Chem. Senses 24: 393-403,
1999
© Oxford University Press 1999
Taste Confusions following Gymnemic Acid Rinse
University of Connecticut Health Center, Farmington, CT 1 John B. Pierce Laboratory and Yale University, New Haven, CT, USA
Correspondence to be sent to: Janneane F. Gent, PhD, Department of Biostructure and Function, University of Connecticut Health Center, 263 Farmington Ave., Farmington, CT 06030-3705, USA. e-mail:gent{at}neuron.uchc.edu
 |
Abstract |
|---|
 Top Abstract IntroductionExperiment 1. Taste perception...Experiment 2. Effect of...Experiment 3. Effect of...General discussion and...References |
|---|
The effect of a gymnemic acid (GA) rinse, which simulated a sweet-taste deficit, was measured on human taste perception and identification. Taste ratings showed that GA reduced the intensities of sucrose and aspartame to 14% of pre-rinse levels; over the recovery interval of 30 min, these values increased linearly to 63% of the pre-rinse levels. Repeated presentations of a set of 10 stimuli (five primarily or partly sweet—sucrose, aspartame, and NaCl–sucrose, acid–sucrose and quinine–sucrose mixtures; and five nonsweet—NaCl, KCl, Na glutamate (MSG), quinineHCl and citric acid) for identification following water and GA rinses produced `taste confusion matrices' (TCMs). Correct identification of the sweet-tasting stimuli was reduced by 23% in presentations closely following the GA rinse, an effect that dissipated with time. Most misidentifications involved sucrose and mixtures containing sucrose. In a second TCM experiment, GA was presented frequently within each session to maintain the sweet taste deficit, which revealed itself as specific confusions. Rinsing with GA impaired discriminability of sweet–nonsweet pairs of stimuli but enhanced discriminability of the aspartame–(NaCl–sucrose) pair. GA had no effect on discriminability of nonsweet stimulus pairs. The results suggest that specific error patterns in the TCM could be used to identify quality-specific taste disorders.
|
Introduction |
|---|
|
TopAbstractIntroductionExperiment 1. Taste perception...Experiment 2. Effect of...Experiment 3. Effect of...General discussion and...References |
|---|
Our interest in exploring the use of stimulus identification reflects our goal of constructing objective, performancebased measures of suprathreshold taste function. This interest arises from our experience with patients presenting to the Taste and Smell Clinic of the Connecticut Chemosensory Clinical Research Center. Measures of identification have two virtues: they reveal perceptual performance at levels of stimulation that characterize much daily experience and they are objective, in the sense that responses can be classified as correct or incorrect. In a recent paper (Hettinger et al., 1999
),
we characterized a stimulus-identification task which has potential for evaluating suprathreshold
taste
function. The resulting `taste-confusion matrix' is composed of identification
response
frequencies to a set of 10 stimuli presented 10 times each; responses were chosen from among a
closed
set of 10 stimulus names. The response matrix reveals not only the number of correct
identifications, but
also the pattern of incorrect choices for each stimulus in the set.
Although taste-quality identification is routinely assessed in chemosensory clinical centers (Frank et al., 1995; Gent et al., 1997), patients are typically asked to identify
perceptual
qualities, such as sweet, sour, salty or
bitter, rather than stimuli, such as sugar, acid, salt or quinine. The data are reported simply in
terms of
frequency of quality identification (Cowart, 1989; Deemset al., 1991). In contrast, clinical tests of smell emphasize stimulus
identification (Doty et al., 1984; Cain and Gent,
1986; Wright, 1987). Patterns of errors in
stimulus-identification tasks may provide
important diagnostic information regarding smell function (Wright et al.,
1991). Such an approach has not been attempted before for taste function.
The array of 10 taste stimuli that we use includes both single substances and mixtures. We
ask the
subjects to identify each substance from a set of response labels containing the names of all of
the
stimuli. Groups of subjects with normal taste function performed consistently at this task (Hettinger et al., 1999). If a stimulus-identification test designed
along these lines is
to be useful in clinical settings, it needs to identify deficits and/or distortions in taste function. To
this
end, the present study seeks to show that the `taste-confusion matrix' (TCM) is
sensitive
to a simulated deficit in sweet taste perception engendered by application of gymnemic acid
(GA).
Work on the olfactory system shows that a smell deficit results in poor stimulus identification (Cowart et al., 1997).
Gymnemic acid, a mixture of triterpene saponins, was discovered in 1847 to temporarily
reduce or abolish the sweet taste of sugar in humans (Hellekant and van der Wel,
1989). Because GA
has a specific (Bartoshuk et al., 1969; Oakley, 1985) and profound (Frank et al.,
1992) effect on the sweetness of sugars and other sweet substances, the application
of GA serves as a
useful model of dissociated taste loss (Tomita and Horikawa, 1986).
Thus, in the present study we
examine how the application of GA to the tongue modifies the pattern of taste identification for a
TCM.
In particular, we hypothesized that pre-rinsing with GA would make stimuli that are primarily
sweet
tasting more difficult to identify. Furthermore, because binary mixtures are normally confused
with the
mixture components (Hettinger et al., 1999), we hypothesized
that after GA treatment,
mixtures containing one sweet-tasting and one nonsweettasting substance (e.g. sucrose plus
NaCl)
would be more difficult to distinguish from the nonsweet component (i.e. NaCl alone).
We report here how the application to the tongue of GA modifies the TCM. From the
matrix of correct and incorrect responses for each subject (the TCM) we calculate two
quantitative
measures of performance, namely, percent correct and transmitted information, T (Hettinger et al., 1999).
|
Experiment 1. Taste perception after GA rinse |
|---|
|
TopAbstractIntroductionExperiment 1. Taste perception...Experiment 2. Effect of...Experiment 3. Effect of...General discussion and...References |
|---|
Experiment 1 measured how the perceived intensity of sucrose, aspartame and NaCl taste varied as a function of time following application of GA.
