Investigating more powerful discrimination tests with consumers:

effects of memory and response bias

Benoıˆt Rousseau*, Stefanie Stroh, Michael O’Mahony

Department of Food Science and Technology, University of California, Davis CA 95616, USA

Received 4 April 2000; received in revised form 19 June 2001; accepted 20 July 2001

Abstract

Two experiments were conducted to investigate the sensitivity of four discrimination methods when they were performed by

consumers. In Experiment I, the influence of memory in the duo-trio method was studied. Three versions of the duo-trio method
with different memory requirements were considered. Calculated d

0

values indicated a higher sensitivity for the duo-trio with the

reference tasted between the two test samples (DTM), illustrating the importance of memory in sensory discrimination testing. In
Experiment 2, four discrimination tests were compared: the triangle, the DTM, the same–different and the dual-pair tests. The dual-
pair test was predicted to increase the d

0

value of the same–different test by eliminating the large response bias variations between

consumers. Results indicated no significant differences in d

0

among the protocols. Thus the dual-pair method was not able to

improve the sample discrimination ability of the same–different test. # 2001 Elsevier Science Ltd. All rights reserved.

Keywords:

Discrimination tests; Power; Consumers; Memory; Response bias

1. Introduction

The area of discrimination testing has recently been

an active area of research in sensory evaluation. Dis-
crimination methods are broadly used both in the
industry and academia, their applications ranging from
daily quality control measurements, to the study of the
impact of ingredient or process changes, to the investi-
gation of the ability of consumers to discriminate
among products. Because of the costly implications that
can be induced by imprecise results or conclusions,
numerous studies have been conducted to attempt to
define the most appropriate discrimination methods. By
‘most appropriate’ was meant the most powerful
method of investigation. The more powerful the test, the
more likely the detection of a sensory difference when it
exists. The analysis of results in early studies was often
based on the guessing model (illustrated by the use of
binomial tables) and yielded conflicting results (Bucha-
nan, Givon, & Goldman, 1987; Byer & Abrams, 1953;
Dawson & Dochterman, 1951; Filipello, 1956; Franc¸ois

& Sauvageot, 1988; Gridgeman, 1955; Helm & Trolle,
1946; Hopkins, 1954; Hopkins & Gridgeman, 1955;
MacRae & Geelhoed, 1992; Pfaffman, 1954; Pokorny,
Marcı´n, & Davı´dek, 1981; Raffensperger & Pilgrim,
1956; Vessereau, 1965). The work of Thurstone (1927),
later used by Frijters (1979) to solve the ‘paradox of dis-
criminatory non-discriminators’ (Byer & Abrams, 1953;
Gridgeman, 1970), allowed an explanation of these
inconsistencies and provided a more consistent analysis
of results obtained from discrimination tests and other
sensory evaluation procedures. This approach involves
the calculation of a d

0

value which is an index describing

the perceived degree of difference between two products.
The larger the value, the more different the products.

This approach has been frequently applied in recent

studies in order to investigate which sensory protocol
would be the most appropriate to study small sensory
differences between products (Delwiche & O’Mahony,
1996; Dessirier & O’Mahony, 1998; Frijters, 1980;
Geelhoed, Macrae, & Ennis, 1994; Hautus & Irwin,
1995; Huang & Lawless, 1998; Irwin, Hautus, & Still-
man, 1992; Irwin, Stillman, Hautus, & Huddleston,
1993; Masuoka, Hatjopoulos, & O’Mahony, 1995;
Rousseau & O’Mahony, 1997, 2000, 2001; Rousseau,
Meyer, & O’Mahony, 1998; Rousseau, Rogeaux, &

0950-3293/01/$ - see front matter # 2001 Elsevier Science Ltd. All rights reserved.
P I I : S 0 9 5 0 - 3 2 9 3 ( 0 1 ) 0 0 0 5 5 - 6

Food Quality and Preference 13 (2002) 39–45

www.elsevier.com/locate/foodqual

* Corresponding author. Tel.: [phone]; fax: [phone]-

4759.

E-mail address:

[e-mail] (B. Rousseau).

O’Mahony, 1999; Stillman, 1993; Stillman & Irwin,
1995; Tedja, Nonaka, Ennis, & O’Mahony, 1994).

