J. AMER. SOC. HORT. SCI. 126(1):115–121. 2001.

Sugars and Acids Influence Flavor Properties of

Mango (Mangifera indica)
T.M.M. Malundo
U. S. Distilled Products, 1607 South 12th Street, Princeton, MN 55371
R.L. Shewfelt
Department of Food Science, University of Georgia, Athens, GA 30602
G.O. Ware
Department of Experimental Statistics, College of Agricultural and Environmental Sciences, University
of Georgia, Athens GA 30602
E.A. Baldwin1
U.S. Department of Agriculture, Agricultural Research Service, Citrus and Subtropical Products
Laboratory,600 Ave. S N.W., P.O. Box 1909, Winter Haven, FL 33883-1909
ADDITIONAL INDEX WORDS. descriptive sensory panel, total soluble solids, titratable acidity
ABSTRACT. Information on important flavor components for fruit and vegetables is lacking and would be useful for
breeders and molecular biologists. Effects of sugar and acid levels on mango (Mangifera indica L.) flavor perception were
analyzed. Twelve treatments, identified using a constrained simplex lattice mixture design, were formulated by adding
sugar (60%), citric acid (40%), and water to an equal volume of mango homogenate. Using 150-mm nonstructured line
scales, a trained panel evaluated the treatments according to 11 flavor descriptors. Titratable acidity (TA), pH, and total
soluble solids (TSS) were also determined. Acid concentration affected ratings for sweet, sour, peachy, pine/terpentine,
astringent, and biting. Except for sour taste, all descriptors were affected by sugar content while increasing water
increased intensities of all flavor notes. TA, pH, and TSS/TA correlated (P < 0.01) with and were useful predictors (r >
0.80) of sour taste and chemical feeling descriptors astringent and biting. TSS, however, was not a particularly good
indicator of sweetness (r = 0.72) or any other descriptor except possibly peachy (r = 0.79). It is evident from this study that
sugars and acids enhance human perception of specific flavor notes in mango, including aromatics.

Mango (Mangifera indica L.) is a very popular specialty fruit and Thompson, 1985; Tandon and Kalra, 1983; Vazquez-Salinas
in the United States (Vance Publishing Corp., 1994) and is of and Lakshminarayana, 1985). The main reducing sugar identified
great economic importance to the tropical regions that produce by Medlicott and Thompson (1985) and Vazquez-Salinas and
them (Mitra and Baldwin, 1997; Narain et al., 1998). Its current Lakshminarayana (1985) was fructose while Selvaraj et al. (1989)
popularity is attributed mainly to its unique flavor, aroma, and reported glucose to be predominant. Conflicting reports on the
appearance. Flavor, especially, is a consumption attribute critical relative concentrations of individual sugars present in mango at
to consumer acceptability of mangoes (Gholap et al., 1986; different stages of ripening were attributed to varying cultivars
Malundo et al., 1996). and storage conditions used (Medlicott and Thompson, 1985).
Lack of attention to flavor compounds in breeding programs The effects of cultivar (Gowda and Ramanjaneya, 1994;
has led to flavor mediocrity in many fruit and vegetables. This has Kapse et al., 1988; Kundu and Gosh, 1992;), stage of maturity
occurred because fruit and vegetable breeders have little informa- (Morga et al., 1979; Shashirekha and Patwardhan, 1976; Tandon
tion on flavor compounds available for use in selecting for this and Kalra, 1983), postharvest treatments (Kumar et al., 1992),
complex trait. Some information on mango flavor has been and storage conditions (Vazquez-Salinas and Lakshminarayana,
reported, but much is still not understood. Organic acids and 1985) on sugar and acid levels in mango have been studied
sugars are key components in the perception of mango flavor as extensively. Acid levels are expressed frequently in terms of pH
in most fruit (Medlicott and Thompson, 1985). The predominant and titratable acidity (TA) while sugar concentrations are re-
acid is citric (Medlicott and Thompson, 1985; Lizada, 1993). On ported as total soluble solids content (TSS) (Kapse et al., 1988;
the other hand, as a result of starch hydrolysis from increased Kumar et al., 1992; Kundu and Gosh, 1992; Medlicott and
amylase activity during ripening (Fuchs et al., 1980; Tandon and Thompson, 1985; Morga et al., 1979; Vazquez-Salinas and
Kalra, 1983), sucrose is the major sugar in the ripe fruit (Medlicott Lakshminarayana, 1985). Few investigators, however, have at-
tempted to understand how varying concentrations of these
Received for publication 1 Nov. 1999. Accepted for publication 25 Aug. 2000.
components affect flavor perception of the fruit. This is important
South Atlantic Area, Agricultural Research Service, U.S. Department of Agricul- information for breeders and molecular biologists who seek to
ture. Mention of a trademark or proprietary product is for identification only and improve mango flavor through genetic manipulation using culti-
does not imply a guarantee or warranty of the products by the USDA. The USDA var selection or identification of genes responsible for flavor
prohibits discrimination in all its programs and activities on the basis of race, quality. Kapse et al. (1988) determined that increasing total
color, national origin, gender, religion, age, disability, political beliefs, sexual
orientation, and marital or family status. We thank Julia Heggie and Holly Sisson soluble solids while decreasing acidity increased flavor ratings of
for technical assistance. The cost of publishing this paper was defrayed in part by mango as determined by a panel of seven judges. However, the
the payment of page charges. Under postal regulations, this paper therefore must usefulness of such conclusions towards understanding the effects
be hereby marked advertisement solely to indicate this fact. of sugar and acids on mango flavor is limited since the investiga-
1
To whom correspondence should be addressed; e-mail ebaldwin@citrus.usda.gov.
tors failed to clearly define the type of panel (i.e., consumer,