Subjects
Six subjects (three women, three men), aged 20–55 (mean = 42, SD = 9 years), participated in one session. Subjects for this and subsequent experiments were recruited from the students and staff of the University of Connecticut Health Center. This study was approved by the Institutional Review Board of the University of Connecticut Health Center. All subjects gave informed consent for participation.
Stimuli and treatment rinse
The stimuli used were 0.3 M sucrose (reagent grade; Baker, Phillipsburg, NJ), 3.0 mM
aspartame (for laboratory use; Searle, Arlington Heights, IL), 0.1 M NaCl (Baker, reagent grade)
and
water (reverse osmosis deionized). Sucrose and aspartame are primarily sweet and NaCl is
primarily
salty (DuBois et al., 1991; Smith and van der
Klaauw, 1995; Hettinger et al., 1996).
The GA treatment rinse was prepared by a modification of a previously published procedure (Warren
and Pfaffmann, 1959). Dried leaves of Gymnema sylvestre (50 g) were
mixed with 1000 ml of
deionized water and heated at 80°C for 4 h. After pressing out the liquid, remaining solids
were
removed by centrifugation. The cooled extract (~800 ml) was acidified with 50 ml of 1 M HCl
and the
tan-brown precipitate collected by centrifugation. The dried product (~1 g) was dissolved in 200
ml of
0.1 M NaHCO 3 to give a concentration of 0.5% w/v. Solutions were refrigerated,
then
used within a few days or frozen. After correcting for losses, we estimate this crude GA
constituted
2.5% of the weight of the original dry leaves. The product was mostly GA because our yield was
comparable to yields of 1.30–1.56% obtained with more complex procedures that likely
resulted in additional losses but gave essentially pure products (Kurihara, 1969; Riskey et al.,
1982). Thus, ~2.5% of dry leaves of Gymnema sylvestre is GA, a useful
figure when
comparing potencies of extracts with purified GA.
Psychophysical method
Subjects rated the intensity of the stimuli on a 10-point scale, ranging from 0 (tasteless) to 9 (extremely strong), using a `sip-and-spit' procedure with water rinses between trials. Stimuli were presented for rating before subjects were asked to rinse with GA (two 1 min rinses with 5 ml of GA, followed by a water rinse) and again immediately after and 5, 10, 20 and 30 min after the GA rinse.
Data analysis
The effect of GA on perceived taste intensity over time was examined using a two-way, repeated measures ANOVA. The two factors of interest were compound (NaCl, sucrose and aspartame) and time (six intensity rating points from before to 30 min after the GA rinse). Final post-treatment ratings were compared with pre-treatment ratings using t-tests. To establish any difference in the effect of GA on the two sweeteners, sucrose and aspartame, the five post-GA ratings were compared with the pre-GA ratings by a two-way, repeated measures ANOVA. The ratings for water had no variance for many of the times it was rated and were therefore excluded from the analysis.
Results and discussion
Rinsing with GA had an immediate effect on the intensity of sucrose and aspartame, but no effect on water or NaCl (Figure 1). The repeated measures tests of within-subjects effects were significant for compound [F(2,10) = 21.96, P< 0.0001], for time [F(5,25) = 24.01, P< 0.0001] and for their interaction [F(10,50) = 10.86, P< 0.0001]. The intensity of sucrose and aspartame was significantly reduced immediately following GA: the mean intensity rating for sucrose fell from 6.7 ± 0.4 to 0.7 ± 0.3 (t = 10.39, P< 0.0001); and that of aspartame fell from 6.7 ± 0.5 to 1.2 ± 0.5 (t = 5.74, P< 0.002). Recovery of intensity showed significant linear trends for both compounds [sucrose: F(1,5) = 52.23, P< 0.001; aspartame: F(1,5) = 14.55, P< 0.012], but neither returned to its pre-treatment level by 30 min after treatment. At time 30 min the rating of sucrose was 4.2 ± 0.3 (t = 5.84, P< 0.002) and the rating of aspartame was 4.3 ± 0.6 (t = 4.18, P< 0.009). The proportional effect of GA on pre-GA taste intensity ratings for the two sweeteners did not differ; neither the main effect of stimulus nor the stimulus by time interaction was significant.
|
Our results are consistent with previous work attributing to GA a selective and profound effect on sweet but not nonsweet stimuli, yet a general and equal effect on most if not all sweet stimuli irrespective of chemical structure (Riskey et al., 1982;
Frank et al., 1992). The time
course of recovery that we observed also compares well with earlier data obtained with a 0.05%
GA
solution (Riskey et al., 1982); the intensity of sucrose (average
for 0.08, 0.16, 0.32 and 0.64
M) rose to ~60% of the pre-GA value in 40 min. |
Experiment 2. Effect of GA rinse on identification |
|---|
|
TopAbstractIntroductionExperiment 1. Taste perception...Experiment 2. Effect of...Experiment 3. Effect of...General discussion and...References |
|---|
Using a within-subjects experimental design, experiment 2 compared effects of GA on identification of sweet and nonsweet stimuli at 0–10 min versus 20–30 min following GA application.
Subjects
Ten subjects participated (three women and seven men), aged 17–38 (mean = 23, SD = 6 years).
Stimuli and treatment rinse
The 10 stimuli (and their names) were identical to those used in our other study (Hettinger et al., 1996), and included five solutions that were primarily or
partly sweet: 0.3 M sucrose
(`sugar'), 3 mM aspartame (`artificial sweetener'), and
NaCl–sucrose (`salt–sugar'), citric acid–sucrose
(`acid–sugar') and quinine–sucrose
(`quinine–sugar')
mixtures; and five that were nonsweet: 0.1 M NaCl (`salt'), 0.1 M KCl
(`salt
substitute'), 0.1 M Na glutamate (`MSG'), 0.1 mM quinine•HCl
(`quinine') and 3 mM citric acid (`acid'). The concentrations used
to
generate components of mixtures were the same as those used in single stimuli. Sucrose, NaCl
and KCl
were all reagent grade compounds from Baker; citric acid (reagent grade), quinine•HCl and
MSG were all from Sigma (St Louis, MO) labeled for laboratory use.