While the majority of these studies have been con-

ducted under very controlled conditions in the labora-
tory, few have been conducted under less controlled
conditions of consumer testing. Stillman (1993) con-
firmed

the

paradox

of

discriminatory

non-dis-

criminators

using

consumers;

the

subjects

could

discriminate significantly better between two versions of
an onion party dip using a 3-Alternative Forced Choice
(3-AFC) than using a triangle test. This illustrated the
higher power of the 3-AFC over the triangle test (Ennis,
1993) because of its more effective cognitive strategy.
One of the drawbacks of the 3-AFC test, or its two-
sample counterpart, the paired comparison test (2-
AFC), is that it requires the specification of the nature
of the difference in the instructions. Yet, this informa-
tion is not always readily available to the experimenter.
Another study by Rousseau et al. (1998) compared the
performance of the duo-trio, triangle and same–differ-
ent tests with consumers. These three protocols have the
advantage of not requiring any description of the dimen-
sion of the difference; however, they lack the statistical
power of the 2-AFC and 3-AFC tests (Ennis, 1993). In
that study, Rousseau et al. found that the same–differ-
ent test was the most powerful and most sensitive of the
three protocols thanks to its more effective cognitive
strategy and lower memory requirements, respectively.

While some studies have shown the importance of

memory when using scaling procedures (Kim &
O’Mahony, 1998), more studies have investigated its
effects in discrimination testing. For example, memory
decay over time has been shown to affect performance
in the same–different test (Cubero, de Almeida, &
O’Mahony, 1995; de Almeida, Cubero, & O’Mahony,
1999). Furthermore, memory is thought to be the main
factor giving an advantage to the same–different over
the triangle test (Rousseau et al., 1998, 1999; Rousseau
& O’Mahony, 2000; 2001). Memory is also suspected of
being partly responsible for the higher sensitivity of the
2-AFC over the 3-AFC (Dessirier & O’Mahony, 1998;
Rousseau & O’Mahony, 1997).

A potentially useful way of further confirming the

significant effects of memory would be to change only
slightly the design of a particular protocol, while alter-
ing its memory properties, and observe its effect on the
protocol’s performance. A candidate for such an inves-
tigation is the duo-trio method. The traditional pre-
sentation involves the tasting of the reference first, then
the two unknown or ‘test’ samples. It is possible that
once the third sample is evaluated, the memory trace or
‘engram’ of the reference stored in the short-term mem-
ory would have degenerated; it would be ‘blurred

0

, and

might cause the judge to provide an incorrect answer.
One could slightly modify the mode of presentation by
introducing the reference between the two alternatives,

so that both alternatives are evaluated immediately
before or after the reference, reducing memory require-
ments for the comparisons. This issue was investigated
in Experiment 1.

While the same–different protocol is more sensitive

and powerful than the traditional triangle and duo-trio
tests, it might be possible to further improve its perfor-
mance. The factor that needs to be taken into con-
sideration is response bias, in the form of their t
criterion, a psychological ‘yardstick’ of degree of differ-
ence. If the perceived difference between the two sam-
ples is smaller than his t criterion, the subject will
consider the stimuli as ‘same’, while a larger difference
will induce him to consider them as ‘different’. Signal
Detection Theory and Thurstonian models successfully
deal with this response bias to yield an unbiased
measure of discrimination, the d

0

value. However, one

of the limitations of combining data from different sub-
jects in the same–different test is that subjects exhibit t
criteria of different sizes. Such combination will cause
an underestimation of the d

0

value (Hautus, 1997; Mac-

millan & Kaplan, 1985).

The dual-pair test is thus investigated in an attempt to

control this effect. In the dual-pair test, also called 4-
Interval AX or 4IAX in psychology (e.g. Kaplan, Mac-
millan, & Creelman, 1978), the subject is presented with
two pairs of samples simultaneously. One is a pair of
identical samples, the other of different samples. In the
pair consisting of two different samples, one of the
samples is the same as the sample used in the pair of
identical stimuli. The task of the subject is to determine
which pair has the same and which pair has different
samples. This protocol has been studied previously
under controlled conditions (Rousseau & O’Mahony,
2000, 2001), while a Thurstonian model has been devel-
oped allowing the investigation of its statistical power
(Rousseau & Ennis, 2001). Information obtained from
these experiments and new models did not indicate an
advantage for the dual-pair test. Its statistical power
was found to be more similar to that of the duo-trio test
and thus inferior to that of the same–different test.
Furthermore, it did not yield larger d

0

values. Not

yielding larger d

0

values than the same–different test

might have been due to the fact that in controlled con-
ditions subjects were familiar with the products under
study, limiting the large variations in the subjects’ t cri-
teria which would have hindered performance on the
same–different test. It is possible that such variations
could have a larger influence on the same–different test
when testing highly ‘uncalibrated

0

subjects such as con-

sumers. Here, an advantage of the dual-pair test over
the same–different test might possibly be observed. This
particular issue was studied in Experiment 2.