J. AMER. SOC. HORT. SCI. 126(1):115–121. 2001. 115

 solution was prepared from a mixture of sucrose (table sugar) and
fructose (D-fructose, U.S.P./N.F., Spectrum Chemical Mfg. Corp.,
Gardena, Calif.) combined in a 2 sucrose : 1 fructose ratio, the
average ratio of sucrose to fructose in mango cultivars studied
(Gowda and Ramanjaneya, 1994; Kumar et al., 1992; Kundu and
Ghosh, 1992; Medlicott and Thompson, 1985; Tandon and Kalra,
1983; Vazquez-Salinas and Lakshminarayana, 1985). Before
blending with an equal volume of mango homogenate (see
below), mixtures of 60% sugar, 40% citric acid and deionized
water were prepared in specific proportions resulting in a three-
component constrained simplex lattice mixture design, as de-
scribed by Cornell (1983) (Fig. 1). A similar design was devel-
oped in studying the effects of formulation on sensory, physical,
and microbial properties (Malundo et al., 1994); and consumer
acceptability (Malundo and Resurreccion, 1993) of liquid coffee
whitener from peanut extract.
To approximate sugar and acid levels in mango cultivars
already studied (Gowda and Ramanjaneya, 1994; Kapse et al.,
1988; Kumar et al., 1992; Kundu and Ghosh, 1992; Medlicott and
Thompson, 1985; Morga et al., 1979; Tandon and Kalra, 1983;
Fig. 1. Constrained region in the simplex coordinate system defined by the Vazquez-Salinas and Lakshminarayana, 1985), the proportions
following restrictions: 0 < x1 < 0.60 and 0 < x2 < 0.08, where x1 and x2 are
proportions of sugar and acid solutions, respectively. Component x3 represents
of sugar and acid solutions used were limited to maximum levels
proportion of deionized water. Points within the constrained region indicate of 0.6 and 0.08, respectively.
design mixtures of a constrained simplex lattice experimental design. Points Twelve mixture blends were identified and corresponded to
correspond to the following (x1, x2, x3 ) coordinates : 1: (0.60, 0.08, 0.32), 2: the 12 points on the simplex coordinate system in Fig. 1. These
(0.60, 0.04, 0.36), 3: (0.60, 0, 0.40), 4: (0.40, 0.08, 0.52), 5: (0.40, 0.04, 0.56), consisted of coordinates of vertices and of uniformly spaced
6: (0.40, 0, 0.60), 7:(0.20, 0.08, 0.72), 8: (0.20, 0.04, 0.76), 9: (0.20, 0, 0.80) 10:
(0.0, 0.08, 0.92), 11: (0.0, 0.04, 0.96), and 12: (0, 0, 1). points on the face and sides of the constrained region of the
polynomial equation representing a response of behavior over the
experienced, or trained) or scale (i.e., intensity or acceptability) restricted simplex region.
used. Frozen mango homogenate was thawed by immersing plastic
This study was conducted to determine the effects of varying containers in tap water. For each treatment, 250 mL sugar/acid
sugar and acid concentrations on flavor properties of mango to mixture was prepared by combining water, and sugar and acid
better understand how flavor components impact our sensory solutions according to the proportions mentioned above. The
perceptions. Another goal was to determine useful measurements resulting mixture was then blended with 250 mL thawed mango
that relate to flavor perception for mango. Currently, breeders and homogenate. A 100 mL subsample was drawn from each treat-
others interested in developing flavor traits in fruit and vegetables ment and stored in a closed plastic container at –20 ± 2 °C until
must rely on sensory analyses which are time consuming, expen- use for chemical analysis. The remaining treatment blends were
sive, and require much expertise. Determining the relationship of evaluated using sensory methods. Two replications were con-
chemical measures to sensory data would allow their use in lieu ducted.
of sensory studies to predict certain flavor properties. Thus, we SENSORY EVALUATION. The treatment blends were evaluated
hope to increase understanding of mango flavor and identify using a method modified from the Spectrum technique (Sensory
useful tools for breeders and molecular biologists for use in Spectrum, Chatham, N.J.) for descriptive analysis (Meilgaard et
improving mango flavor quality. al., 1991). The procedure calibrated panelists based on reference
standards established for the Spectrum method but rated samples
Materials and Methods using 150-mm nonstructured line scales (Galvez and Resurreccion,
1990). This modified method was used previously in descriptive
SAMPLE PREPARATION AND EXPERIMENTAL DESIGN. ‘Tommy analysis of muffins (Holt et al., 1992a), tortillas (Holt et al.,
Atkins’ mangoes were purchased in July, 1996 from a commer- 1992b), whipped topping (Abdullah et al., 1993b), and coffee
cial mango distributor based in Homestead, Fla. Excess fruit were whiteners (Abdullah et al., 1993a; Malundo et al., 1994; Malundo
bought and only fully ripe mangoes, in good condition, and and Resurreccion, 1993), and mungbean noodles (Galvez et al.
harvested at the morphologically mature stage as described by 1995).
Medlicott et al. (1988) were used. Ten judges, trained previously in the descriptive analysis
Ten batches of 600 g pulp from 30 fruit were homogenized technique employed, were chosen to participate in the evalua-
using a 12-speed Oster blender (Oster Corp., Milwaukee, Wis.) tions. All had consumed mangoes previously and did not have any
and combined to form a composite sample. Mango homogenate negative reaction towards the fruit.
was distributed among plastic containers which were immedi- The panelists were trained to evaluate the flavor of the treat-
ately capped and stored at –20 ± 2 °C until use. The plastic ment blends in six 1-h training sessions by the principal investi-
containers were purchased specifically for the storage of mango gator. During training, panelists tasted 100% and 50% mango
samples. homogenate (diluted with an equal volume of deionized water)
Solutions of 60% (w/v in deionized water) sugar and 40% samples. The flavor terminology was developed using diluted
(w/v) citric acid (U.S.P.–F.C.C., J.T. Baker, Inc., Phillipsburg, mango homogenate. Panelists suggested descriptors and refer-
N.J.) were prepared and stored at 5 ± 2 °C until used. The sugar ences for evaluating the flavor of samples presented (i.e., the

