Difference test sensitivity: Comparison of three versions of the
duo–trio method requiring different memory
schemes and taste sequences
Hye-Seong Lee, Kwang-Ok Kim
*
Department of Food Science and Technology, Ewha Womans University, Seoul 120-750, South Korea
Received 30 January 2007; received in revised form 21 May 2007; accepted 9 July 2007
Available online 6 August 2007
Abstract
The duo–trio test is a simple non-directional difference test method suitable for discriminating subtle differences between confusable
stimuli. It is suitable for consumers as well as trained panelists. In the present study, the influence of memory and tasting sequence in the
duo–trio was investigated, using various low concentrations of NaCl in a roving discrimination design. Three versions of duo–trio pro-
tocols were considered. The first protocol was the traditional duo–trio with the reference tasted first (DTF). The second protocol was the
duo–trio with the reference tasted in the middle, between the two test samples (DTM). The third protocol was the duo–trio with the
reference tasted twice, first and last as a reminder (DTFR). In some conditions, the higher concentration of NaCl solution was the ref-
erence (W-odd); in others, the lower concentration was the reference stimulus (S-odd). Calculated d
0
values indicated a superior perfor-
mance for the DTFR over the DTF and DTM. In the DTF, S-odd yielded a lower d
0
than W-odd, probably due to NaCl taste
adaptation. In the DTFR, the effects of adaptation were counterbalanced for all the stimuli including the reference, resulting in no sig-
nificant sequence effect. In the DTM, the sequence of the first two stimuli in the test had a significant effect on the performance, probably
due to the decision strategy used for the protocol. Manipulation of the reference in the duo–trio test introduces issues such as attention as
well as memory and adaptation.
Ó 2007 Elsevier Ltd. All rights reserved.
Keywords: Difference tests; d
0
; duo–trio; Memory; Sequence effects; NaCl taste adaptation; Cognitive decision strategy
1. Introduction
The duo–trio test is a commonly used three stimulus dif-
ference test in which the attribute that differs is not specified
(
ASTM, 1968; Lawless & Heymann, 1996; Meilgaard,
Civille, & Carr, 1991; Peryam, 1958
). The duo–trio is partic-
ularly useful for many important academic and business
objectives such as quality control, ingredient or process
changes, and storage or packaging changes which often
involve subtle multidimensional differences between stimuli.
The same–different test and the triangle test share the same
characteristic of not having to describe what constitutes the
difference. However, they differ in that same–different test
involves two stimuli while the triangle involves three.
Considering difference tests, three main issues can be
identified; cognitive strategies, the sequence of tasting, and
memory effects. Regarding cognitive strategies, a Thursto-
nian context can be invoked. It is generally assumed that
the duo–trio test as well as the triangle test uses what has
been called the ‘comparison of distances’ cognitive strategy
(
Ennis, 1993; Lee & O’Mahony, 2004; O’Mahony, Mas-
). It is necessary to know the cognitive
strategy for the computation of d
0
values. Because of this,
research into cognitive strategies is important.
The second topic concerns the sequence of tasting. This
has been described by the Sequential Sensitivity Analysis
0950-3293/$ - see front matter
Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.foodqual.2007.07.004
*
Corresponding author. Tel./fax: +82 0 2 3277 3095.
E-mail address:
(K.-O. Kim).
www.elsevier.com/locate/foodqual
Food Quality and Preference 19 (2008) 97–102
(SSA) model (
Lee & O’Mahony, 2007a; O’Mahony &
) as well as conditional stimulus model (
). The SSA model predicts for a distilled
water/low concentration NaCl system that a duo–trio with
a weaker (water) reference (S-odd) would give superior per-
formance to one with a stronger (NaCl) reference (W-odd)
(
).
The third topic concerns memory. The memory problem
in difference tests has been discussed in reviews of Thursto-
nian modeling (
Lee & O’Mahony, 2004; O’Mahony &
). A consideration of memory or forgetting
effects would predict that a difference test with two stimuli
would give superior performance to one using three stimuli.
This can be argued from two points of view: memory decay
and interference. For the former, it can be argued that in a
difference test with two stimuli, the final stimulus is com-
pared to the decaying memory trace of the adjacent first
tasted stimulus. In a test with three stimuli, the final stim-
ulus is compared with the decaying memory trace of the
adjacent second tasted stimulus and the even more decayed
memory trace of the first tasted stimulus. Performance in
this triadic test would be expected to be inferior.