Thus, in summary, the present experiments further

investigated the effect of memory in discrimination test-
ing and the possible experimental advantages of the

40

B. Rousseau et al. / Food Quality and Preference 13 (2002) 39–45

dual-pair paradigm. In Experiment 1 concerning mem-
ory, the sensitivities of three versions of the duo-trio test
were compared using 96 consumers. The objective was
to investigate the effect of the position of the reference
versus that of the two alternative samples. In Experi-
ment 2, the relative performances of the dual-pair and
same–different tests were compared using 144 con-
sumers. The duo-trio and triangle tests were also inclu-
ded in the study so as to provide more complete
information about the advantages and disadvantages of
the traditional discrimination protocols.

2. Experiment 1

2.1. Materials and methods

2.1.1. Judges

Ninety-six (96) subjects (51 M, 45 F; age range 13–56

years average age 22.7 years) participated in Experiment
I. They were students, faculty, staff and their friends
intercepted on the campus of the University of Cali-
fornia, Davis.

2.1.2. Products

The stimuli for the experiment consisted of two ver-

sions of a non-carbonated orange flavored beverage
(The Gatorade company, Chicago, IL, USA), available
as an instant mix. One stimulus (A) was prepared using
80 g of product per liter, mixed with deionized water.
The second stimulus (B) was prepared the same way but
10 g of sucrose (Fischer Grade, Fisher Scientific, Fair
Lawn, NJ, USA) were also added to create a small sen-
sory difference. Samples were served in 10 ml aliquots in
approximately thirty-milliliter clear plastic cups (Serco
1-oz Plastic Cups, S. E. Rykoff & CO., Los Angeles,
CA, USA).

2.1.3. Procedure

Having intercepted the judge, an experimenter first

established rapport and then took personal details. A
second experimenter then explained the methods and
performed the test.

Each judge performed three versions of the duo-trio

test: the traditional duo-trio with the reference tasted
first, the duo-trio with the reference tasted between the

two test samples (DTM) and a third version designated
as the 2-Distance-AFC or 2-D-AFC. For the first two
protocols, the subject was asked to indicate which of the
two alternatives was the most similar to the reference.
The 2-D-AFC had the same presentation order as the
DTM, but the subject was asked to indicate which
change of sensation was the larger. Thus, they reported
whether the sensation change was greater between the
first and the second sample, or between the second and
the third sample, rather than indicate which sample was
the more similar to the middle stimulus. Here, they were
not told that among the three samples, two were identical.

One deionized water rinse was taken before each test.

Before tasting the first sample of the triad, a primer
sample was taken in order to prevent the distorting
effect of the rinsing water on the taste of the first sam-
ple. The primer was always sample A. Subjects were
required to take the whole 10 ml of each sample at once
and swallow it. Water rinses were also swallowed. No
retasting was allowed so as to allow memory to have a
larger influence. The order of presentation of the tests
and the order in which samples appeared in a test were
counterbalanced over judges. In order to prevent differ-
ential sequence effects among the various protocols,
only two presentation orders were used, namely ‘AAB’
and ‘BBA’. Each judge encountered only one of the two
possible triads over the three protocols performed. Jud-
ges responded verbally, no feedback was given. Session
lengths

from

interception

to

termination

lasted

approximately 10–15 mins.

2.2. Results

The d

0

values for the three versions of the duo-trio test

are presented in Table 1. The d

0

values for the duo-trio

and DTMwere obtained from the proportions of cor-
rect responses using tables (Ennis, 1993). The d

0

value

for the 2-D-AFC was also obtained using the tables,
since it can be argued that the cognitive strategy used by
the subjects to perform this particular protocol is the
same as that assumed for the duo-trio.