The GA rinse was prepared in the manner described for experiment 1, resulting in a 0.5% GA solution.
Psychophysical method
As in our other study (Hettinger et al., 1999), 100 trials were
presented in 10
replicates of 10 stimuli; within each replicate, all 10 stimuli were presented in random order.
Using the
`sip-and-spit' method, with water rinses between trials, subjects were asked to
taste 5
ml of solution, then name the solution using only names on the list provided. Subjects received
no
feedback as to the correct label for each of the stimuli. One hour or less was needed to complete
each
of two sessions; each replicate took 5–6 min to run. In addition, subjects received a 2 min
treatment rinse before replicates 1 and 6 (trials 1 and 51). The rinse was water in session 1 and
0.5%
GA in session 2.
Data analysis
A matrix containing frequencies of each response for all stimuli was generated for each
subject—for examples see Table 1 and our other study (Hettinger et al., 1999). Fromeach
subject's TCM we derived the percent correct identification for each
stimulus. Using
percent correct for the sweet stimuli from session 2, in which GA was applied, we used a t-test to examine the effects of time after GA rinse. Because the effect of GA fades with time
(experiment 1), performance should be reduced more for replicates closest to GA rinse
(replicates 1, 2,
6 and 7) than for replicates farthest from the rinse (replicates 4, 5, 9, 10). The effect of the GA
rinse
was examined separately for the sweet and nonsweet stimuli with a two-way analysis of variance
(ANOVA) with two main effects: session (1: water rinse, 2: GA rinse) and stimulus. Separate
analyses
were necessary because GA selectively affects sweet intensity (Riskey et al.,
1982), and
identification performance with the nonsweet stimuli improves with practice (Hettinger et al., 1999). Given the result of the t-test described
above, we also repeated the two-way
ANOVA
using data for sweet stimuli from replicates 1, 2, 6 and 7.
|
Results and discussion
In session 2, the sweet stimuli were identified less accurately in the four replicates immediately following the GA rinse (36 ± 5% correct) compared with the four replicates further from the GA rinse (59 ± 6% correct) (t = 3.36, P< 0.008). Patterns of errors for `near' replicates (1, 2, 6 and 7) were compared with patterns of errors for `far' replicates (4, 5, 9 and 10). On average, the three mixtures were more frequently misidentified as a nonsweet mixture component in the `near' (34.2%) than `far' replicates (12.5%) and sucrose was more frequently misidentified as a mixture or nonsweet stimulus in the `near' (20.0%) than `far' replicates (7.5%).
The ANOVA comparing percent correct for all 10 replicates of the sweet stimuli in
water
and GA sessions did not reveal a significant effect of rinse (P< 0.062), even though
average
percent correct fell from 59 ± 4% in session 1 (water) to 50 ± 3% in session 2
(GA).
However, rinse condition had a significant effect on performance in the four `near'
replicates: 1, 2, 6 and 7 [F(1,9) = 17.39, P< 0.002], with percent correct
falling
from 59% in session 1 (water) to 36% in session 2 (GA) (Figure 2).
Collapsed across sessions, the
accuracy with which the five different sweet stimuli were correctly identified in all 10 replicates
differed
[F(4,36) = 3.22, P< 0.023], as expected, given that normal individuals
identify
aspartame less accurately than the other sweet stimuli (Hettinger et al., 1999). More to the
point, there was a significant interaction of rinse condition and sweet stimulus both for all 10
replicates [F(4,36) = 2.793, P< 0.041] and for the four `near'
replicates: 1, 2, 6
and 7 [F(4,36) = 2.787, P< 0.041]. Post-hoc contrasts (
= 0.01)
demonstrated significant decreases in percent correct in session 2 (GA) compared with session 1
(water). For sucrose, percent correct fell from 69.6 ± 6% to 40 ± 10% in all 10
replicates, and to 28 ± 11% in replicates 1, 2, 6 and 7; for the quinine– sucrose
mixture
it fell from 62 ± 10% to 30 ± 6% in replicates 1, 2, 6 and 7.
|
The ANOVA for the nonsweet compounds showed no significant effect of rinse or rinse by stimulus interaction. There was a significant effect of stimulus [F(4,36) = 8.78, P< 0.0001], confirming the differences in correct identification for the stimuli shown in our other report (Hettinger et al., 1999
). In summary, the profound decrease in the intensity of sweet stimuli immediately following application of a 0.5% solution of GA (experiment 1) results in decreased accuracy in identifying sweet stimuli. Accuracy of identification improves as the effect of GA wears off with time. In fact, for the replicates farthest from the GA rinse (4, 5, 9 and 10), identification performance in session 2 (GA) was the same as in session 1 (water) at 59%.
In experiment 1 we showed that the perceived taste intensity of either sucrose or
aspartame,
which have a sweet taste quality, was reduced to ~ 15% of pre-rinse intensity by a GA rinse but
recovered linearly with time. Intensity recovered another 15% of its pre-rinse value every 10 min.
Given
that each session took about 60 min to run, in the 20 min spent running replicates 1–3 or
6–8, intensity would recover to ~50% of its value before GA. Apparently, subjects could
identify stimuli at half strength as accurately as stimuli at full strength. In our other report, we
addressed
effects of stimulus intensity on identification (Hettinger et al., 1999). Those data and the current
finding support the idea that identification improves as intensity increases until a peak is reached,
beyond which further increase in stimulus intensity may result in decreased accuracy.
|
Experiment 3. Effect of multiple GA rinses on identification |
|---|
|
TopAbstractIntroductionExperiment 1. Taste perception...Experiment 2. Effect of...Experiment 3. Effect of...General discussion and...References |
|---|
In order to maintain its maximal effect on sweet stimulus intensity, subjects in experiment 3 rinsed with GA frequently during the course of a session. A between-subjects experimental design was used to assess the consequent decrease in sweetness on the identification of sweet and nonsweet stimuli. The measures of stimulus identification used in the analyses were percent correct and measures derived from information theory that quantify consistency and discriminability. We discuss the results of experiment 3 in the context of the oft-cited confusions in taste labels made by normal people (Kuznicki et al., 1983;
Cowart et al., 1997; Ossebaardet al., 1997), and in the face
of evidence that practice and training improve accuracy of tastestimulus identification (Hettinger et al., 1999). Subjects
Twenty-four subjects participated in experiment 3. Twelve served in a control (water rinse) group (six women and six men), aged 22–52 (mean = 38, SD = 12 years), and 12 served in the GA-rinse group (eight women and four men), aged 20–64 (mean = 36, SD = 13 years).