A second argument regarding inferior performance for
the triadic test would be that while tasting the final third
stimulus, the memory traces of the initial two stimuli would
be distorted because of their mutual interference.
working with triangle
tests, showed that this interference effect was more impor-
tant than mere stimulus decay.
Comparisons have been made of the performance on
three stimulus and two stimulus tests.
O’Mahony (1999) and Rousseau and O’Mahony (1997)
reported that the three stimulus 3-AFC yielded lower d
0
values than the two stimulus 2-AFC. Although this could
be due to memory effects, the comparison is complicated
because
2-AFC
procedure
has
favorable
sequences.
Although SSA predicts superior performance of the trian-
gle test over the same–different test, based on sequence
effects, it has been found that triangle test resulted in equal
or lower d
0
values than same–different test, suggesting the
importance of memory effects (
). Yet,
memory effects would not have appeared to dominate
sequence effects when the duo–trio and same–different test
had comparable d
0
values (
). Again, this comparison is complicated by
tau-variance with the same–different test.
Interestingly, despite comparable expected memory and
sequence effects, the duo–trio was shown to elicit higher d
0
values than the triangle test (
). This could be chance results or it
could suggest some further factors.
The duo–trio with the reference presented first, was orig-
inally developed by the Joseph E. Seagram Quality Labora-
tory in 1941 (
). Compared with the
triangle test, the duo–trio could be viewed as a simpler test
because more information is provided: namely, the refer-
ence stimulus will not be the ‘odd’ one.
introduced a modification of the
duo–trio test. For this protocol, the reference stimulus was
tasted in the middle of the triad. They named the test the
DTM (Duo–Trio Middle). They compared the perfor-
mance of the DTM with the traditional duo–trio test using
an orange flavored beverage with different concentrations
of sugar and found that the DTM elicited significantly
higher d
0
values. They hypothesized that the memory load
for the DTM was reduced because the two test stimuli were
evaluated immediately adjacent to the reference.
Considering the traditional duo–trio, should distortion
of the memory for the first (reference) stimulus be a prob-
lem for the assessment of the second test stimulus (third
tasted stimulus), the duo–trio could be modified by repeat-
ing the reference stimulus at the end of the test as a remin-
der. As with the DTM, each test stimulus would be tasted
adjacent to the same reference. Such a test will be called the
DTFR (Duo–Trio with reference given First and last as a
Reminder). The tasting sequences for the DTM and DTFR
are different. The question becomes whether the perfor-
mance elicited by the DTM and DTFR are also different.
Investigating this question was the goal of the present
experiment.
2. Materials and methods
2.1. Judges
Six judges (age range 20–25 y), female students at Ewha
Womans University, Seoul, South Korea, participated in
the experiment. The judges were requested not to eat or
drink (except water) for at least half an hour before the
tests. All were naı¨ve to the specific aim of the study,
although some of them had participated in psychophysical
experiments before.
2.2. Stimuli
The stimuli consisted of three confusable pairs of differ-
ent low concentration NaCl solutions: 7 and 10 mM, 10 and
13 mM, and 12 and 15 mM. The stimulus pair to be discrim-
inated was not fixed. It was varied randomly over the three
pairs, in order to have judges focus only on the differences
between the stimuli in the pair and to minimize any learning
effects. For convenience, the stimulus with lower concentra-
tion in each stimulus pair will be denoted as ‘W’ (as the
weaker stimulus) and the stimulus with higher concentra-
tion will be denoted as ‘S’ (stronger stimulus).
The NaCl solutions were prepared by dissolving reagent
grade NaCl (Ducksan Pure Chemical Co., Ltd., Ansan,
Gyeonggido) in deionized water. The NaCl solutions were
dispensed in 10 ml aliquots using Labmax Bottle-Top
Dispensers (Witeg Scientific, Berlin, Germany), and served
98
H.-S. Lee, K.-O. Kim / Food Quality and Preference 19 (2008) 97–102
in 2-oz semitransparent plastic cups (Jesam Co., Sungnam,
Gyeonggido). All stimuli were presented at constant room
temperature at 20 ± 2
°C, on plastic trays, with two duo–
trio tests per tray.
