You are on page 1of 8

Sensory and Objective Measurement of Sandiness in Duke de Leche,

a Typical Argentine Dairy Product


GUILLERMO HOUGH, EDGARW MARTINEZ, and ADRIANA CONTARlNl
lnstituto Superior Experimental de Tecnologia Alirnentaria
H. lrigoyen 931
6500 Nueve de Julio
Buenos Aires, Argentina

ABSTRACT is concentrated to 70% total solids by boiling at


atmospheric pressure. From milk with 12% to-
Lactose crystallization in Dulce de
tal solids and 4.5% lactose, lactose concentra-
Leche can produce a sandiness defect. tion in Dulce de Leche is 9.85 d l 0 0 g. Consid-
The objective measurement and sensory ering the water phase, the lactose concentration
evaluation of this defect were the thrusts is 33 @lo0 g water. Solubility of lactose at 15
of this study. The technique for the mi- and 30°C is 16.9 and 24.8 g/100 g water,
croscopic measurement of the number respectively (20). So even without interfer-
and size of crystals is described in detail. ences, lactose in Duke de Leche is initially in a
Sensory panel members found that iac- supersaturated solution. This is compounded by
tose crystals added to Dulce de Leche the simultaneous presence of sucrose (146 g/
produced a sandiness defect equivalent to 100 g water), which substantially reduces lac-
added sand crystals. The latter are far tose solubility (14). Under these conditions,
easier to define by size so they were used lactose crystallization is inevitable in Dulce de
for magnitude estimation, where the ex- Leche, which leads to the prevalent defect of
ponent of Steven’s law was approxi- sandiness.
mately 1. Based on this, seven-point a Technological methods have been developed
sensory scale for sandiness was devel- to eliminate this defect in Dulce de Leche (2,
oped using Dulce de Leche mixed with 16, 17, 18, 19) and in sweetened condensed
sand of crystal sizes varying between 0 milk (1, 4, 7, 9), which is of similar composi-
and 1500 pm. A double enay table is tion.
presented whereby the sensory sandiness To evaluate experimental results these au-
threshold can be read as a function of size thors used arbitrary category scales for the sen-
and number of crystals. With crystal size sory measurement of sandiness, and in some
below 6 pm sandiness is not detected, cases measured crystal size microscopically (5,
even if all lactose in Dulce de Leche is 7, 9, 17).
crystallized. Above this size, the detec- Hunziker (7) is the only investigator to re-
tion threshold depends on number of late quantitatively the number and size of crys-
crystals. tals with a sensory scale for degree of sandi-
(Key words: Dulce de Leche, sandiness, ness. Santos et al. (17) noted a contradiction in
sensory evaluation) that “no sandiness” was the score for crystals
smaller than 10 pm yet Hunziker recommended
INTRODUCTION seeding with 400-mesh lactose, which has a
37-pm aperture. According to Hunziker (7),
Dulce de Leche, a typical Argentine dairy absence of sandiness is obtained when there are
product similar to sweetened condensed milk, more than 4 x lo8 crystals/cm3. With this
was described by Mor0 and Hough (12) in a number, he calculated a corresponding 10-p
previous paper. A standard initial formulation crystal size; in the calculation he used a lactose
is 10 parts of milk and 2 parts of sucrose. This solubility value, which does not coincide with
the solubility of lactose in the presence of
sucrose reported by Nickerson (14). Other
Received December 1 , 1988. points of Hunziker’s sandiness scale are based
Accepted July 24, 1989. on an arbitrary reduction in number of crystals.

1990 J Dairy Sci 73:604-611 604


MEASUREMENT OF SANDINESS IN DULCE DE LECHE 605
Clearly, for research and quality control of was randomly placed over the slide. Magnifica-
sandiness in Dulce de Leche, the following tion was selected until approximately 20 crys-
aspects need developing: 1) a sensory percep- tals appeared in the field. If there were too
tion scale of sandiness in relation to objectively many crystals at maximum magnification, a
measured size and number of crystals, and 2) reticulated portion of the slide (ln, 1/4, etc.)
sensory threshold of sandiness as a function of containing approximately 20 crystals was used.
size and number of lactose crystals. The objec- Average size of individual measurements was
tive of this study was to cover both these taken.
aspects. In some samples size stratification was clear.
In these cases, the number of crystals are re-
MATERIALS AND METHODS ported separately for each size.