116 J. AMER. SOC. HORT. SCI. 126(1):115–121. 2001.

Table 1. Flavor descriptors used to evaluate mango samples.
Descriptorz Definition
Tastes
Sweet Taste stimulated by sugars like sucrose, fructose, and glucose
Sour Taste stimulated by acids such as citric and malic
Bitter Taste stimulated by substances such as caffeine
Aromatics
Peachy Aromatic associated with ripe peach
Pine/turpentine Aromatic common to both pine tar and turpentine
Sweet potato Aromatic associated with boiled sweet potato
Banana Aromatic associated with ripe banana
Grassy Aromatic associated with newly cut grass
Orange peel Aromatic associated with orange rind
Chemical feeling factors
Astringent The shrinking or puckering of the tongue surface caused by substances such as tannin and alum
Biting The stinging sensation felt on the tongue after drinking carbonated drinks such as soda
zThe panelists were presented with peaches, turpentine, boiled sweet potato, banana, cut grass, and orange rind for confirmation of descriptors.

panelists developed their own terminology, such as sweet potato, two replications).
banana, or peach. and were then presented with those materials Following sample preparation procedures described previ-
during the training for confirmation). An initial list of 18 descrip- ously, samples were prepared 1 h before evaluation and coded
tors was generated during the first and second sessions. This with three-digit random numbers. Twenty-five-milliliter por-
number was reduced to 11 after subsequent training in which tions were then placed in 60-mL plastic soufflé cups which were
standard references were introduced to clarify terms (Rainey, capped immediately. Malundo et al. (1997) determined that the
1986). The final list of descriptors used during evaluations, flavor profile of 50% mango homogenate generated through
together with corresponding definitions, is tabulated in Table 1. headspace analysis did not change significantly when the samples
Except for the descriptors sweet and sour, as noted below, were left at room temperature in covered containers for up to 2 h.
panelists were calibrated using reference standards established Panelists rated samples individually in partitioned sensory
for Spectrum Analysis (Meilgaard et al., 1991). Intensities of evaluation booths under red light. Samples were served one at a
standards, developed on a 15-point category scale, were adapted time in random order. Using scoresheets provided, panelists were
to the 150-mm nonstructured line scale by multiplying by a factor requested to rate samples relative to intensity ratings of standards
of 10. Since some treatment blends were expected to contain 18% established during training (Table 2). They were served the
to 28% sugar and 1.60% to 1.85% acid, sucrose (12%, 24%, and standards, and unsalted crackers and water to clear their palates
36%) and citric acid (0.7%, 1.4%, and 2.1%) solutions were used between samples.
as reference standards for the descriptors sweet and sour, respec- CHEMICAL ANALYSES. Frozen samples were thawed under
tively. These solutions were assigned intensity ratings by panel- running tap water. For each of the 12 treatments, a 40 mL sample
ists along the 150-mm line scale. was centrifuged at 12,100 gn for 15 min. TSS content (%) and pH
Based on ratings of reference standards already established, of the supernatant fluid were determined using a refractometer
panelists then rated two mango standards, labeled Standard A and (Palatte PR-101; Atago Co., Ltd., Japan) and titrator (model
Standard B, for all flavor descriptors listed in Table 1. The mango 230A; ATI Orion, Beverly, Mass.). TA, expressed as percentage
standards were used by panelists as warm-up samples during citric acid, was calculated after titrating 10 mL supernatant fluid
evaluations (Malundo et al., 1994; Rutledge and Hudson, 1990; to a pH of 8.1 with standardized 0.1 mol·L–1 NaOH. A TSS/TA
St. Angelo et al., 1992). These were 50% mango homogenate, ratio for each treatment was also calculated. There were two
other than the experimental samples, selected after preliminary replicates per sample.
screening of test samples with varying intensities for flavor STATISTICAL ANALYSIS. Data were analyzed using procedures
properties. The mango standards were intended to represent of the Statistical Analysis System (SAS Institute Inc., 1990).
extremes in the expected range of mango flavor (but not necessar- Performance of individual panelists was evaluated using cluster
ily representing the extremes on the 150 mm scale which was analysis (PROC VARCLUS). PROC MEANS was used to calcu-
established with the Spectrum standards, Meilgaard et al., 1991). late means and SDs of panel ratings. Flavor components which
Standards A and B were prepared and stored at –20 ± 2 °C until affected the response of the descriptive variables were identified
needed. The intensity ratings assigned to all reference standards by regression analysis. With mean scores across replicates as
during training are listed in Table 2. dependent variables (O’Mahony, 1995; Schutz, 1983), a reduced
A standard for good performance was also set during training. second order polynomial was fit to the descriptive data. The
Panelists were considered to be performing well when ratings polynomial took the form: Σ(y) = β1x1 + β2x2 + β3x3 + β12x1x2 +
were within ±10 mm from the mean based on a 150-mm line scale β13x1x3 + β23x2x3 + δ12x1x2(x1 – x2) + δ13x1x3(x1 – x3) + δ23x2x3(x2 –
(Meilgaard et al., 1991). x3), where Σ(y) was the expected value of the response of the
Each replication of the experimental design was evaluated dependent flavor variables; x1, x2, and x3, respectively (i.e.,