2.3. Procedure
Each judge performed three versions of the duo–trio
test. For the first protocol, the traditional duo–trio, the ref-
erence stimulus was tasted first. On tasting the two test
stimuli judges were required to report which one was the
same as the reference. For the purposes of this article, this
protocol will now be denoted as the ‘DTF’ (Duo–Trio
First). For the second protocol, the ‘DTM’ procedure
was used (
), while for the third proto-
col, the ‘DTFR’ procedure was used.
For all three protocols, for half of the tests, the higher
concentration NaCl solution was used as the reference
stimulus, while for the other half, the lower concentration
was used. All three pairs of stimuli were used for each pro-
tocol. Thus within a protocol the stimulus pairs were con-
stantly changing. This is called a roving design (
). For each pair of stimuli and each pro-
tocol there were four possible sequences of tasting. This
gives a total of 36 possible sequences of tasting (3 stimuli
pairs
3 protocols 4 sequences of tasting). Each judge
performed all 36 sequences once in an experimental session.
The tests were grouped according to the protocol (12 DTFs
followed by 12 DTMs followed by 12 DTFRs). The six
possible orders of protocols were counterbalanced for each
judge. Accordingly each judge performed six experimental
sessions and the order of presentation of the protocols was
counterbalanced over judges. Five minute breaks were
taken between protocols within a session to minimize the
fatigue effects. The experimental sessions were separated
by at least 1 day and no more than 3 days. This gave a
grand total of 216 tests per judge.
One deionized water rinse was taken before each test.
Before tasting the first stimulus of the duo–trio test, a pri-
mer stimulus was taken in order to make the NaCl adapt-
ing conditions for the first stimulus similar to the second
and the third stimulus in the test and prevent the distorting
effect of the rinsing water on the taste of the first stimulus.
The primer presented was always same as the weaker stim-
ulus (W) in the pair of stimuli. Judges were told that this
primer was not part of the test and thus, its sensory char-
acteristics should be ignored.
Judges were required to sip the whole 10 ml of each
stimulus at once and expectorate it. No retasting was
allowed in order to examine memory and sequence effects.
Judges responded verbally, and no feedback was given so
as to minimize learning and strategy change. The experi-
ment was conducted in a one to one interview style. Proto-
col testing lengths lasted 5–11 min, 4–13 min, and 5–15 min
for DTF, DTM, and DTFR, respectively. Total session
lengths from starting testing to brief interview at the end,
lasted 30–50 min.
2.4. Statistical analysis
Values of d
0
were computed for each sequence of the
three versions of the duo–trio test from the pooled data
over the six judges, using IFPrograms software, assuming
the comparison of distances cognitive strategy (Institute
for Perception, Richmond, VA). These can also be com-
puted from the proportion of correct responses using tables
(
). Significance of the differences between d
0
val-
ues was tested using chi-square tests (d
0
-test, IFPrograms,
Institute for Perception, Richmond, VA).
3. Results
The d
0
values obtained from the three versions of duo–
trio test are given in
. The DTFR resulted the high-
est d
0
value (2.28) and then the DTF resulted the second
highest d
0
value (1.95) and the DTM (1.72) resulted the
lowest d
0
value (1.72). The difference in d
0
values between
the DTFR and the DTM was significant (p = 0.01), indi-
cating a superior performance of DTFR over DTM.
Values of d
0
were also calculated for each sequence for
each protocol (
). For the DTF and DTFR, the d
0
values for four different sequences were not significantly
different from each other. Yet, for the DTM, there were
differences in d
0
among the sequences. The test sequences
having the first stimulus different from the reference –
hWSSi and hSWWi, resulted in a higher d
0
than the
sequences having the first stimulus the same as the refer-
ence –
hWWSi and hSSWi.
shows the d
0
values
obtained from the results combined according to whether
the first two stimuli were the ‘same’ or ‘different’. It indi-
cates a significant difference only between the two different
sequences for the DTM (p < 0.001).