Size and Number of Crystals Sensory Scale

There is no published standard for measur- Where crystal-free Duke de Leche (CFDL)
ing the size and number of lactose crystals in was necessary, it was provided from a fresh
d a y products. Crystalline lactose in milk prod- batch by Mastellone Hnos, Buenos Aires, Ar-
ucts can be detected simply by means of a gentinc who used P-galactosidase to reduce
polarizing microscope (15). In this study an lactose content and prevent sandiness. To en-
Olympus BHI Model BH-2 (Tokyo, Jpn) mi- sure no crystals were present, samples were
croscope was used equipped with a polarizing first observed under the microscope with polar-
filter, PM-6/ENM-7 Olympus photography and ized light.
ocular with calibrated scale to allow sectorizing Sand used for mixing with CFDL to simu-
an observed field and measurement of crystal late lactose crystals was obtained from local
axis. construction stores. The sand was washed,
Lactose crystal numbers and size were mea- dried, and finally placed overnight in an oven at
sured as follows: a microscope slide and cover 130°C for sterilization. To obtain different
were weighed. Approximately ,003 g of Dulce granulomeay, the sand was sieved through
de Leche was then weighed onto the slide. The standard-sized screens.
center of the cover slip was placed over the A 12-member panel was used for sensory
Duke de Leche, and a slight pressure was evaluations. Training included selection by per-
exerted. Care was taken not to crush crystals; if formance on taste tests and sequential analysis
this was observed through the microscope, the (8), followed by 10 d of intensive training in
slide was prepared again. The resulting sample sensory practices (8).
on the slide was circular in appearance. Crys- Evaluations consisted of four 1-h sessions
tals were counted in each of 10 randomly cho- where the panel evaluated different samples
sen microscopic fields and then averaged to using as a scale a IO-cm line marked on one
calculate: end “no sandiness” and “extreme sandiness” on
the other. This scale was used during training
N = nf rt R2/(FW) and development of a consensus evaluation
procedure. Also at this stage, samples of
where N = number of crystals per gram of CFDL, mixed with washed and sterilized sand
Dulce de Leche, nf = average number of crys- of different granulometry, were presented to the
tals observed in F, R = radius of sample (mm), panel to compare sensory perception of this
F = area of microscopic field (mm), and W = sandiness with that of samples with lactose
weight of Dulce de Leche sample. The same crystals.
prepared slide was used for measuring crystal The next step was to establish sensory ver-
size. The length of the longest axis of a-lactose sus objective evaluation of sandiness. The fixed
crystal is the most relevant to sensory analysis; modules magnitude estimation method was
therefore, this is the measure in which sizes are chosen (1 1). For this, lactose crystals of defi-
expressed. Within a Dulce de Leche sample, nite size are needed, but these are difficult to
lactose crystal sizes are not uniform; therefore, obtain. To allow lactose naturally present in
averaging was required. A microscopic field Dulce de Leche to crystallize to different sizes
Journal of Dairy Science Vol. 73, No. 3, 1990
606 HOUGH ET A L
TABLE 1. Sieves. resulting crystal size, and concentration of sand used for magnitude estimation of sensory versus
objective crystal size.
Concentration
Sieve CrYd of sand in Sensory
aperture size Dulce Leche scale'
( m 7 Wloog)
... ... 0 0
125 115 5 1
250 307 5 3
700 622 5 6
840 895 10 9
1250 1207 10 12
I500 1490 10 15
'Standard number as presented to panelists.