Stimuli and treatment rinse
Stimuli and their names were the same as those used in experiment 2. The GA rinse was prepared in the manner described in experiment 1 and 2, resulting in a 0.5% GA solution.
Psychophysical method
Subjects were asked to choose correct stimulus names for the 10 test stimuli, again presented 10 times each in a single session without feedback, as in experiment 2. Each subject received a 2 min rinse every two replicates (i.e. prior to replicates 1, 3, 5, 7 and 9), instead of every five replicates as in experiment 2. Twelve subjects (control group) rinsed with water and another 12 (GA group) rinsed with 0.5% GA.
Data analysis
The resulting TCMs were used to generate measures of performance based on information
theory. Two measures of T, bits of information transferred, were calculated as described
elsewhere (Hettinger et al., 1999). The first was an overall
measure of consistency of
performance using the full 10 x 10 response matrix: T 10. The second
was T 2, a measure of pairwise stimulus discriminability using the forty-five
2
x
10 response matrices. T 2 quantifies the difference in the response patterns
for
each pair of stimuli. In this analysis, T = Hx + Hy – Hxy bits of information, where Hx = 
[px
x log2(1/px)], Hy =[py x log 2(1/py)] and Hxy =[pxy x log2(1/pxy)]. Hx is the
information contained in the 10 stimuli, Hy is the information in
the 10 response labels and Hxy is the
information in the 100 stimulus– response combinations. The probabilities of
stimulus presentations (px) were set at 0.1, i.e. 10 out of 100;
probabilities of response–label use (py) and of each
stimulus– response combination (pxy) were determined by
a
subject's performance on the identification task (seeTable 1 for an
example). Theoretically, T 10 can range from 0 to 3.32 bits; likewise, T 2 can range
from 0 to 1.00 bit. T is maximal when the 10 response labels are used with equal
frequency
and each response label is used for one stimulus only (e.g. for T 10, Hx = 3.32, Hy = 3.32 and Hxy = 3.32). T is minimal when each response label is used for every
stimulus (e.g. for
T 10, Hx = 3.32, Hy = 3.32
and Hxy = 6.64) and all stimulus–response combination
frequencies are 1. Close to maximum T (perfect performance) has been achieved by
some
subjects, but minimum T is very unlikely with merely 10 replicates of random
assignment. For
further discussion of these topics, see our other study (Hettinger et al., 1999).
Subjects' taste confusion matrices were also used to generate the measures of performance used in experiment 2: percent correct and error patterns for each sweet stimulus.
We tested with a one-tailed, independent t-test whether the overall
consistency of
response, as measured by T 10, was greater for the water-rinse than the
GA-rinse
group. We examined with three separate two-way ANOVAs the effect of GA on T2 (Hettinger et al., 1996). The three ANOVAs were for
(1)
the 25 stimulus pairs
composed of one primarily or partly sweet and one nonsweet stimulus; (2) the 10 stimulus pairs
composed of two sweet stimuli; and (3) the 10 stimulus pairs composed of two nonsweet stimuli.
The
main variables were rinse group and stimulus pair.
Analysis of percent correct was similar to that described for experiment 2. Separate,
repeated measures ANOVAs were performed for the sweet and nonsweet stimuli. The
between-subjects factor was rinse (water or GA); the withinsubjects repeated measure was
stimulus.
Percent correct was calculated in two ways: (1) for 10 `correct labels' as in
experiment
2 and (2) with `correct' defined in terms of five response categories. Using
response
categories based on errors made by normal subjects (Hettinger et al., 1996) minimized
between-group differences due to unfamiliar stimuli or stimulus names. The five categories were
created
as follows: (1) a response of `salt', `salt substitute' or
`MSG' was treated as `correct' for salts: NaCl, KCl, and MSG; (2)
either `sugar' or `artificial sweetener' was treated as correct for
sucrose
and aspartame; (3) either `quinine' or `acid' was treated as correct
for
quinine•HCl and citric acid; (4) either `acid–sugar' or
quinine–sugar' was treated as correct for the acid–sucrose and
quinine–sucrose mixture; and (5) `salt– sugar' remained the only
correct
label for the NaCl–sucrose mixture.
Percent correct calculated on the basis of the five response categories was used for additional analyses that examined the effect of GA rinse on the pattern of response errors for the three sucrose mixtures. A two-way ANOVA with multiple comparisons was run with rinse condition (either GA or water) as the between-subjects factor and response label (limited to three categories: the mixture label and the labels for each of the components) as the within-subjects factor.
Results and discussion
The use of frequent rinses maintained the efficacy of GA throughout the test session. Average percent-correct identification of the five sweet stimuli was 36 ± 5% both for replicates 1, 3, 5, 7 and 9, which immediately followed a GA rinse, and for replicates 2, 4, 6, 8 and 10, which immediately preceded the next GA rinse. Thus, we succeeded in maintaining a maximal effect of GA on stimulus identification.
Overall consistency of performance as measured by T 10, which has a
maximum value 3.32 bits, was lower for the GA-rinse group (mean = 1.84 ± 0.07 bits)
than for
the waterrinse group (mean = 2.08 ± 0.10 bits) (t = 2.04, P< 0.026).