within two sessions. Therefore, each panelist rated six treatments expected value of the random variable = the descriptive re-
per evaluation. Two replications of the design were evaluated sponse), corresponded to the proportions of sugar, acid, and water
within the same day, one in the morning and the other in the in the blend; and β1, β2, β3, β12, β13 β23, δ12, δ13, and δ23 were
afternoon (i.e., each panelist evaluated each treatment in a total of coefficients associated with the model terms (Snee, 1979). This

J. AMER. SOC. HORT. SCI. 126(1):115–121. 2001. 117

model was derived after omitting the three-way interaction term During training, panelists noted that flavor properties were
(β123x1x2x3) from the full cubic polynomial since there were not detected more readily in diluted samples than in 100% mango
enough data points for the full model. PROC STEPWISE (back- homogenate. Although this observation should be verified with a
ward elimination) was used to determine whether a lesser number time-intensity study, results presented in Table 3 indicate that
of terms than that in the reduced cubic polynomial was adequate increasing water level increased intensities of all flavor descrip-
to predict the response of the flavor variables. The criterion for the tors evaluated. In a previous study, Malundo et al. (1997) ob-
elimination was a significance level of P ≥ 0.15. When an served that concentration of volatile compounds released into the
equation with lesser terms was sufficient, an F statistic was headspace increased when mango homogenate was diluted up to
calculated to ascertain adequacy of this model in replacing the 50%. In another study, Wilson et al. (1990) showed that the aroma
reduced cubic polynomial (Cornell, 1983). Finally, correlation and flavor thresholds of lactones, a flavor component isolated in
analysis (PROC CORR) was conducted to detect significant mango, is lower (i.e., detected at lower concentrations) when
correlations between flavor properties of mangoes and chemical placed in water than in mango puree as determined by a panel of
methods for measuring sugars and acids.
Table 2. Intensity ratings of standards for attributes used to evaluate
Results and Discussion mango flavor.
Standard Ratingz
PANEL PERFORMANCE. Cluster analysis was used to detect
outliers among the trained panelists (Powers, 1988). One indi- Tastes
vidual was determined to be a consistent outlier. Therefore, Sweet
evaluations of only 9 of 10 trained judges were used in subsequent Standard B 20
analysis. 12 % sugar 55
SDs (not presented) were used to assess panel performance.
Standard A 63
Based on the Empirical Rule, the range of intensities assigned by 24 % sugar 124
practically all the nine panelists to a treatment was within ±2 SDs 6 % sugar 150
from the mean (Ott, 1988). The descriptor intensities of the test Sour
samples fell within the ranges illustrated in Table 2. Standard B 29
EFFECT OF SUGARS AND ACIDS ON FLAVOR PROPERTIES. Regres- 0.7 % acid 76
sion analyses and calculated F statistics indicated that a reduced Standard A 85
cubic canonical polynomial was not necessary to explain the 1.4 % acid 117
response of all descriptive variables evaluated (Table 3). Equa- 2.1 % acid 150
tions with fewer terms were, in fact, adequate. The models Bitter
generated had relatively low CVs implying very little variation Standard A 22
between replicates in mean panel ratings. Therefore, despite 0.08 % caffeine 50
variations observed in panelists’ responses, using mean scores to 0.15 % caffeine 100
measure the intensity of flavor attributes as suggested previously Aromatics
(O’Mahony, 1995; Schutz, 1983) would seem like a logical Peach
approach. Standard B 26
Of the 11 properties evaluated, acid levels affected only six, Standard A 53
namely, sweet, sour, peachy, pine, astringent, and biting. The Pine/turpentine
effects of acid on sourness and astringency were expected. Acids Standard A 15
are noted for their sour taste. In fact, these are even used as Standard B 50
reference standards in defining the sour perception during de- Sweet potato
scriptive analysis (Meilgaard et al., 1991). In addition, several Standard A 19
investigators have characterized acids as having an astringent Standard B 31
note (Hartwig and McDaniel, 1995; Rubico, 1993; Rubico and Banana
McDaniel, 1992; Straub, 1992). The effect of acids in potentiat- Standard A 24
ing sweetness was not expected but the description of acids as Standard B 38
possessing a sweet taste by all 12 judges in a study involving free- Grassy
choice profiling (Hartwig and McDaniel, 1995) provides a pos- Standard A 20
sible explanation. Standard B 53
Sugar enhanced the perception of all flavor attributes except Orange peel
for sour taste (Table 3). Between the two taste components Standard A 19
studied, only sugar had any significant effect on the intensities of Standard B 23
the descriptors bitter, sweet potato, banana, grassy, and orange Chemical feeling factors
peel. The positive contribution of sugar to bitter taste was ex- Astringent
pected. Although several investigators have observed a bitter Standard A 22
flavor note in acids (Hartwig and McDaniel, 1995; Rubico, 1993; Standard B 37
Rubico and McDaniel, 1992), present understanding of the rela- Biting
tionship between molecular structure and taste receptor sites Standard A 35
would tend to support the results of this study as sweet and bitter Standard B 46
sensations are triggered by similar molecular features (Lindsay, zIntensity ratings based on 150-mm unstructured line scales anchored

