Abstract
A series of physical softness models were developed for facial tissue products. To this end, subjective softness data were obtained by panelists by means of round-robin pair-comparison methods. Overall softness was found to mainly consist of bulk and surface softness. Bulk softness was determined by measuring the tensile modulus (TM) from tensile testing. In contrast, the surface softness considered the mean absolute deviation (RMAD) from the roughness average (Ra) and the mean absolute deviation (FMAD) from the average coefficient of friction ( ), respectively, which were determined by profilometry. The developed models exhibited strong correlations with subjective softness. In particular, surface softness was found to contribute more to the overall softness than bulk softness. Overall, the developed models can serve as guidance for developing tissue products.
Download PDF
Full Article
Developing Physical Softness Models for Facial Tissue Products
Ji Min Lee,a Young Chan Ko,a Byoung Geun Moon,b Yong Ju Lee,a Soon Wan Kweon,a and Hyoung Jin Kim a,*
A series of physical softness models were developed for facial tissue products. To this end, subjective softness data were obtained by panelists by means of round-robin pair-comparison methods. Overall softness was found to mainly consist of bulk and surface softness. Bulk softness was determined by measuring the tensile modulus (TM) from tensile testing. In contrast, the surface softness considered the mean absolute deviation (RMAD) from the roughness average (Ra) and the mean absolute deviation (FMAD) from the average coefficient of friction (), respectively, which were determined by profilometry. The developed models exhibited strong correlations with subjective softness. In particular, surface softness was found to contribute more to the overall softness than bulk softness. Overall, the developed models can serve as guidance for developing tissue products.
DOI: 10.15376/biores.19.1.116-133
Keywords: Facial tissue; Physical softness model; Pair-comparison test; Interval scale value (ISV); Bulk softness; Surface softness; Tensile modulus (TM); Surface roughness (Ra and RMAD); Surface friction ( and FMAD)
Contact information: a: Department of Forest Products and Biotechnology, Kookmin University, 77 Jeongneung-ro, Seongbuk-gu, Seoul 02707 Republic of Korea; b: Consumer Product Division Packaging Technology Center, Korea Conformity Laboratories, 199, Gasan Digital 1-ro, Geumcheon-gu, Seoul, 08503, Republic of Korea; *Corresponding author: hyjikim@kookmin.ac.kr
INTRODUCTION
Strength, softness, and absorbency are key attributes that consumers consider in hygiene paper products, such as bathroom tissues, facial tissues, and paper towels. However, it is critically important to realize that these attributes are subjective, which can be quite different from objective or physical properties. These may even be contradictory, resulting in mistakes when evaluating the in-use properties of the products.
To avoid making such mistakes while measuring the physical properties, various hygiene paper producers still heavily rely on subjective evaluation, especially sensory panel tests (SPTs) with well-trained panelists under controlled environments (Hollmark and Ampulski 2004; Rosen et al. 2014; Ko et al. 2015; 2016; 2017a; Lee et al. 2017a; Park 2017). Nevertheless, there have been continuous efforts to develop the physical test methods which can be used for predicting the subjective attributes because their benefits are too great to be ignored (Ko et al. 2015; 2016; Pawlak et al. 2022). Table 1 lists some of the potential benefits of physical test methods.
Softness is considered the most subjective property for hygiene papers, among the attributes of strength, softness, and absorbency. In particular, softness is extremely difficult to define because several factors may contribute to the softness attributes. It may even be impossible to unequivocally define softness. Since the pioneering work by Hollmark, steady progress on developing physical softness models of hygiene paper products has been made (Hollmark 1976; 1983a; 1983b). An excellent and thorough review on this subject has been done by Ramasubramanian (2002). Currently, it is generally accepted that subjective softness mainly consists of two components, namely bulk- and surface-softness (Ampulski et al. 1991; Beuther et al. 2012; Ko et al. 2017a; Pawlak et al. 2022).
Table 1. Benefits of Physical Test Methods
Ko et al. (2015; 2017a; 2018) developed several softness models for bathroom tissue products. To this end, the tensile modulus (TM), which is defined as the slope between two selected points in a load–elongation curve, has been determined for bulk softness. Meanwhile, surface roughness and friction have been determined using a surface profilometer for surface softness (Ko et al. 2015; 2017a; 2018). Friction alone was established to be sufficient for bathroom tissue products.
To date, existing works on the softness of hygiene paper have mainly been limited to bathroom tissue products. Limited works have been available on other hygiene paper products, such as facial tissues, paper towels, and napkins. Among these products, softness is believed to be especially important to facial tissue products.
As such, the main objective of this study is to develop physical softness models of facial tissue products following the same principles which have been used for developing the models for bathroom tissue products. Another objective is to demonstrate the use of physical softness models as a guidance for product development.
Test Methods of Subjective Attributes
Reliable and reproducible data of the in-use properties from users is a prerequisite in developing physical models to predict subjective attributes. In particular, the physical properties should be relevant to the in-use properties.
Fig. 1. Relevance vs. reproducibility and variability of different test methods
As subjective test methods to obtain the in-use properties, the home-use tests (HUTs), the Central Location Tests (CLTs), the Building Panel Tests (BPTs), and the sensory panel tests (SPTs) have been used (Lee et al. 2017a; Ko et al. 2018).