The
percent correct for all 10 stimuli (sweet and nonsweet stimuli combined) was 45 ± 3%
for both
the GA-rinse and water-rinse groups. Thus consistency for the entire 10 x 10 matrix was
more
sensitive to the effect of GA than percent correct, a result that is explained by a consistent but
incorrect
use of labels by some subjects (Hettinger et al., 1996). T 10 and average
percent correct (10 response labels) were correlated for all 24 subjects: r = 0.77
(significantly
greater than 0, P< 0.0001). A similar significant correlation between T10
and percent correct was also observed for 42 subjects (Hettinger et al., 1999). As was
pointed out in that study, a high level of percent correct necessarily yields high values of T10; however, low values of percent correct may not yield low values of T 10
if subjects use `incorrect' labels consistently.
Repeated measures ANOVAs of the pairwise stimulus discriminability, as measured by T 2, showed the ability of the GA-rinse group to discriminate the 25 pairs
representing
one nonsweet and one sweet stimulus to be inferior to the ability of the water-rinse group. The
average
value of T 2 of 0.79 ± 0.02 bit for the GA-rinse group was lower [F(1,22) = 25.36, P< 0.0001] than the average T 2
of 0.93
± 0.02 bit
for the water-rinse group. The value for the water-rinse group is close to the theoretical
maximum 1.0
bit and is identical to the performance for sweet–nonsweet pairs observed previously (Hettinger et al., 1999). The reduced intensity of the sweeteners
presumably made it more
difficult to
distinguish them consistently from nonsweet stimuli.
The significant interaction term for rinse and stimulus pair [F(24,528) = 5.06, P< 0.0001] suggests a variable effect of GA on T 2, the nature of
which
we
addressed with additional analyses. Table 2 shows results of analyses for
the 25
nonsweet–sweet pairs grouped as follows: the 10 pairs involving a nonsweet stimulus and
sucrose or aspartame and the three sets of five pairs involving a nonsweet stimulus and a sucrose
mixture. For each of these sets of stimulus pairs, the water-rinse group performed better than the
GA-rinse group. Figure 3 shows the significant post-hoc comparisons
( = 0.01), all of which
involve discriminations of the sweet mixtures with a nonsweet stimulus. An analysis of the exact
errors
that the subjects in the GA group made follows after our presentation of percent-correct measures
below.
|
|
Rinse condition had no overall effect on the 10 pairs representing the discrimination of two sweet stimuli. However, there was a significant interaction of rinse and stimulus pair [F(9,198) = 3.09, P< 0.002]. The GA-rinse group was significantly better at discriminating aspartame from the salt–sucrose mixture (T 2 = 0.97 ± 0.03 bit) than the waterrinse group (T 2 = 0.65 ± 0.09 bit) (post-hoc comparisons,
= 0.01). Table 1 is the matrix of an individual in the GA group who
discriminated this pair of stimuli
perfectly (T 2 = 1.0); there was no overlap in responses. However, subjects
who
rinsed with water identified the NaCl–sucrose mixture as `sugar' or
`artificial sweetener' 21% of the time (Hettinger et al., 1999). Such mistakes
would be expected to be less frequent, given the reduction in sweetness of sucrose in this
mixture, and
would improve discriminability. Finally, rinse had no effect and there was no rinse by pair interaction for the 10 pairs representing the discrimination of two nonsweet stimuli.
The results of the repeated measures ANOVAs of the percent-correct responses calculated using the original 10 response labels are shown in Table 3. Rinse condition (group) failed to reveal a significant effect on correct identification of the sweet solutions; although averages were 50 ± 5% correct for the water-rinse group and 36 ± 5% for the GA-rinse group. However, the rinse condition by compound interaction was significant and, as in experiment 2, the greatest effect was for identification of sucrose. The water-rinse group identified sucrose correctly 67 ± 8% of the time but the GA group identified it correctly 32 ± 8% of the time.
|
The results for the sweet stimuli were more straightforward when percent correct was calculated from the five response–label combinations (Table 3.). The average 55 ± 6% correct performance for sweet stimuli by the GA group was significantly lower than the average 76 ± 3% correct performance by the water control group. There was no rinse by stimulus interaction.
Although there was no between-subjects group effect of rinse condition for the nonsweet
stimuli for either the 10or the 5-label analysis, there was a significant group by compound
interaction
with the 10-label analysis (Table 3.). This result may be explained by a
fortuitous assignment of subjects:
the water-rinse group correctly identified quinineHCl and citric acid less frequently than the
GA-rinse
group (44 ± 11 versus 71 ± 10% and 44 ± 11 versus 64 ± 9%,
respectively) . This disparity in percent-correct performance, attributable to the taste-quality label
confusions made by normal people (Kuznicki et al., 1983; Cowart et al., 1997;
Ossebaard et al., 1997), necessarily disappeared when the
response labels for quinine and
acid were combined. Similarly, the significant group by compound interaction for the 10-label
analysis
for the sweet stimuli is partly attributable to the same fortuitous group difference, specifically
with regard
to correct identification of the quinine–sucrose and acid– sucrose mixtures. In this
case,
when labels were combined, the effect of GA on group performance was revealed.
Figure 4 shows the effect of GA rinse on the pattern of five response–label combinations used for each sweet stimulus. For the analysis of the sucrose mixtures, we were interested in the number of times out of the 10 stimulus presentations that each subject used the `correct' mixture label compared with one of the two mixture-component labels. Label had a significant effect [F(6,17) = 10.90, P< 0.0001], which indicated that subjects applied different labels to the different mixtures. More relevant to the effects of GA, the label by rinse-group interaction was also significant [F(6,17) = 5.42, P< 0.003]. Contrasts for interactions of labels and rinse group were significant for all three mixtures: NaCl–sucrose [F(1,22) = 11.42, P< 0.003]; acid–sucrose [F(1,22) = 11.04, P< 0.003]; and quinine–sucrose [F(1,22) = 6.29, P< 0.02] (Figure 4a–c). In each case, the label for the nonsweet component was used more frequently and the mixture label less frequently by the GA group compared with the control group. The patterns for sucrose and aspartame are also shown (Figure 4d,e). Sucrose was correctly identified significantly less often by the GA group than by the water group (t = 3.09, P< 0.005) and the results for aspartame showed a similar trend (t = 1.56, P< 0.07).