1988). with the descriptors weak (=12.5) and strong (=137.5).

118 J. AMER. SOC. HORT. SCI. 126(1):115–121. 2001.

Table 3. Results of regression analyses conducted to model the responses of flavor properties to varying proportions of sugar (x1), acid (x2), and water
(x3) in mango homogenate.
Descriptorz
Variable Sweet Sour Bitter Peachy Pine Sweetpotato Banana Grassy Orangepeel Astringent Biting
x1 * NS * * * * * * * * *
x2 * * NS * * NS NS NS NS * *
x3 * * * * * * * * * *
x1 x2 * NS NS * NS NS NS NS NS * *
x1 x3 * NS NS * NS NS NS * NS NS NS
x2 x3 * * NS NS * NS NS NS NS * *
x1 x2(x1–x2) * NS NS NS * NS NS NS NS NS *
x1 x3(x1–x3) NS NS * NS NS * * NS NS NS NS
x2 x3(x2–x3) * NS NS * * NS NS NS NS NS NS
F statisticy <0.01 2.71 0.90 0.33 0.95 0.61 0.91 0.85 0.82 0.25 0.31
CV 8.76 11.66 6.38 4.36 12.46 3.30 3.20 10.28 5.72 3.94 2.50
R2 0.97 0.94 0.83 0.88 0.53 0.90 0.90 0.67 0.45 0.96 0.99
zCalculated to determine whether chosen model can replace reduced cubic canonical polynomial, P = 0.05.
*Significant at P = 0.15; interaction terms significant at P = 0.15 were included only if corresponding linear terms were significant.