was impractical because of lack of size unifor- The reason crystal sizes are larger than the
mity. Because of these difficulties, sand sieved corresponding aperture is because in measuring
through different-sized screens was used as a size, the longest axis is considered. For the
replacement. Concentration of sand in CFDL smallest crystal size, Nestle's (Buenos Aires,
was determined by approximating the number Argentina) microcrystalline lactose (used for
of sand crystals to the number of lactose crys- d i n g sweetened condensed milk) was em-
tals of similar size present in commercial Dulce ployed as a standard. For this lactose, a 1.5%
de Leche. The sieves, sand crystal sizes, and mixture with CFDL provides a size and number
concentrations used for magnitude estimation distribution as shown in Table 3.
are in Table 1 . Sizes cover the range found in The sensory analysis method used for
Dulce de Leche. Prior to the test, the panel threshold determination is described by Lun-
received an orientation session (13). Magnitude dah1 et al. (10). The same testing panel de-
estimation was done in duplicate on different scribed earlier was used. To determine sensory
days. threshold four experiments were performed.
Based on magnitude estimation, a sensory Experiment 1 . Sandiness threshold was de-
scale was developed. With this, samples of termined using six sizes of lactose crystals;
Dulce de Leche with different size and number each one was mixed with CFDL in a 1.5%
of lactose crystals were measured to control concentration. Crystal sizes used were 45, 57,
scale validity. Also, for the same crystal size, 69, 80, and 105 pn and microcrystalline lac-
sand concentration was varied to see if it signif- tose (see Table 3). Mixtures of CFDL with
icantly affected sensory evaluation. These vari- lactose were done within an hour of sensory
ations are summarized in Table 2. evaluation. Crystal-free Dulce de Leche was
used as the control sample (10). All samples,
Sandiness Threshold
As mentioned, for larger crystal sizes, sand
replaced lactose. For threshold sizes, solubiliza- TABLE 2. Sand concentration used to study influence of
tion of lactose during sensory evaluation would this parameter on sensory perception of sandiness.
have probably influenced perception; therefore,
Concentrations
analytical grade lactose was used. It was of sand
crushed in a mortar and then passed through a Sensory used in
series of sieves with apertures of 37, 44, 53, scale Dulce
and 62 p.Smaller apertures are not currently size de Leche
available. The size of resultant crystals was (g/lOo/g)
measured by dispersing them in CFDL in ap- 1 2.5, 5. 10
proximately 1?b concentration, obtaining the 6 2.5. 5, 10
following sizes: 45, 57, 69, 80, and 105 pm, 12 5 , 10, 20

Journal of Dairy Science Vol. 73, NO. 3, 1990


MEASUREMENT OF SANDINESS IN DULCE DE LECHE 607
TABLE 3. Size and number of lactose crystals in sample against CFDL was performed; panelists an-
prepared with 1.5% microcrystalline lactose in crystal-free swered which of the two samples presented
Duke de Leche.
sandiness. The seeded sample and CFDL dif-
crystal fered in color and flavor; therefore, to avoid
size CNSdS bias, two deliberately confusing pairs were
added at each session. These were: CFDL with
1.5% added lactose of 105-pm crystal size
(slightly above threshold), tested against the
seeded sample or against CFDL. Each panelist
tested 10 pairs in total, divided in two different
sessions. Subtracting the confusing pairs gave
six answers per panelist.
including control, were vigorously agitated
with a spoon in a serving cup before serving, Statistical Analysis
and panelists were instructed to do likewise Magnitude estimation results were analyzed
before taking the sample to mouth. This was as described by Mori et al. (11) and threshold
necessary to mix the lactose thoroughly and to results as proposed by Lundahl et al. (10). In all
ensure uniformity in consistency of all samples. cases, significance was determined at Pc.05.
The determination was done by duplicate.
Experiment 2. Sandiness threshold was de-
termined using 45-pm lactose crystals in six RESULTS AND DISCUSSION
different concentrations: . I , .25, .5, .75, 1, and
1.5%. Control and agitation were the same as Size and Number of Crystals
previous experiment.
Experiment 3. Experiment 3 was conducted The microscopic method to measure size and
as Experiment 2 but lactose crystals were 105 number of crystals was replicated 11 times on
P- the same Dulce de Leche. Average and stan-
Experiment 4 . Presence of sensory sandiness dard deviation for number of crystals were,
was determined in Dulce de Leche seeded with respectively, 2.2 x lo9 and 9.9 x lo8 crystals/g.
microcrystalline lactose. To this purpose, Dulce As to size, the mean was 4.1 pn and the
de Leche was made following a classical recipe standard deviation was .28 pm. If quadruplicate
(12) and then seeded with Nestle's microcrys- measurements are performed, the confidence
talline lactose according to the procedure given interval (b.05)for number of crystals is 2.05
by Hunziker (7) for sweetened condensed milk. x 109 crystais/g; for size it is .5 pm. Both are
After 15 d of storage at room temperature, size considered adequate for practical purposes.
and number of crystals were measured micro-
scopically. Also, sensory evaluation was done Sandiness Scale
to see if the seeded sample was above or below
threshold. The sensory scale developed from The first responses from the sensory panel
the magnitude estimation test was used. To provided a consensus sensory technique for
c o n f m the result, a paired comparison test evaluating sandiness: one-half teaspoonful of