Variances for testing significant differences were

obtained using the approach proposed by Bi, Ennis, and
O’Mahony (1997). It was found that the d

0

value

obtained with the DTMwas significantly higher than
those obtained with the traditional duo-trio and the

Table 1
Results for the three protocols compared in Experiment 1—proportion of correct answers, d

0

, variance of d

0

and 95% confidence intervals

a

Protocol

Number of correct answers (out of 96)

d

0

Variance

95% Confidence intervals

Duo-Trio

62

1.4a

0.08

(0.8; 2.0)

DTM74

2.2b

0.08

(1.6; 2.7)

2-D-AFC

61

1.3a

0.09

(0.8; 1.9)

a

DTMd

0

value significantly different from duo-trio (P=0.05) and 2-D-AFC (P=0.03). DTM, duo-trio with reference in the middle; 2-D-AFC,

2-distance-alternative forced choice.

B. Rousseau et al. / Food Quality and Preference 13 (2002) 39–45

41

2-D-AFC (specifically at P=0.05 and 0.03, respectively,
after an overall screening at P=0.06), while the latter
two procedures did not show any significant difference
(P=0.8).

For each of the three duo-trio protocols, no differ-

ences were found between the two different sequences of
tasting (AAB, BBA). d

0

values are given in Table 2.

Differences were not significant (P 5 0.33).

It was decided to use the DTMin Experiment 2 since it

appeared to be the most sensitive version of the duo-trio.

3. Experiment 2

3.1. Materials and methods

3.1.1. Judges

One hundred and forty-four (144) subjects (78 M, 66

F; age range 18–71; mean age 24.3 years) participated in
Experiment 2. They were selected from the same pool of
subjects as in Experiment 1.

3.1.2. Products

The stimuli for the experiment were the same as those

used in Experiment 1. The only difference was a reduc-
tion of the sugar concentration in sample B to 8 g/l from
10 g/l, to assure better confusability of the stimuli.

3.1.3. Procedure

The testing procedures were similar to those used in

Experiment 1. This included the rinsing protocol and
the use of a primer. Consumers swallowed the stimuli
and the rinses. Each consumer performed one test of
each protocol: triangle, DTM, dual-pair and same–dif-
ferent. It should be noted that the last protocol con-

sisted of two successive pairs of samples, one being of
identical samples, the other of different samples. How-
ever, subjects were not aware of this arrangement. The
version of the same–different test used here, sometimes
labeled as the ‘longer’ version (Rousseau et al., 1999), is
statistically more powerful than the traditional triangle
and duo-trio tests (Ennis, 1996). For each pair of sti-
muli, subjects were asked to state whether they thought
the samples were the same or different and whether they
were sure or not so sure of their answer.

Each consumer was first introduced to the two sam-

ples A and B to be tested and was allowed to sample
both. This ‘familiarization procedure’ was performed to
allow the judge to have some preconceived idea of the
extent of the difference to be detected. This allowed
them to adjust their criterion of difference (Macmillan &
Creelman, 1991) such that if they detected a difference,
they would report it.

The order of presentation of the tests and the order in

which samples appeared in a test were counterbalanced
over judges. A block of 48 subjects permitted a complete
counterbalancing. Thus three of these blocks were used
here. Judges responded verbally, no feedback was given.
Session lengths from interception to termination lasted
approximately 10–15 min.

4. Results

Results are presented in Table 3. For the DTM, dual-

pair and triangle tests, d

0

values and their variances were

obtained from the proportion of correct responses using
tables (Ennis, 1993; Rousseau & Ennis, 2001). For the
same–different test, an analysis used in previous studies
(Rousseau et al., 1998, 1999; Rousseau & O’Mahony,
2000; 2001) was applied. This analysis was based on
Thurstone’s law of categorical judgment (Ennis, 1998;
Thurstone, 1927). The analysis was applied to difference
ratings and used the method of maximum likelihood to
estimate d

0

, t and the variance–covariance matrix for

these estimates.

Using the approach proposed by Bi et al. (1997), no

significant differences were observed among the four d

0

values (P=0.7). However, a trend was observed for the
triangle test to be slightly less sensitive than the other
three procedures.

Table 2
d

0

values for each duo-trio sequence (n=48)

Protocol

AAB

BBA

Duo-Trio

1.6

a

1.1

a

DTM1.9

b

2.5

b

2-D-AFC

1.5

c

1.1

c

a,b,c

For each protocol, d

0

values for sequences AAB and BBA are

not significantly different (P 5 0.33). DTM, duo-trio with reference in
the middle; 2-D-AFC, 2-distance-alternative forced choice.