|
This analysis points to the similarity in results from experiments 2 and 3. Given the GA rinse, subjects more frequently mistook the nonsweet component of the sucrose-containing mixture for the mixture (experiment 3, Figure 4), which resulted in a reduced percent correct (experiment 2, Figure 2). The subjects misidentified sucrose and aspartame more frequently in the GA-rinse group, but in these cases responses were distributed across many incorrect label categories (Figure 4d,e). Finally, stimulus-paFir discriminability, as measured by T 2, was poorer for mixture versus nonsweet component pairs in the GA group (Figure 3). This effect can be seen in the individual matrix for a member of the GA group presented in Table 1. For this subject, T 2 calculated for the `citric acid' versus `acid–sucrose' rows reflects 80% response overlap. In the average matrix for 42 people who were not treated with GA, 12% of the responses in these rows overlapped (Hettinger et al., 1999
).
Measures of T 2 for the NaCl–sucrose versus nonsweets
showed that subjects selected the label for KCl as well as NaCl instead of the mixture label. This
is
understandable because use of response labels `salt' and `salt
substitute'
for KCl and NaCl overlapped 39% in subjects not treated with GA (Hettinger et
al., 1999).
Finally, comparison of Figure 4a with 4e further
demonstrates why T 2 for the
NaCl– sucrose mixture versus aspartame pair was larger for the GAthan the water-rinse
group
(also see above): mistaking the NaCl–sucrose mixture for aspartame was less likely when
the
perceptual intensity of sucrose was weak.
|
General discussion and conclusions |
|---|
|
TopAbstractIntroductionExperiment 1. Taste perception...Experiment 2. Effect of...Experiment 3. Effect of...General discussion and...References |
|---|
In this general discussion we address the utility of an objective, performance-based, stimulus-identification task, which results in a response–frequency matrix (TCM), in the diagnosis of a specific taste deficit. We comment on prevalence of dissociated taste losses in patients, the relationship between taste-stimulus identification performance and subjective intensity ratings, and specific misidentification patterns for binary stimuli made by individuals experiencing specific taste deficits. We also address the value of performance measures based on `correct identification' versus measures based on information theory (consistency of identification or discrimination).
Dissociated loss of the sweet taste in humans
Rinsing with GA produced a deficit specific for sweet compounds in normal subjects. This
simulates the nearly complete loss of sweet taste, but not other taste qualities, reported for a
Japanese
population (Tomita and Horikawa, 1986). However, such
specific sweet-taste losses are rare
among the patients of our Taste and Smell Clinic at the University of Connecticut Health Center.
During
the 24 month period of 09/96 to 09/98, we identified 88 patients (38 male, mean age 53.2
±
13.4 years, and 50 female, mean age 53.5 ± 16.4 years) with hypogeusia/ageusia (Gent et
al., 1997) for at least one of four prototypic taste stimuli: sucrose, NaCl, citric
acid and
quinine•HCl. Of these patients, 45 were hypogeusic/ageusic for sucrose, rating sucrose
intensity
(totaled for five concentrations: 0.01. 0.03, 0.10, 0.30 and 1.0 M), on average, half as strong as
did a
control population. However, only two of these patients experienced a
`dissociated' loss
for sweet sucrose, rating nonsweet NaCl, citric acid and quinine•HCl in the normal range of
values. Thus, in our patients, a sweet taste loss is rarely `dissociated;' rather, it is
usually
accompanied by losses in other taste modalities.
Correct identification is related to perceived intensity in a complex way
Correct identification of sweet-tasting stimuli was reduced by GA but the effect
disappeared
once stimulus intensity had recovered to 50% of pre-GA intensity. This information is useful for
the
design of a TCM that would identify hypogeusic patients. Data from a psychophysical function
for
sucrose (magnitude estimation of intensity versus concentration) produced by normal controls (n = 159) reveal that the intensity of 0.1 M sucrose is rated half as strong (14.20
± 0.60) as
0.3 M sucrose (28.16 ± 0.84) (Bartoshuk and Marks, 1986). Our
results on the recovery of
identification performance with time following GA rinses suggest that 0.1 M sucrose would be as
readily identified as 0.3 M sucrose. However, weaker sensations, such as those produced by
lower
concentrations in normals, would yield more misidentifications. Given that, on average, our
hypogeusic
patients perceive sucrose as half as intense as normals do, they would generally not
make more errors than normals if 0.3 M sucrose were used in a TCM. However, they would be
expected to make more errors than normals if 0.1 M sucrose were in the TCM. This idea could be
tested in the TCM on normal subjects by using 0.04 M sucrose, which normals perceived as half
as
intense as 0.1 M sucrose.
When the sucrose concentration increased from 0.3 to 0.9 M, more rather than fewer
misidentifications of sweet stimuli resulted (Hettinger et al., 1999). In particular,
sucrose-containing mixtures were mistaken for sucrose more frequently. Although 0.9 M sucrose
results in a 30% increase over 0.3 M in perceived intensity in normals, discriminability as
measured by T 2 for sweet–sweet pairs was reduced. Thus,
discriminability of
sweet–sweet pairs worsens when sucrose intensity is increased above midrange, whereas
discriminability of sweet–nonsweet pairs worsens when sucrose intensity is decreased
below
midrange. Thus, discriminability and stimulus intensity are related in a complex way and,
compared with
midrange intensities (e.g. 0.1–0.3 M sucrose), distinct patterns of errors are seen for
weaker
and stronger intensities, as may be experienced by patients.
Identification of sucrose-containing mixtures was most sensitive to sweet deficit
Subjects treated with GA had particular difficulty correctly identifying the sucrose mixtures
as mixtures, labeling them instead as the nonsweet components. It is interesting to compare this
result
with our observations of 42 subjects who performed the same identification task after water
rinses
(Hettinger et al., 1999). These subjects mistook components for
the mixture 23% of the time,
on average, but only 8% of the choices were for the nonsweet component; 15% were for the
sweet
component. In the current studies, the weaker `sweetness' of the sucrose
component
after GA resulted in an increase in mistakes of identifying the mixture as its nonsweet
component.