TABLE 4. Sensory sandiness scores, and crystal sizes and numbers of different Duke de Leche samdes.

crystal Sensory scores


Sample size cryds Mean 95% Cll
(w) (no./@
A 190 4 x 104 1.7 .5
B 650 1 x 103 1.0 .8
C 78 7 x 104 .55 .39
'Confidence interval.

Journal of Dairy Science Vol. 73. No. 3, 1990


608 HOUGH ET AL.

Figure 1 . Duke de Leche with a) laclose crys(als and b) sand crystals as seen through microscope ( 1 0 0 ~ )with
polarized light. Both samples were classified as equivalent sandiness by panel. Each scale division corresponds to 14.7 pm.

Journal of Dairy Science Vol. 73. No. 3, 1990


MEASUREMENT OF SANDINESS IN DULCE DE LECHE 609
TABLE 5. Size and number of lactose crystals present in l sizes found in Dulce de Leche, i.e., 0 to
t
a
Dulce de Leche &ed with microcrystalline lactose during 1500 pm. Intermediate standards prepared with
cooling.
CFDL and sand are shown in Table 1. For
Crystal size crystals evaluation, a 15-cm line anchored with stan-
(W) (no./g) dards at equivalent lengths is used for presenta-
6 8 x lo8 tion to panelists. Three samples of Dulce de
15 9 x 106 Leche with different sizes of lactose crystals
20 5 x 104 were measured objectively and also using the
sensory scale. Results are in Table 4. Close
correspondence is observed, confirming validity
of scale. Analysis of variance of results from
h l c e de Leche was placed on tongue, pressed sand concentration variations (see Table 2) did
against the palate, and size of crystals as per- not show significant (F5.05) differences in sen-
ceived between the tongue and palate was eval- sory evaluation by the panel.
uated. From these results a reproducible sensory
The sensory panel concluded that use of sandiness scale was developed, which is equiv-
sand in replacement of lactose crystals to stim- alent to microscopic measurement of crystal
ulate sandiness produced an equivalent re- size.
sponse. Similarities in crystal appearance are
shown in Figure 1. A t test on sensory evalua-
tion means of these samples (using the 10 cm Sandiness Threshold
unstructured scale) showed equivalent sandi- Experiment I . Sandiness threshold was
ness. found to be between the sample of CFDL
Results from magnitude estimation were mixed with 1.5% microcrystalline lactose and
correlated according to Steven's law (13): the sample of CFDL mixed with 1.5% lactose
crystals 45 pm in size. Therefore, sandiness
s = C" will not be perceived at or below the crystal
where S = perceived sandiness, C = crystal sizes and numbers shown in Table 3.
size, and n = exponent of Steven's law. Experiment 2. Sensory threshold was be-
In Equation [l], n is the parameter of inter- tween the sample of CFDL mixed with . l %
est, because it indicates whether variations in lactose crystals 45 pm in size and the sample of
crystal size are amplified or diminished through CFDL mixed with .25% lactose crystals of the
sensory evaluation. In this case, n was 1.08, same size. The number of 45-pm crystals in
with its 95% confidence limits (6) including 1. concentrations of .1 and .25% were measured
Thus, sensory versus objective evaluation of microscopically. Results were 1.2 x 16 and 3.1
sandiness can be considered linear. Current lit- x I@ crystals/g, respectively. Thus, for crystals
erature (3, 13) offers no comparative values for of 45 pm or smaller, threshold will be between
n; the closest comparison would be visual ob- these numbers of crystals.
servation of length (3) with n = 1. Experiment 3. As in experiment 2, only the
Considering n = 1, a sensory scale of sandi- lowest concentration was below threshold.
ness was developed covering the range of crys- Numbers of 105-pm crystals in concentrations