Table 3
Results for the four protocols compared in Experiment 2 — proportion of correct answers, d

0

, variance of d

0

and 95% confidence intervals

Protocol

Number of correct answers (out of 144)

d

0

Variance

95% Confidence intervals

Triangle

62

1.1

a

0.06

(0.6; 1.6)

DTM96

1.5

a

0.05

(1.1; 2.0)

Same–different

N/A

1.4

a

0.04

(1.0; 1.8)

Dual-pair

90

1.4

a

0.06

(0.9; 1.9)

a

d

0

values are not significantly different (P=0.7).

42

B. Rousseau et al. / Food Quality and Preference 13 (2002) 39–45

As for Experiment 1, no clear trend was apparent

when looking at possible sequence effects. This can be
observed separately for different test protocols. First,
for the same–different test, the pattern of sequences
obtained with the four possible pairs of stimuli could be
examined. This is an approach similar to that used in
the sequential sensitivity analysis (O’Mahony & Odbert,
1985). The appropriate way of performing the analysis
would be to carry out a

2

test including the four

sequences. However, this analysis is complicated by the
existence of response bias. For instance, if the subjects
exhibited a very large t criterion, the results of the pair
AA and BB would tend to appear more ‘correct’ than
those of the pairs AB and BA, giving a significant

2

test. This would obscure the effect of adaptation and
sequencing since the results would be contaminated by a
psychological bias. However, it is still relevant to com-
pare the responses of the pairs AA with BB on one
hand, and AB with BA on the other hand (Table 4). No
significant differences are observed (

2

, P=0.7 for either

comparison).

For the remaining tests, the effects can be examined

by looking at the two halves of each of the triangle,
DTMand dual-pair tests. Results are presented in
Table 5. Using the same analysis as in Experiment I
(Table 2), no significant difference between the two
halves (‘A-odd

0

and ‘B-odd

0

) were observed (P 5 0.17).

5. Discussion

The results from Experiment 1 confirmed the sig-

nificant effect played by memory in discrimination tests.
Merely changing the position of the reference in the
duo-trio test significantly improved the sensitivity of the
protocol. This result substantiates the higher sensitivity
of tests with only two samples, such as the same–differ-
ent and 2-AFC tests over those with three samples,
observed in previous studies. This first result seems to
indicate that tasting the reference first in the duo-trio
test is not the most appropriate design for this parti-
cular protocol. In order to obtain the best discrimina-
tion between products, the reference should be tasted
between the two alternative samples.

However, it needs to be noted that subjects were not

allowed to retaste the samples, therefore permitting
memory effects to have a non-negligible influence on the
outcome of the test. It is possible that retasting might
limit the advantage of the DTM; further experimenta-
tion is needed to investigate this issue. Nevertheless,
when the nature of the product does not allow retasting
because of sensory fatigue (e.g. red pepper, hard liquors,
. . .), the DTMappears like the most appropriate version
of the duo-trio test.

The 2-D-AFC did not show any particular advantage.

It may have been expected to give a higher proportion of
tests correct like the DTM. Yet, subjects reported that the
task was difficult and they might have reverted to an
alternative strategy. Perhaps more practiced and trained
subjects in laboratory conditions may have more success.

Experiment 2 did not show any significant differences

among the sensitivities of the four protocols. These
results would confirm Thurstonian predictions. How-
ever, trends seen in previous research (Rousseau et al.,
1998, 1999; Rousseau & O’Mahony, 2000; 2001; Still-
man & Irwin, 1995) were observed. Of the four proto-
cols, the same–different test was the most powerful as
indicated by its lower variance.

The dual-pair did not yield a larger d

0

value than that

of the same–different test as predicted. The reason could
be that larger memory requirements counterbalanced
any stabilizing effect on the judges’ t criteria that was
provided by using a forced-choice method. A second pos-
sibility is that it might be due to the fact that the familiar-
ization procedure performed by the subjects prior to
starting the experiment was sufficient to prevent large
variations in the size of the subjects’ t criteria in the
same–different test and so not reduce its performance.

Nevertheless, this result indicates that the dual-pair

method does not appear to be a good alternative to the
same–different method. It exhibits at best the same sen-
sitivity as the same–different, while its statistical power
is more limited. Thus, the same–different test with a
familiarization procedure appears more appropriate as
a sensory discrimination method for consumer testing.