Sequence effects were also examined in terms of whether
the reference stimulus was the weaker stimulus (W) or
stronger stimulus (S) (
). When the duo–trio test is
considered as a triad, the duo–trio test with a weaker refer-
ence becomes a triad with ‘one stronger, two weaker stim-
uli’: S-odd. The duo–trio with a stronger reference becomes
a triad with ‘one weaker and two stronger stimuli’: W-odd.
For the DTM and the DTFR, there was no significant dif-
ference between the tests having different reference stimuli.
Table 1
Number of correct responses and d
0
values obtained for the three duo–trio
protocols
Number of correct responses (out of
432)
d
0
Variance of
d
0
DTF
319
1.95
cd
0.017
DTM
303
1.72
c
0.017
DTFR
341
2.28
d
0.019
a
d
0
values with different superscripts are significantly different
(p = 0.01).
b
DTF = duo–trio with reference tasted first, DTM = duo–trio with
reference tasted in the middle, DTFR = duo–trio with reference tasted
first as well as at the end as a reminder.
H.-S. Lee, K.-O. Kim / Food Quality and Preference 19 (2008) 97–102
99
The DTF tests with a stronger reference (W-odd) had sig-
nificantly higher d
0
values than the DTF tests with a weaker
reference (S-odd).
4. Discussion
The secondary goal of the present study was to general-
ize to different stimuli the superior performance of DTM
over the traditional duo–trio test (DTF), because the
DTM had both test stimuli adjacent to the reference. An
alternative way of achieving this adjacency was to use a
DTFR design and the primary goal of this study was to
determine its efficacy. In the assessment of these tests, a
roving design was used in order to minimize stimulus learn-
ing from repeated testing, which could have led to cognitive
strategy change. From the results, however, the DTM did
not elicit superior performance to the DTF. The DTFR
elicited superior performance (not always significant) to
the DTF and DTM, despite a possible hypothesis that hav-
ing four stimuli might have increased ‘fatigue’ and memory
load. The repeated presentation of reference appeared to
have no ill effects.
found that the DTM yielded sig-
nificantly higher d
0
values than the DTF for discriminating
between two orange flavored beverages with different sugar
concentrations. In the present study, the DTF elicited
higher d
0
values albeit non-significantly. Examining the
individual sequences (
) only the DTM exhibited
variation in sensitivity. The sequences where the first two
stimuli were the same elicited much lower d
0
values and
so reduced the overall sensitivity. This effect could be
explained by a hypothesized response bias whereby there
was a tendency to report confusable stimuli as ‘different’.
This would reduce performance for the two sequences
where the initial two stimuli were the same. This was sup-
ported by subjective reports that when performing DTM
judges tended to concentrate on determining whether the
first two stimuli were the same or different while the third
was merely tasted as a check.
Comparing these data with the results of
, it would have been convenient had they not used
these ‘bad’ sequences. However, these were the only
sequences that they used, so bad sequences do not explain
the differences between the present study and that of Rous-
seau et al. The two studies used different stimuli. It may be
hypothesized that judges would be more familiar with
changes due to sugar concentration in orange flavored
drinks than with the various taste changes due to NaCl
concentration in saline solutions (
). Further it may be hypothesized that the above
response bias would be less effective with more familiar
stimuli. This might furnish explanation for the differences
between the two studies.
Having discussed the inferior performance of the DTM
compared with the DTF, it remains to discuss why the
DTFR elicited better performance than the DTF. It can
be hypothesized that should there be doubt regarding
Table 2
d
0
values for each sequence for the three duo–trio protocols
Reference
stimulus
Sequence
including
reference
No. of correct
responses
(out of 108)
d
0
Variance
of d
0
DTF
W
WWS
74
1.63
d
0.069
WSW
76
1.74
d
0.068
S
SWS
84
2.20
d
0.072
SSW
85
2.26
d
0.074
DTM
W
WWS
71
1.46
e
0.073
WSS
89
2.53
f
0.082
S
SWW
78
1.85
ef
0.068
SSW
65
1.13
e
0.090
DTFR
W
WWSW
87
2.40
g
0.077
WSWW
91
2.69
g
0.088
S
SWSS
80
1.96
g
0.069
SSWS
83
2.14
g
0.071
a
d
0
values obtained from different sequences are not significantly dif-
ferent for DTF (p = 0.22) and for DTFR (p = 0.28). For DTM, d
0
values
not sharing the same superscript are significantly different (p = 0.001).
b
DTF = duo–trio with reference tasted first, DTM = duo–trio with
reference tasted in the middle, DTFR = duo–trio with reference tasted
first as well as at the end as a reminder.
c
‘W’ refers to the lower concentration of NaCl solution and ‘S’ refers to
the higher concentration of NaCl solution in a pair of stimuli.