TABLE 6. Sensory threshold perception of sandiness according to number and size of crystals.
Size
crystals 105 p 45 p 15 P 6Pn
WJg)
Between 0 and 4 x IO3 NQ No No No
Between 4 x Id and 1 6 Yes No No No
Between 16 and IO7 Unreal Yes No No
Between lo7 and IO9 unreal UIUeal ? No

Journal of Dairy Science Vol. 73, No. 3, 1990


610 HOUGH ET AL.

TABLE 7. Maximum number of lactose crystals present in (7) indicated that the longest axis of this crystal
Dulce de Leche, according to size. is 1.33 times the diameter of an equal volume
crystal Maximum sphere. With the volume of a single crystal, and
size crystals the volume occupied by lactose in totally crys-
(w) (no&) tallized Dulce de Leche, the maximum number
6 1.3 x 109 of crystals per gram were calculated. Results
15 9.2 x 10’ are shown in Table 7.
45 3.1 x 106 Hunziker’s threshold for sandiness in swee-
105 2.4 x 16 tened condensed milk is more than 4 x lo8
crystals/cm3 (7). This is within the order of
values from Tables 6 and 7, considering Dulce
de Leche’s density approximately 1.3 s/cm3
of .1 and .25% were 3800 and 8500 crystal& (12).
respectively. Thus, for 105 pm or smaller,
threshold will be between these number of
ACKNOWLEDGMENTS
crystals.
Experiment 4. Using the sensory sandiness Authors acknowledge financial support of
scale described, the mean score for Duke de the Argentine Scientific and Technological Sec-
Leche seeded with microcrystalline lactose was retariat (SECYT) and Buenos Aires Province
.22 (scale from 0 to 15) with a standard devia- Research Committee (CIC).
tion of .54. This is not significantly (b.05)
different from 0, i.e., no sandiness present.
REFERENCES
Confirming this result, the paired difference
tests showed no significant (b.05)difference 1 Buyze, H. G. 1952. Seed-lactose for the manufacture of
between seeded Dulce de Leche and CFDL. condensed milk. Neth. Milk Dairy J. 6:218.
Results of size and number of crystals mea- 2Castelao. E., 0. Perin, R. Reyna, 0. Sbodio, J. E.
Bacheta. and M. Pauletti. 1978. Duke de leche a partir de
sured microscopically for seeded Dulce de leche hidrolisada-ensayo previo. Heladeria-confiteria
Leche are in Table 5 . Thus, below values Latinoam. 517.
shown in Table 5 , sandiness will not be per- 3 Guirao, M. 1980. Page 69 in Los sentidos, bases de la
ceived. percepcion. Alhambra, Madrid, Spain.
Results indicate that according to the size of 4 Giec. A.. N. Bialuga, and J. Skupin. 1985. Application of
beta-galactosidase preparations in production of con-
the lactose crystals present, the sandiness densed milk. Acta Aliment. Pol. 9:215.
threshold will depend on the number of crys- 5 Hosken, F. S. 1969. Doce de leite - durabilidade e
tals; i.e., there is no single threshold. A sum- cristalizacao. Rev. Inst. Candido Tostes 24:lO.
mary of these results is presented in Table 6. 6Hunter. J. S. 1981. Calibration and the straight line:
This table is useful for quality control of the current statistical practices. J. Asscc. Offic. Anal. Chem.
64574.
sandiness defect: the size and number of crys- 7 Hunziker, 0.F. 1946. Pages 164-165 in Condensed milk
tals of a sample can be measured microscopi- and milk powder. Publ. by author, La Grange, IL.
cally, and values from the table will indicate 8 Jellinek, G. 1985. Sensory evaludon of food, theory and
whether sandiness will be perceived or not. In practice. Ellis Horwood Ltd., Chichester, Engl.
9 Kisza, J., J. Switka, A. Kruk, and A. Surazynski. 1973.
Table 6, for certain sizes, it is indicated when E& d’utilisation de la &D-galactosidase pour la fabri-
the number of crystals would probably not oc- cation du lait condense sucre. Le Lait (527):430.
cur Le., when more lactose crystallized is indi- 10 Lundahl, D. S . . 8 . K. Lukes. M. R. MeDaniel, and L. A.
cated than is normally present in Dulce de Henderson. 1986. A semi-ascending paired difference
Leche. The maximum number of crystals for a method for determining sensory thresholds of added
given size were calculated as follows: subsmces to background media. J. Sensory Stud. 1 :291.
11 Mori, E. E.. I. Shirose, and M. C. Junqueira. 1984. 0.
As mentioned in the introduction, lactose metodo de estimativa de grandeza: revisao e aplicacao.
concentration in Dulce de Leche is 9.85 dl00 g Bol. Ital. 21991.
of Dulce de Leche. Lactose density is 1.54 g/ 12 Moro, 0.. and G. Hough. 1985. Total solids and density
cm3 (20), therefore, supposing it is completely measurements of Dulce de Leche, a typical Argentine
dairy product. J. Dairy Sci. 68521.
crystallized, it occupies .0639 cm3/g of Dulce 13 Motowitz, H.R. 1977. Magnitude estimation: notes on
de Leche. To calculate the volume of a single what, how, when, and why to use it. I. Food Qual. 3:195.
tomahawk-shaped a-lactose crystal, Hunziker 14 Nickerson, T. A., and E. E. Moore. 1972. Solubility