12 experienced judges. The effect of water was most probably due ing, subject to error due to the variability of human judgement,
to its ability to release entrapped volatiles or disrupt apparent and, therefore, expensive. Most breeders and molecular biolo-
interactions between the mango pulp and flavor compounds, thus gists do not have access to trained panels, yet flavor quality is an
increasing the release of the latter (Harrison and Hills, 1997; important characteristic to be considered when evaluating culti-
Malundo et al., 1997). This may also indicate that the level of var selections. Understandably, there have been constant at-
juiciness in mango fruit is important to flavor perception. In tempts to replace sensory panels with instrumental and/or chemi-
addition, some of the interaction terms were also significant for cal measures that are not susceptible to fatigue nor psychological
some of the descriptors (for sweet in particular) which is not fluctuations which characterize human performance (Szczesniak,
surprising since, for example, acid affects perception of sweet- 1987; Trant et al., 1981), yet relate to the sensory experience.
ness and vice versa. In addition, the levels of sugars and acids Chemical measures for sugar and acid levels are currently
interacted to affect the responses to some of the aromatic descrip- standard indices for evaluating flavor quality in fruit and veg-
tors. etables. However, while sugar and acid concentrations indicate
The R2 values presented in Table 3 indicated that varying degree of ripeness, these do not necessarily relate to flavor
levels of sugar, acid, and/or water explained much of the varia- properties (Alavoine et al., 1990). The extent to which these
tions observed in mango samples in terms of the flavor descrip- measurements correlate with descriptive ratings must be estab-
tors sweet, sour, bitter, peachy, sweet potato, banana, astringent, lished (Trant et al., 1981) for them to be useful tools for breeders
and biting. Based on the discussions presented above, sugars and and others interested in screening genetic material for flavor
acids would be expected to account for much of the sweetness quality.
variations observed among the samples evaluated. In addition, A chemical measure can be used with confidence as a predictor
acid would logically be responsible for most of the sour taste in of descriptive score if the simple correlation coefficient between
mango. A variety of compounds contributing to the bitterness and the two variables is ≥0.90. When the coefficient is ≥0.80 but
astringency of fruit are known (Kays, 1991), however, results in
Table 3 indicate that variations in sugars and acids accounted for Table 4. Pearson correlation coefficients (r) between flavor variables and
much of the bitterness and astringent sensations in mango. In chemical measurements (n = 24).
addition, the contribution of these two in explaining variations in Chemical measurement
fruity flavor notes (i.e., peachy, sweet potato, and banana) were
Flavor variable TSS pH TA TSS/TA
also major. It is uncertain whether these observations were due to
the psychological association of sugars and acids with fruit in the Taste
minds of the panelists or from actual chemical or physical Sweet 0.72* 0.48** –0.44** 0.60*
interactions of the two taste components with other flavor com- Sour NS –0.96* 0.95* –0.84*
pounds. However, the importance of sugars and acids as potentia- Bitter –0.49** –0.54* 0.48** –0.57*
tors of some aromatic flavor sensations was clearly established. Aromatics
Sugar and acid also explained some of the variations observed Peachy 0.79* NS NS NS

among the samples in terms of the flavor notes pine/turpentine, Pine/turpentine –0.52* NS NS NS

grassy, and orange peel. However, because of the relatively low Sweet potato 0.41** 0.74* –0.72* 0.74*
coefficients of variation (R2) resulting from regression analysis, Banana 0.47** 0.63* –0.64* 0.62*
it is obvious that other flavor compounds, such as aroma volatiles, Grassy –0.66* NS NS NS

also are factors in the sensation of these attributes as was shown Orange peel NS –0.56* 0.51** –0.60*
by Malundo et al. (1996). Chemical feeling
CORRELATIONS BETWEEN FLAVOR AND CHEMICAL MEASURE- Astringent NS –0.96* 0.91* –0.87*
MENTS. Despite test protocols that minimize bias, use of descrip-
Biting NS –0.96* 0.91* –0.91*
NS ,*,**
tive panels to evaluate product quality is complex, time-consum- Nonsignificant or significant at P < 0.05 or 0.01, respectively.