Table 3
d
0
values (and variance of d
0
values) for tests having each type of first two
stimuli for the three duo–trio protocols
Duo-trio having the first
two stimuli, ‘same’:
hWWi
or
hSSi
Duo-trio having the first
two stimuli, ‘different’:
hWSi
or
hSWi
DTM
1.30
d
(0.039)
2.17
e
(0.036)
DTF
1.93
c
(0.034)
1.96
c
(0.034)
DTFR
2.26
f
(0.037)
2.30
f
(0.037)
a
d
0
values obtained from tests with each type of first two stimuli (same
pair vs. different pair) are significantly different for DTM (p < 0.001) and
are not significantly different for DTF (p = 0.91) and DTFR (p = 0.88) for
comparisons within protocols. For comparisons between protocols, d
0
values of the protocols sharing the vertical line are not significantly dif-
ferent (p = 0.05).
b
DTF = duo–trio with reference tasted first, DTM = duo–trio with
reference tasted in the middle, DTFR = duo–trio with reference tasted
first as well as at the end as a reminder.
Table 4
d
0
values (and variance of d
0
values) for tests using each type of references
for the three duo–trio protocols
Duo-trio with the weaker
(W) reference: S-odd
Duo-trio with the stronger (S)
reference: W-odd
DTM
1.66
e
(0.034)
1.79
e
(0.034)
DTF
1.68
c
(0.034)
2.23
d
(0.036)
DTFR
2.53
f
(0.041)
2.05
f
(0.035)
a
For comparisons within protocols, d
0
values obtained from tests with
each type of reference are significantly different for DTF (p = 0.04) and are
not significantly different for DTM (p = 0.62) and DTFR (p = 0.08). For
comparisons between protocols, d
0
values of the protocols sharing the
vertical line are not significantly different (p = 0.01).
b
DTF = duo–trio with reference tasted first, DTM = duo–trio with
reference tasted in the middle, DTFR = duo–trio with reference tasted
first as well as at the end as a reminder.
100
H.-S. Lee, K.-O. Kim / Food Quality and Preference 19 (2008) 97–102
which test stimulus is the same as the reference in the DTF,
the introduction of the final reminder stimulus would allow
a second chance for the judgment to be made.
It is interesting that this effect was only apparent when
the weaker stimulus was the reference. From
, it
can be seen that for the DTF, the d
0
was smaller when
the reference was the weaker stimulus. Also, the reverse
was the case for the DTFR. Therefore, the difference
between the DTFR and DTF was enhanced when the ref-
erence was the weaker stimulus. However, these effects
require explanation.
For the DTF, performance was inferior when the refer-
ence was the weaker stimulus. Consider the following
hypotheses for the four sequences of tasting. For the
hW
R
WS
i sequence, adaptation to the reference (W
R
) could
cause the first test stimulus (W) to appear much weaker
while the second test stimulus (S) weakened by adaptation
might become more similar to ‘W
R
’ than ‘W’. For the
hW
R
SW
i sequence, adaptation to the reference (W
R
) could
cause the first test stimulus (S) to appear weaker. Greater
adaptation effects due to the first two stimuli would cause
the third stimulus (W) to appear even weaker and thus less
similar to ‘W
R
’.
For the
hS
R
WS
i sequence, adaptation to the reference
(S
R
) could cause the first test stimulus (W) to appear much
weaker. Adaptation to ‘W’ would be likely to be less strong
and affect the third stimulus (S) to a lesser extent. There-
fore, there could be a tendency to judge ‘S
R
’ and ‘S’ as
more similar. For the
hS
R
SW
i sequence, adaptation to
the reference (S
R
) could cause the first test stimulus (S)
to appear weaker. Greater adaptation effects due to the first
two stimuli would cause the third stimulus (W) to appear
even weaker and thus less similar to ‘S
R
’.
