes in the flavor difference scores of the stored products
were represented with polynominal functions of the storage
times and temperatures.

for the present investigation, this estimated value is not reliable.


The rates of changes in the flavor scores with changes in
storage time or temperature may be easily estimated from
respective partial derivatives of Eq (1):
(as/at), = a, + a,T + a,T2

OF SHORT-TERM

page 943

McCall, R.B. 1970. Fundamental


Statistics for Psychology,
Harcourt,
Brace. and World, New York.
Reichmann.
W.J. 1961. Use and Abuse of Statistics.
Methuen & Co..
Ltd., London.
Roessler.
E.B., Warren, J. and Guymon.
J.F. 1948. Significance
in
triangular
taste tests. Food Res. 13: 503.
Roessler, E.B., Baker, G.A. and Amerine, M.A. 1953. Corrected normal
and chi-square
approximations
to the binomial
distribution
in
organoleptic
tests. Food Res. 18: 625.
Roessler, E.B.. Baker, G.A. and Amerine,
M.A. 1956. One-tailed
and
two-tailed
test in organoleptic
comparisons.
Food Res. 21: 117.
Stahl. W.H. and Einstein.
M.A. 1973. Sensorv testing methods.
In
Encyclopedia
of Industrial
Chemical Analysis,
17: 608.
Stone. H. and Sidel. J.L. 1978. Computing
exact probabilities
in sensory discrimination
tests. J. Food Sci. 43:
MS received 7125177; revised 1114177; accepted 11/10/77.
Values not appearing in the table may be obtained from Tables of
the Cumulative
Binomial
Probabilitv
Distribution.
Annals
of the
Computation
Laboratory
of Harvard
University,
Harvard
University
Press, Cambridge.
Mass., 35:1955.

Volume 43 (1978kJOURNAL

OF FOOD SCIENCE-

947

También podría gustarte