J. AMER. SOC. HORT. SCI. 126(1):115–121. 2001. 119

<0.90 then the chemical measure may still be used as a predictor Baldwin, E.A., J.W. Scott, M.A. Einstein, T.M.M. Malundo, B.T. Carr,
but with less confidence. A correlation ≥0.70 but <0.80 indicate R.L. Shewfelt, and K.S. Tandon. 1998. Relationship between sensory
that the chemical index is only a marginal predictor of sensory and instrumental analysis for tomato flavor. J. Amer. Soc. Hort. Sci.
ratings (Bourne, 1982). 123:906–915.
Bourne, M.C. 1982. Food texture and viscosity: Concept and measure-
Although there were a number of significant correlations
ment. Academic Press, New York.
between the descriptive and chemical properties of the experi- Cornell, J.A. 1983. How to run mixture experiments for product quality.
mental blends, not all of them were high enough to be considered Amer. Soc. for Quality Control, Milwaukee, Wis.
important (Table 4). In addition, correlations that were not Fuchs, Y., E. Pesis, and G. Zuberman. 1980. Changes in amylase
indicative of a cause-and-effect relationship had limited value. activity, starch, and sugar contents in mango fruit pulp. Scientia Hort.
The TSS/TA ratio, often used as a measure of sweetness, was 13:155–160.
not useful in measuring this descriptor in mango (r = 0.60), but Galvez, F.C.F. and A.V.A. Resurreccion. 1990. Comparison of three
was a reasonable predictor of the descriptors sour and astringent descriptive analysis scaling methods for the sensory evaluation of
(r = –0.84 and 0.87, respectively), although pH and TA were noodles. J. Sensory Studies 5:251–263.
much better predictors of the two descriptors (r > |0.90|). Similar Galvez, F.C.F., A.V.A. Resurreccion, and G.O. Ware. 1995. Formula-
tion and process optimization of mungbean noodles. J. Food Process-
results were found for tomato (Lycopersicon esculentum Mill.)
ing and Preservation 19:191–205.
fruit (Baldwin et al., 1998). TA, pH, and TSS/TA were also very Gholap, A.S., C. Bandyopadhyay, and G.B. Nadkarni. 1986. Aroma
good predictors of biting (r > |0.90|). TSS was not a particularly development in mango fruit. J. Food Biochem. 10:217–229.
useful predictor of any of the descriptive variables evaluated., Gowda, I.N.D., and K.H. Ramanjaneya. 1994. Studies on physico-
except possibly peachy aroma (r = 0.79). One could argue that chemical characteristics of some commercial cultivars of mango.
peachy aroma and sweetness may interact in terms of the psychol- Indian Food Packer 48:45–49.
ogy of our perception. No measure was a reasonable predictor of Hartwig, P. and M.R. McDaniel. 1995. Flavor characteristics of lactic,
sweetness (TSS was marginal at best with r = 0.72). malic, citric, and acetic acids at various pH levels. J. Food Sci. 60:384–
388.
Harrison, M. And B.P Hills. 1997. Mathematical model of flavor release
Conclusions
from liquids containing aroma-binding macromolecules. J. Agr. Food
Chem. 45:1883–1890.
Before this study, no in-depth research on the role of sugars Holt, S.D., K.H. McWatters, and A.V.A. Resurreccion. 1992a. Valida-
and acids on the perception of mango flavor had been conducted. tion of predicted baking performance of muffins containing mixtures
Therefore, in the past, the significant contribution to mango of wheat, cowpea, peanut, sorghum, and cassava flour. J. Food Sci.
flavor attributed to these two taste components (Medlicott and 57:470–474.
Thompson, 1985) was merely an assumption perhaps based on Holt, S.D., A.V.A. Resurreccion, and K.H. McWatters. 1992b. Formu-
generalizations established for other fruit. lation, evaluation, and optimization of tortillas containing wheat,
Understanding the chemical components that contribute to cowpea, and peanut flours using mixture response methodology. J.
flavor and how to best measure them is necessary for flavor Food Sci. 57:121–127.
improvement efforts in breeding programs. Data generated in this Kapse, B.M., D.A. Rane, and D.M. Khedkar. 1988. Correlation between
bio-chemical parameters and organoleptic evaluation in mango variet-
study were consistent with the present understanding of the
ies. Acta Hort. 231:756–762.
chemical basis of flavor and results of related studies. Therefore, Kays, S.J. 1991. Postharvest physiology of perishable plant products.
many of the conclusions drawn fell within the range of normal Van Nostrand Reinhold, New York.
expectations. However, the extent of the influence of sugars and Kumar, R., R.A. Kaushik, and A.S. Chharia. 1992. Effect of post-harvest
acids on the intensity of specific flavor notes, especially the treatments on the quality of mango during storage. Haryana J. Hort. Sci.
aromatics and chemical feeling factors, was new and noteworthy. 21:46–55.
This study does not minimize the importance of volatiles and Kundu, S. and S.N. Ghosh. 1992. Studies on physico-chemical charac-
other flavor compounds on perception of mango flavor, but does teristics of mango cultivars grown in the laterite tract of West Bengal.
establish sugars and acids as primary taste compounds in mango Haryana J. Hort. Sci. 21:129–134.
and as important potentiators of the aromatic components of Lindsay, R.C. 1988. Flavors, p. 723–765. In: O.R. Fennema (ed.). Food
chemistry. Marcel Dekker, Inc., New York.
mango flavor. However, measurements of TSS or TSS/TA were
Lizada, C. 1993. Mango, p. 255–271. In: B. Seymour, J.E. Taylor, and
not useful predictors of sweetness, although they are often used G.A. Tucker (eds.). Biochemistry of fruit ripening. Chapman and Hall,
for this purpose. Meanwhile pH, TA, and TSS/TA related well to New York.
sourness, astringency, and biting. This should be taken into Malundo, T.M.M., E.A. Baldwin, M.G. Moshonas, R.A. Baker, and R.L.
account when these measurements are used to screen breeding Shewfelt. 1997. Method for the rapid headspace analysis of mango
material. (Mangifera indica L.) homogenate volatile constituents and factors
affecting quantitative results. J. Agr. Food Chem. 45:2187–2194.
Literature Cited Malundo, T.M.M., E.A. Baldwin, G.O. Ware, and R.L. Shewfelt. 1996.
Volatile composition and interaction influence flavor properties of
Abdullah, A., T.M.M. Malundo, A.V.A. Resurreccion, and L.R. Beuchat. mango. Proc. Proc Fla. State Hort. Soc. 109:264–268.
1993a. Descriptive sensory profiling for optimizing the formulation of Malundo, T.M.M. and A.V.A Resurreccion. 1993. Optimization of
a peanut milk-based coffee whitener. J. Food Sci. 58:120–123. liquid whitener from peanut extract. Lebensm.-Wiss. u.-Technol.
Abdullah, A., A.V.A. Resurreccion, and L.R. Beuchat. 1993b. Formula- 26:552–557.
tion and evaluation of a peanut milk based whipped topping using Malundo, T.M.M., A.V.A. Resurreccion, G.O. Ware, and L.R. Beuchat.
response surface methodology. Lebensm.-Wiss. u.-Technol. 26:162– 1994. Sensory, physical, and microbiological properties of liquid
166. whitener from peanuts. J. Food Sci. 59:338–343, 349.
Alavoine, F., M. Crochon, and C. Bouillon. 1990. Practical methods to Medlicott, A.P., S.B. Reynolds, S.W. New, and A.K. Thompson. 1988.
estimate taste quality of fruit: How to tell it to the consumer. Acta Hort. Harvest maturity effects on mango fruit ripening. Tropical. Agr.
259:61–68. 65:153–157.