Among the subjects from our other study (Hettinger et al., 1999)
who had water rinses, the
0.3 M sucrose was hardly ever mistaken for a nonsweet stimulus. These TCM results are
compatible
with the perspective that taste is an analytic system and that each of the mixture components is
separately identifiable within the mixture (Bartoshuk and Gent, 1985; Schiffman and Erickson, 1993).
Those arguing that taste is synthetic point out that the subject is not presented with a response
label
appropriate for the mixture in most protocols. In the identification protocol, however, both
mixture and
component labels are presented.
Subjects who rinsed with water (Hettinger et al., 1999)
hardly ever mistook 0.3 M
sucrose or 3 mM aspartame for a nonsweet stimulus. Thus, if GA were to reduce sweetness
intensity to
zero, there would be no appropriate response such as `no taste' for sucrose and
aspartame. For these `pure sweet' stimuli, erroneous choices were thus distributed
across the incorrect categories following the GA rinse.
Data from a study on the effect of stimulus intensity on TCM (Hettinger et
al., 1999) suggest that the off-taste of aspartame (DuBois et al., 1991) resulted in
mistaking 3 mM
aspartame (the same concentration used in the present study) for a sucrose mixture 7% of the
time and
mistaking the more intensely sweet 20 mM aspartame for a sucrose mixture 30% of the time.
Rinsing
with GA, which reduced the intensity of the 3mM aspartame in this study, decreased the
frequency of
mistaking aspartame for the NaCl–sucrose mixture, which increased discriminability (T 2). In conclusion, although responses to sucrose mixtures are sensitive to a
taste deficit
produced by GA, some response errors become more likely, such as identifying a
salt–sugar
mixture as a salt, but others become less likely, such as identifying the mixture as a
`pure'
sweet stimulus.
The TCM measures based on information transmitted (T) fared well
compared with
simple percent-correct measures but both benefited by considering normal error patterns.
Measures
from information theory were used to quantify consistency in identification and pairwise
discrimination
of stimuli. Consistency in general and discrimination of sweet from nonsweet stimuli in
particular were
significantly impaired in subjects treated with GA. T measures consistency of
identification, not
correct identification. For example, if a subject consistently identified quinine as
`acid'
and acid as `quinine', information would be transmitted perfectly about the tastes
of the
stimuli by the gustatory system. T has an advantage over correct identification when a
subject
is unfamiliar with names of stimuli, especially if the `wrong' name has been
learned for a
substance. This advantage was observed in the current study, with T 10
showing a
significant effect of GA on the entire 10 x 10 TCM. T 10
`ignored' differences in use of incorrect names for nonsweets by the two groups.
The
discriminability of stimulus pairs, as measured by T 2, was also useful,
given
expected effects of GA on particular stimulus pairs. T 2 has the advantage
of
quantifying overlapping responses for pairs of stimuli; again consistency, not correct
identification, is
key. However, overall T 2 may show no effect; discriminability would be
expected
to improve for some pairs but decline for other pairs, given a particular taste weakness. The
`percent correct' measure can better detect deficits if response categories based on
common errors in normal populations are devised. Then naming errors that people frequently
make
becomes irrelevant to the result. Response categorization with an error analysis demonstrated the
effect
of GA very well. A deficit for the taste of one component of a binary, heterogeneous mixture (McBride,
1989) decreased the frequency of mistaking the mixture for that component but
increased the frequency
of mistaking it for the other component. We conclude that a dissociated sweet deficit would be
identified with greatest sensitivity by examining key changes in error patterns (more mistakes
toward
nonsweet and less toward sweet stimuli) and key changes in pairwise discriminability (decreases
for
sweet versus nonsweet, increases for sweet versus partially sweet). The key changes would be a
`signature' for this particular dissociated taste disorder, which would differ from
the
`signatures' for other dissociated deficits affecting other single taste modalities.
However,
given that 74% of our patients who have taste loss perceive stimuli of several taste qualities
weakly,
most patients would display multiple shifts in error patterns and discriminability.
Summary and conclusions
Treatment of the tongue with GA produced a sweet-taste intensity deficit in subjects who were then asked to identify 10 chemicals, each presented 10 times. Consistent sweetstimulus identification, as measured by information transmitted (T), was specifically impaired, suggesting that `objective' tests could serve to characterize taste deficits in patients.
|
Acknowledgments |
|---|
This work was supported by NIH grant P50 DC00168 and a University of Connecticut Health Center Summer Fellowship to Quinterol Mallette. We also thank Amanda Hesla of the University of Nebraska for technical assistance and Dr Robert A. Frank of the University of Cincinnati for sharing his supply of gymnema leaves.
|
References |
|---|
|
TopAbstractIntroductionExperiment 1. Taste perception...Experiment 2. Effect of...Experiment 3. Effect of...General discussion and...References |
|---|
Bartoshuk, L.M. and Gent, J.F. (1985) Taste mixtures: an analysis of synthesis. In Pfaff, D.W. (ed.), Taste, Olfaction and the Central Nervous System. Rockefeller University Press, New York, pp. 210–232.
Bartoshuk, L.M. and Marks, L.M. (1986) Ratio scaling. In Meiselman, H.L. and Rivlin, R.S. (eds), Clinical Measurement of Taste and Smell. Macmillan Publishing Co., New York, pp. 50–65.
Bartoshuk, L.M., Dateo, G.P., Vandenbelt, D.J., Buttrick, R.L. and Long, L. Jr(1969) Effects of Gymnema sylvestre and Synsapalum dulcificum on taste in man. In Pfaffmann, C. (ed.), Olfaction and Taste III. Rockefeller University Press, New York, pp. 436–444.