Table 4
Distribution of responses for the four pairs presented in the same–
different test (in bold, responses which can be scored as ‘correct’)

a

Pair

Number of answers (/72)

‘‘Correct’’

b

‘‘Incorrect’’

b

Same

AA

39

33

BB

41

31

Different

AB

42

30

BA

44

28

a

The pattern of responses is not significant (

2

, P=0.7 for com-

parison of ‘same’ pairs, P=0.7 for comparison of ‘different’ pairs).
The categories ‘‘Different sure’’ and ‘‘Different not sure’’ on one hand,
and ‘‘Same sure’’ and ‘‘Same not sure’’ on the other have been com-
bined to permit easier comparison.

b

For the pairs AB and BA, a response ‘Different’ will be con-

sidered correct, while for the pairs AA and BB, a response ‘Same’ will
be considered correct.

Table 5
d

0

values for each type of sequence for three testing protocols (n=72)

Protocol

‘B’ odd

‘A’ odd

Triangle

0.9

a

1.2

a

DTM1.9

b

1.2

b

Dual-pair

1.7

c

1.0

c

a,b,c

For each protocol, d

0

values of each half of the tasting sequen-

ces (‘B’ odd vs. ‘A’ odd) are not significantly different (P 5 0.17).
DTM, duo-trio with reference in the middle.

B. Rousseau et al. / Food Quality and Preference 13 (2002) 39–45

43

Acknowledgements

This research was supported by a grant from the

Groupe Danone, France.

References

Bi, J., Ennis, D. M., & O’Mahony, M. (1997). How to estimate and

use the variance of d

0

from difference tests. Journal of Sensory

Studies

, 12, 87–104.

Buchanan, B., Givon, M. M., & Goldman, A. (1987). Measurement of

discrimination ability in taste tests: an empirical investigation.
Journal of Marketing Research

, 24, 154–163.

Byer, A. J., & Abrams, D. (1953). A comparison of the triangular and

two-sample taste-test methods. Food Technology, 7, 185–187.

Cubero, E., de Almeida, T. C. A., & O’Mahony, M. (1995). Cognitive

aspects of difference testing: memory and interstimulus delay. Jour-
nal of Sensory Studies

, 10, 307–324.

Dawson, E. H., & Dochterman, E. (1951). A comparison of sensory

methods of measuring differences in food qualities. Food Technol-
ogy

, 5, 79–81.

de Almeida, T. C. A., Cubero, E., & O’Mahony, M. (1999). Same–

different discrimination tests with interstimulus delays up to one
day. Journal of Sensory Studies, 14, 1–18.

Delwiche, J., & O’Mahony, M. (1996). Flavour discrimination—an

extension of Thurstonian paradoxes to the tetrad method. Food
Quality and Preference

, 7, 1–5.

Dessirier, J. M., & O’Mahony, M. (1998). Comparison of d

0

values for

the 2-AFC (paired comparison) and 3-AFC discrimination meth-
ods: Thurstonian models, sequential sensitivity analysis and power.
Food Quality and Preference

, 10, 1–8.

Ennis, D. M. (1993). The power of sensory discrimination methods.

Journal of Sensory Studies

, 8, 353–370.

Ennis, D. M. (1996). Personal communication.
Ennis, D. M. (1998). Personal communication.
Filipello, F. (1956). A critical comparison of the two-sample and tri-

angular binomial designs. Food Research, 21, 235–241.

Franc¸ois, P., & Sauvageot, F. (1988). Comparison of the efficiency of

pair, duo-trio and triangle tests. Journal of Sensory Studies, 3, 81–
94.

Frijters, J. E. R. (1979). The paradox of discriminatory non-

discriminators resolved. Chemical Senses & Flavor, 4, 355–358.

Frijters, J. E. R. (1980). Three-stimulus procedures in olfactory psy-

chophysics: an experimental comparison of Thurstone-Ura and
three-alternative forced-choice models of signal detection theory.
Perception and Psychophysics

, 28, 390–397.

Geelhoed, E. N., Macrae, A. W., & Ennis, D. M. (1994). Preference

gives more consistent judgments than oddity only if the task can be
modeled as forced choice. Perception and Psychophysics, 55, 473–
477.

Gridgeman, N. T. (1955). Taste comparisons: two samples or three?

Food Technology, 9, 148–150.

Gridgeman, N. T. (1970). A re-examination of the two-stage triangle

test for the perception of sensory differences. Journal of Food
Science

, 35, 87–91.

Hautus, M. J. (1997). Calculating estimates of sensitivity from group

data: pooled versus averaged estimators. Behavior Research Meth-
ods Instruments & Computers

, 29, 556–562.