Considering these four hypotheses, contradictory to the
prediction of Sequential Sensitivity Analysis (SSA), there
would be greater tendency for errors and lower d
0
values
when the reference was the weaker stimulus for the DTF
protocol. Now consider hypotheses for the DTFR
protocol.
The task of the judge was to compare the two test stim-
uli (in the middle of the sequence) to the two reference
stimuli at the beginning and end of the sequence. This task
is made easier if the references both appear to be either
stronger or weaker. It is made more difficult if one refer-
ence appears to be stronger while the other appears to be
weaker. When the reference is the stronger stimulus, the
first reference will appear stronger but adaptation from
the stronger stimuli in the test will render the final reminder
reference to be weaker. The references would appear to be
different and so cause confusion. When the reference is
weaker, the first reference would appear weaker and any
adaptation effects will render the final reminder reference
to also appear weaker. The two references would appear
to be close enough to be the ‘same’.
These explanations were based on adaptation effects and
specific to the NaCl/NaCl model system that used in this
study. Using different model system or food system would
be expected to give different results. Introduction of other
sensory modalities for example trigeminal stimulation
(
) would be expected to give still differ-
ent results. Regarding DTF, in fact, some studies reported
the superior performance of the weaker reference (S-odd)
over the stronger reference (W-odd) version of duo–trio
(
working with blended whiskies;
working with a purified water/low
concentration NaCl model system) while others reported
no difference in performance between S-odd and W-odd
(
using vanilla flavored yogurts,
using orange flavored drinks).
Many previous studies investigating the difference test
sensitivity, including the studies that developed the Sequen-
tial Sensitivity Analysis (SSA), used a purified water/low
concentration NaCl model system (
etc.). Yet, water stimuli have contrasting properties when
tasted adjacent to NaCl stimuli (
) and different food stimuli would elicit different
extent of adaptation. Therefore, there should be a caution
when these results were generalized to the performance of
difference tests using different, often more complex, food
stimuli.
Acknowledgement
The authors would like to thank Michael O’Mahony for
his helpful comments and advice on the manuscript which
improved its clarity and readability.
References
ASTM (1968). Manual on sensory testing methods. Philadelphia: American
Society for testing and materials.
Carstens, E., Carstens, M. I., Dessirier, J.-M., O’Mahony, M., Simons, C.
T., Sudo, M., et al. (2002). It hurts so good: Oral irritation by spices
and carbonated drinks and the underlying neural mechanisms. Food
Quality and Preferences, 13, 431–443.
Dessirier, J. M., & O’Mahony, M. (1999). Comparison of d
0
values for the
2-AFC (paired comparison) and 3-AFC discrimination methods:
Thurstonian models, sequential sensitivity analysis and power. Food
Quality and Preference, 10, 51–58.
Dessirier, J. M., Siffermann, J. M., & O’Mahony, M. (1999). Taste
discrimination by the 3-AFC method: Testing sensitivity predictions
regarding particular tasting sequences based on the sequential sensi-
tivity analysis model. Journal of Sensory Studies, 14, 271–287.
Ennis, D. M. (1993). The power of sensory discrimination methods.
Journal of Sensory Studies, 8, 353–370.
Ennis, D. M., & O’Mahony, M. (1995). Probabilistic models for sequential
taste effects in triadic choice. Journal of Experimental Psychology:
Human Perception and Preference, 5, 1088–1097.
Kim, H.-J., Jeon, S. Y., Kim, K.-O., & O’Mahony, M. (2006). Thursto-
nian models and variance I: Experimental confirmation of cognitive
strategies for difference tests and effects of perceptual variance. Journal
of Sensory Studies, 21, 465–484.
H.-S. Lee, K.-O. Kim / Food Quality and Preference 19 (2008) 97–102
101
Kuesten, C. L. (2001). Sequential use of the triangle, 2-AC, 2-AFC, and
same–different methods applied to a cost-reduction effort: consumer
learning acquired throughout testing and influence on preference
judgements. Food Quality and Preferences, 12, 447–455.
Lau, S., O’Mahony, M., & Rousseau, B. (2004). Are three-sample
tasks less sensitive than two-sample tasks? Memory effects in the
testing of taste discrimination. Perception and Psychophysics, 66,
464–474.