Cain, W.S. and Gent, J.F. (1986) Use of odor identification in clinical testing of olfaction. In Meiselman, H.L. and Rivlin, R.S. (eds), Clinical Measurement of Taste and Smell. Macmillan, New York, pp. 170–186.
Cowart, B.J. (1989) Relationship between taste and smell across the adult life span. In Murphy, C., Cain, W.S. and Hegsted, D.M. (eds), Nutrition and the Chemical Senses and Aging. New York Academy of Sciences, New York, 561, pp. 39–55.
Cowart, B.J., Young, I.M., Feldman, R.S. and Lowry, L.D. (1997) Clinical disorders of smell and taste. Occupat. Med., 12: 465–483.
Deems, D.A., Doty, R.L., Settle, R.G., Moore-Gillon, V., Shaman, P., Mester, A.F., Kimmelman, C.P., Brightman, V.J. and Snow, J.B. Jr (1991) Smell and taste disorders, a study of 750 patients from the University of Pennsylvania Smell and Taste Center. Arch. Otolaryngol. Head Neck Surg., 117, 519–528.
Doty, R.L., Shaman, P. and Dann, M. (1984) Development of the University of Pennsylvania Smell Identification Test: a standardized microencapsulated test of olfactory function.Physiol. Behav. , 32, 489–502.[Medline]
DuBois, G.E., Walters, D.E., Schiffman, S.S., Warwick, Z.S., Booth, B.J., Pecore, S.D., Gibes, K., Carr, B.T. and Brands, L.M. (1991) Concentration–response relationships of sweeteners: a systematic study. In Walters, D.E. Orthoefer, F.T. and DuBois, G.E. (eds), Sweeteners: Discovery, Molecular Design and Chemoreception. American Chemistry Society, Washington, DC, pp. 261–276.
Frank, M.E., Hettinger, T.P. and Clive, J.M. (1995) Current trends in measuring taste. In Doty, R.L. (ed.), Handbook of Olfaction and Gustation. Marcel Dekker, Inc., New York, pp. 669–688.
Frank, R.A., Mize, S.J., Kennedy, L.M., de los Santos, H.C. and
Green, S.J.(1992) The effect of Gymnema sylvestre extracts on the
sweetness of eight sweeteners. Chem. Senses, 17, 461–479.
Gent, J.F., Frank, M.E. and Mott, A.E. (1997) Taste testing in clinical practice. In Seiden, A.M. (ed.), Taste and Smell Disorders. Thieme, New York, pp. 146–158.
Hellekant, G. and van der Wel, H. (1989) Taste modifiers and sweet proteins. In Cagan, R. (ed.), Neural Mechanisms in Taste. CRC Press, Boca Raton, FL, pp. 85–96.
Hettinger, T.P., Frank, M.E. and Myers, W.E. (1996) Are the tastes of Polycose
and monosodium glutamate unique? Chem. Senses, 21, 341–347.
Hettinger, T.P., Gent, J.F., Marks, L.E. and Frank, M.E. (1999) A confusion matrix for the study of taste perception. Percept. Psychophys., 61, in press.
Kurihara, Y. (1969) Antisweet activity of gymnemic acid A1 and its derivatives. Life Sci.,8 , 537–543.[Web of Science][Medline]
Kuznicki, J.T., Hayward, M. and Schultz, J. (1983) Perceptual processing of
taste quality. Chem. Senses, 7, 273–292.
McBride, R.L. (1989) Three models for taste mixtures. In Laing, D.G., Cain, W.S., McBride R.L. and Ache, B.W. (eds.), Perception of Complex Smells and Tastes. Academic Press, San Diego, CA, pp. 265–282.
Oakley, B. (1985) Taste responses of human chorda tympani nerve. Chem. Senses, 10, 469–482.
Ossebaard, C.A., Polet, I.A. and Smith, D.V. (1997) Amiloride effects on taste quality: comparison of single and multiple response category procedures. Chem. Senses, 22,267 –275.
Riskey, D.R., Desor, J.A. and Vellucci, D. (1982) Effects of gymnemic acid
concentration and time since exposure on intensity of simple tastes: a test of the biphasic model
for the
action of gymnemic acid. Chem. Senses, 7: 143–151.
Schiffman, S.S. and Erickson, R.P. (1993) Psychophysics: insights into transduction mechanisms and neural coding. In Simon, S.A. and Roper, S.D. (eds), Mechanisms of Taste Transduction. CRC Press, Boca Raton, FL, pp. 395–424.
Smith, D.V. and van der Klaauw, N.J. (1995) The perception of saltiness is
eliminated by NaCl adaptation: implications for gustatory transduction and coding. Chem. Senses,20
, 545–557.
Tomita, H. and Horikawa, Y. (1986) Dissociated taste disorder.AurisNasus-Larynx (Tokyo) , 13(suppl. I), s17–s23.
Warren, R.M. and Pfaffmann, C. (1959) Suppression of sweet sensitivity by potassium gymnemate. J. Appl. Psychol., 14, 40–42.
Wright, H.N. (1987) Characterization of olfactory dysfunction. Arch. Otolaryngol.Head Neck Surg. , 113, 163–168.
Wright, H.N., Sheehe, P.R. and Leopold, D.A. (1991) The odorant confusion matrix as an aid to diagnosis. Chem. Senses, 16, 601.
Accepted March 31, 1999
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  M.-F. Wang, L. E. Marks, and M. E. Frank Taste Coding after Selective Inhibition by Chlorhexidine Chem Senses, October 1, 2009; 34(8): 653 - 666. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. Maehashi, M. Matano, M. Nonaka, S. Udaka, and Y. Yamamoto Riboflavin-Binding Protein Is a Novel Bitter Inhibitor Chem Senses, January 1, 2008; 33(1): 57 - 63. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. F. Gent, M. E. Frank, and T. P. Hettinger Taste Confusions Following Chlorhexidine Treatment Chem Senses, January 1, 2002; 27(1): 73 - 80. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. K. Formaker and M. E. Frank Taste Function in Patients with Oral Burning Chem Senses, October 1, 2000; 25(5): 575 - 581. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||