Hautus, M. J., & Irwin, R. J. (1995). Two models for estimating the

discriminability of foods and beverages. Journal of Sensory Studies,
10

, 203–215.

Helm, E., Trolle, B. (1946). Selection of a taste panel. Wallerstein Lab.

Communications, 9, 181–194.

Hopkins, J. W. (1954). Some observations on sensitivity and repeat-

ability of triad taste difference tests. Biometrics, 10, 521–530.

Hopkins, J. W., & Gridgeman, N. T. (1955). Comparative sensitivity

of pair and triad flavor intensity difference tests. Biometrics, 11, 63–
68.

Huang, Y. T., & Lawless, H. T. (1998). Sensitivity of the ABX dis-

crimination test. Journal of Sensory Studies, 13, 229–239.

Irwin, R. J., Hautus, M. J., & Stillman, J. A. (1992). Use of the recei-

ver operating characteristic in the study of taste perception. Journal
of Sensory Studies

, 7, 291–314.

Irwin, R. J., Stillman, J. A., Hautus, M. J., & Huddleston, L. M.

(1993). The measurement of taste discrimination with the same–dif-
ferent task: a detection theory analysis. Journal of Sensory Studies,
8

, 229–239.

Kaplan, H. L., Macmillan, N. A., & Creelman, C. D. (1978). Tables of

d

0

for

variable-standard

discrimination

paradigms.

Behavior

Research Methods & Instrumentation

, 10, 796–813.

Kim, K. O., & O’Mahony, M. (1998). A new approach to category

scales of intensity I. Traditional versus rank rating. Journal of Sen-
sory Studies

, 13, 241–249.

Macmillan, N. A., & Kaplan, H. L. (1985). Detection theory analysis

of group data: estimating sensitivity from average hit and false-
alarm rates. Psychological Bulletin, 98, 185–199.

MacRae, A. W., & Geelhoed, E. N. (1992). Preference can be more

powerful than detection of oddity as a test of discriminability. Per-
ception and Psychophysics

, 51, 179–181.

Masuoka, S., Hatjopoulos, D., & O’Mahony, M. (1995). Beer bitter-

ness detection: testing Thurstonian and sequential sensitivity analy-
sis models for triad and tetrad methods. Journal of Sensory Studies,
10

, 295–306.

O’Mahony, M., & Odbert, N. (1985). A comparison of sensory differ-

ence testing procedures: sequential sensitivity analysis and aspects of
taste adaptation. Journal of Food Science, 50, 1055–1058.

Pfaffman, C. (1954). Variables affecting difference tests In: N.l.A. o.

S.-N. R. Council, Peryam, D.R., Pilgrim, F.J., Peterson, M.S.
Food acceptance testing methodology, a symposium

(pp. 4–20)

Washington DC: Nat’l Academy of Sciences-National Research
Council.

Pokorny, J., Marcı´n, A., & Davı´dek, J. (1981). Comparison of the

efficiency of triangle and tetrad tests for discriminatory sensory
analysis of food. Die Nahrung, 25, 561–564.

Raffensperger, E. L., & Pilgrim, F. J. (1956). Knowledge of the sti-

mulus variable as an aid in discrimination tests. Food Technology,
10

, 254–257.

Rousseau, B., & Ennis, D. M. (2001). A Thurstonian model for the

dual pair (4IAX) discrimination method. Perception & Psycho-
physics

(in press).

Rousseau, B., Meyer, A., & O’Mahony, M. (1998). Power and sensi-

tivity 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:

Thurstonian 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–differ-
ent 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 dis-

crimination by same–different and triangle tests: aspects of irrita-
tion, memory and t criteria. Food Quality and Preference, 10, 173–
184.

Stillman, J. A. (1993). Response selection, sensitivity, and taste-test

performance. Perception and Psychophysics, 54, 190–194.

Stillman, J. A., & Irwin, R. J. (1995). Advantages of the same–differ-

44

B. Rousseau et al. / Food Quality and Preference 13 (2002) 39–45

ent 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.

Thurstone, L. L. (1927). A law of comparative judgment. Psychologi-

cal Review

, 34, 273–286.

Vessereau, A. (1965). Les me´thodes statistiques applique´es au test des

caracte`res organoleptiques. Annales de la Nutrition et de l’Alimenta-
tion

, 19, 103–140.

B. Rousseau et al. / Food Quality and Preference 13 (2002) 39–45

45