Lawless, H. T., & Heymann, H. (1996). Sensory Evaluation of Food:
Principles and practices. New York: Chapman and Hall.
Lee, H.-S., & O’Mahony, M. (2004). Sensory difference testing: Thursto-
nian models. Food Science and Biotechnology, 13, 841–847.
Lee, H.-S., & O’Mahony, M. (2007a). The evolution of a model: A review
of Thurstonian and conditional stimulus effects on difference testing.
Food Quality and Preference, 18, 369–383.
Lee, H.-S., & O’Mahony, M. (2007b). Difference test sensitivity: Cognitive
contrast effects. Journal of Sensory Studies, 22, 17–33.
Liggett, R. E., & Delwiche, J. F. (2005). The beta-binomial model:
Variability in overdispersion across methods and over time. Journal of
Sensory Studies, 20, 48–61.
Macmillan, N. A., & Creelman, C. D. (2005). Detection Theory: A User’s
Guide (second ed.). New York: Cambridge University Press.
Meilgaard, M., Civille, G. V., & Carr, B. T. (1991). Sensory evaluation
techniques (second ed.). Boca Raton, Florida: CRC Press.
Mitchell, J. W. (1956). The effect of assignment of testing materials to the
paired and odd position in the duo–trio taste difference test. Food
Technology, 10, 169.
O’Mahony, M., Masuoka, S., & Ishii, R. (1994). A theoretical note on
difference tests: Models, paradoxes and cognitive strategies. Journal of
Sensory Studies, 9, 247–272.
O’Mahony, M., & Odbert, N. (1985). A comparison of sensory difference
testing procedures: Sequential sensitivity analysis and aspects of taste
adaptation. Journal of Food Science, 50, 1055–1058.
O’Mahony, M., & Rousseau, B. (2002). Discrimination testing: A few
ideas, old and new. Food Quality and Preference, 14, 157–164.
O’Mahony, M., Thieme, U., & Goldstein, L. R. (1988). The warm-up
effect as a measure of increasing the discriminability of sensory
difference tests. Journal of Food Science, 53, 1848–1850.
Peryam, D. R. (1958). Sensory difference tests. Food Technology, 12,
231–236.
Peryam, D. R., & Swartz, V. W. (1950). Measurements of sensory
differences. Food Technology, 4, 390.
Rousseau, B., Meyer, A., & O’Mahony, M. (1998). Power and sensitivity
of the same–different test: Comparison with triangle and duo–trio
methods. Journal of Sensory Studies, 13, 149–173.
Rousseau, B., & O’Mahony, M. (1997). Sensory difference tests: Thurs-
tonian and SSA predictions for vanilla flavored yogurts. Journal of
Sensory Studies, 12, 127–146.
Rousseau, B., & O’Mahony, M. (2000). Investigation of the effect of
within-trial retasting and comparison of the dual-pair, same–different
and triangle paradigms. Food Quality and Preference, 11, 457–464.
Rousseau, B., & O’Mahony, M. (2001). Investigation of the dual-pair
method as a possible alternative to the triangle and same–different
tests. Journal of Sensory Studies, 16, 161–178.
Rousseau, B., Rogeaux, M., & O’Mahony, M. (1999). Mustard discrim-
ination by same–different and triangle tests: Aspects of irritation and s
criteria. Food Quality and Preference, 10, 173–184.
Rousseau, B., Stroh, S., & O’Mahony, M. (2002). Investigating more
powerful discrimination tests with consumers: Effects of memory and
response bias. Food Quality and Preference, 13, 39–45.
Stillman, J. A., & Irwin, R. J. (1995). Advantages of the same–different
method over the triangular method for the measurement of taste
discrimination. Journal of Sensory Studies, 10, 261–272.
Tedja, S., Nonaka, R., Ennis, D. M., & O’Mahony, M. (1994). Triadic
discrimination testing: Refinement of Thurstonian and sequential
sensitivity analysis approaches. Chemical Senses, 19, 279–301.
Thieme, U., & O’Mahony, M. (1990). Modifications to sensory difference
test protocols: The warmed up paired comparison, the single standard
duo–trio and the A-Not A test modified for response bias. Journal of
Sensory Studies, 5, 159–176.
102
H.-S. Lee, K.-O. Kim / Food Quality and Preference 19 (2008) 97–102