120 J. AMER. SOC. HORT. SCI. 126(1):115–121. 2001.

Medlicott, A.P. and A.K. Thompson. 1985. Analysis of sugars and Technol. 37(11):46–48, 62.
organic acids in ripening mango fruit (Mangifera indica L. var. Keitt) Selvaraj, Y., R. Kumar., and D.K. Pal. 1989. Changes in sugar, organic
by high performance liquid chromatography. J. Sci. Food Agr. 36:561– acids, amino acids, lipid constituents and aroma characteristics of
566. ripening mango (Mangifera indica L.) fruit. J. Food. Sci. Technol.
Meilgaard, M., G.V. Civille, and B.T. Carr. 1991. Sensory Evaluation 26:308–313.
Techniques. CRC Press, Inc., Boca Raton, Fla. Shashirekha, M.S. and M.V. Patwardhan. 1976. Changes in amino acids,
Mitra, S. and E.A. Baldwin. 1997. Mango, p. 85–122. In: S. Mitra (ed.). sugars, and nonvolatile organic acids in a ripening mango fruit
Postharvest physiology and storage of tropical and subtropical fruits. (Mangifera indica, Badami variety). Lebensm.-Wiss. u.-Technol.
CAB International, New York. 9:369–370.
Morga, N.S., A.O. Lustre, M.M. Tunac, A.H. Balagot, and M.R. Soriano. Snee, R.D. 1979. Experimenting with mixtures. Chemtech. 9:702–710.
1979. Physico-chemical changes in Philippine Carabao mangoes St. Angelo, A.J., B.T. Vinyard, and K.L. Bett. 1994. Sensory and
during ripening. Food Chem. 4:225–234. statistical analysis in meat flavor research p. 267–290. In: F. Shahidi
Narain, N., P.S. Bora, R. Narain, and P.E. Shaw. 1998. Mango, p. 1–77. (ed.). Flavor of meat and meat products. Elsevier Science Publishers,
In: P.E. Shaw, H.T. Chan, Jr., and S. Nagy (eds.). Agscience, Inc., Amsterdam, The Netherlands.
Auburndale, Fla. Straub, A.M. 1992. Power function determination of sourness and
O’Mahony, M. 1995. Sensory measurement in food science: Fitting astringency and time-intensity measurements of sourness and astrin-
methods to goals. Food Technol. 49(4):72–82. gency for selected acids. MS thesis. Oregon State Univ., Corvallis.
Ott, L. 1988. An introduction to statistical methods and data analysis. 3rd Szczesniak, A.S. 1987. Correlating sensory with instrumental texture
ed. PWS-Kent Publishing Co., Boston. measurements—An overview of recent developments. J. Texture Studies
Powers, J.J. 1988. Using general statistical programs to evaluate sensory 18:1–15.
data. Food Technol. 38(6):74–82. Tandon, D.K. and S.K. Kalra. 1983. Changes in sugars, starch and
Rainey, B.A. 1986. Importance of reference standards in training panel- amylase activity during development of mango fruit cv. Dashehari. J.
ists. J. Sensory Studies 1:149–154. Hort. Sci. 58:449–453.
Rubico, S.M. 1993. Perceptual characteristics of selected acidulants by Trant, A.S., R.M. Pangborn, and A.C. Little. 1981. Potential fallacy of
different sensory and multivariate methods. PhD diss., Oregon State correlating hedonic responses with physical and chemical measure-
Univ., Corvallis. ments. J. Food Sci. 46:583–588.
Rubico, S.M. and M.R. McDaniel. 1992. Sensory evaluation of acids by Vance Publishing Corporation. 1994. Fresh trends: A profile of the fresh
free-choice profiling. Chemical Senses. 17:273–289. produce consumer. The Packer C(54):6–103.
Rutledge, K.P. and J.M. Hudson. 1990. Sensory evaluation: Methods for Vazquez-Salinas, C. and S. Lakshminarayana. 1985. Compositional
establishing and training a descriptive flavor panel. Food Technol. changes in mango fruit during ripening at different storage tempera-
44(12):78–84. tures. J. Food Sci. 50:1646–1648.
SAS Inst., Inc. 1990. SAS user’s guide: Statistics. 6th ed. SAS Inst., Inc., Wilson, III, C.W. P.E. Shaw, and R.J. Knight, Jr. 1990. Importance of
Cary, N.C. some lactones and 2,5-dimethyl-4-hydroxy-3(2H)-furanone to mango
Schutz, H.G. 1983. Multiple regression approach to optimization. Food (Mangifera indica L.) aroma. J. Agr. Food Chem. 38:1556–1559.

J. AMER. SOC. HORT. SCI. 126(1):115–121. 2001. 121