Journal of Dairy Science Vol. 73, No. 3. 1990


MEASUREMENT OF SANDINESS IN DULCE DE LECHE 61 1
interrelationsof lactose and sucrose. J. Food Sci. 3750. Ital. 14:61.
15 Roeunan,K. 1981.Methodsforthequantitativedetermi- 18 Silva, T. J., A. J. Pinheiro, P.T. Coelho, A. S. Pereira and
nation of crystalline lactose in milk products. Neth. Milk J. B. Chaves. 1984. Utilizacao da beta-D-galactosidase
Dairy J. 35:l. no process0 continuo de fabricacao de doce de kite
16Sabioni. J. G . . A. J. Pinheiro, D. 0. Silva, J. B. Chaves, homogeneizado. Rev. Inst. Candido Tostes 3919.
and A. C. Borges. 1984. Control of lactosecrystallization 19 Valle. J. L., I. B. Figueiredo. 1980. Utilizacao de beta-
in Duke de Leche by beta-Dgalactosidase activity from galactosidase para prevenir a cristalizacao do doce de
permeabilizcd Kluyveromyces lacfis cells. J. Dairy Sci. leite. Bol. Ital 17:309.
67:2210. 20 Webb, B., A. Johnson. and J. Alford. 1980. Ch. 6 in
17 Santos, D. M.. J. F. Martins, and N.C. dos Santos. 1977. Fundamentals of dairy chemistry. AVI h b l . Co., West-
Arenosidade e ouuos porblemas do doce de leite. Bol. pn, m.

Journal of Dairy Science Vol. 73, No. 3, 1990

You might also like