Figure 1 shows the relevance of each test method against its reproducibility and variability. HUTs have the highest relevance for obtaining in-use properties; however, this approach provides the largest variability and lowest reproducibility. The results of physical methods are in contrast to those of HUT, i.e., physical methods have the lowest relevance but the lowest variability and highest reproducibility.
In practice, among subjective tests, SPTs have been the most widely used because it is the cheapest and the easiest practice to collect data from well-trained panelists. In this method, a panelist acts like a human-machine under a highly controlled environment to obtain good correlation with physical tests.
Data Acquisition Methods for the Subjective Tests
Once the subjective evaluation method was chosen, the data acquisition method from the rating, ranking order, and pair-comparison was selected. The advantages and disadvantages of each method have been discussed in a previous study (Lee et al. 2017a; Ko et al. 2018).
The Thurstone Interval Scale
In developing physical models for predicting subjective evaluation, it is critically important that the subjective evaluation value should be linear and continuous on equal intervals, as is the case with physical measurements, such as length, weight, and temperature. However, the numbers from the ranking order and rating are not linear and continuous on an equal scale (Ko et al. 2017a; Lee et al. 2017a).
Thurstone (1927) introduced the interval scale value (ISV), which is determined using pair-comparison testing. Percentage data is converted to ISV, which is linear and continuous on equal intervals. Statistically speaking, it is the Z-value from normal deviates (Lee et al. 2017a, Ko et al. 2018). ISV performs like a physical measurement. Table 2 shows the P, %- preference vs. ISV (Z-Value). It shows that ISV is linear and continuous, unlike P, %-preference.
Table 2. P, %-preference vs. ISV (Z-value) in a Pair-comparison Test
The Round-Robin Pair-Comparison Tests
Although the Thurstone interval scale theory constitutes a method for obtaining ISV using pair-comparison tests, it does not show for the case where more than two products are to be pair-compared.
When more than two-products are to be pair-compared, a full round-robin pair-comparison test is necessary (Ko et al. 2017a). As a numerical illustration, when five samples are compared (n = 5), 10 pair-comparison tests would be required, according to Eq. 1,
where N is the number of pair-comparison tests required; and n is the number of test products.
However, there are two major problems in full round-robin pair-comparison tests. First, N increases rapidly as n increases, according to Eq. 1. For example, when n = 10, N is 45. Such a large number of tests may be time consuming and impractical.
Second, ISV is extremely sensitive to small changes in the %-preference for P values that are lower than 10 or higher than 90, as shown in Fig. 2. This indicates that it is difficult to obtain reliable ISVs from a pair of products whose softness attributes are widely different. Nonetheless, pair-comparison testing is extremely effective in obtaining reliable ISVs from a pair of products whose attributes are very close.
Fig. 2. ISV (Z-value) vs. P, %-preference
The Subgrouping Round-Robin Pair-Comparison Method (The Ko-Method)
To solve the above problems with the full round-robin pair-comparison, Ko et al. (2017a) have suggested that products should be divided into subgroups containing fewer test samples. This method is referred to as the Ko-method for convenience.
For n = 10, each panel should be tested with N = 45 for a full round-robin pair-comparison 
Full Round-Robin Pair-Comparison: Sample 1~10
However, if the full group is divided to two subgroups;
Subgroup One: Sample 1, 2, 3, 4, 5
Subgroup Two: Sample 5, 6, 7, 8, 9, 10
It is noted that Sample 5 appears in both subgroups as a linker. Such a sample linking two subgroups is referred to as an anchor sample. In this case, the number of pair-comparison tests by two subgroups required;
To compare with the full round-robin pair–comparison tests which require N = 45, the subgrouping method requires N = 25, which is a reduction of approximately 45% from the full-robin pair-comparison tests (N = 45), thereby addressing its first problem.
The second problem of the full round-robin method is the high sensitivity of ISV to small changes in P, % preference. This can be addressed by avoiding the pairing of two products where P is lower than 20 or higher than 80, as shown in Fig. 2.
Thus, the Ko-method can solve the problems of full round-robin pair-comparison method and provide reliable ISV from a pair of two products whose attributes are very close.
The Softness Models of Hygiene Paper Products
Hollmark’s pioneering work on tissue softness in the early 1980s established two components for the subjective softness of tissue, namely bulk- and surface-softness (Hollmark 1983a,b; Ko et al. 2017a).
For surface softness, surface roughness and friction have been determined (Kawabata and Niwa 1989; Yokura et al. 2002; Hollmark and Ampulski 2004; Ko et al. 2015; 2017a; 2018).
Tensile Properties for Bulk Softness Determination
Since Ko et al. (2017a), it has also been accepted that bulk softness should be determined using TM from tensile testing (Hollmark 1983a; 1983b; Habeger et al. 1989; Ampulski et al. 1991; Spendel 1990; Harper et al. 2002; Dwiggins et al. 2003; Beuther et al. 2012; Ko et al. 2015; 2017a; 2018).
TM is the slope between two points in a load-elongation curve (Ko et al. 2015). ISO 12625-4 (2022) defines the slope between two force points at 2 and 22 N/m from a tensile force–elongation curve, as calculated according to Eq. 2 (unit: N/m),
where TM is the tensile modulus (N/m); l and w are the gauge length (100 mm) and initial width of the test piece (50 mm), respectively; f1 and f2 are the forces closest to 2.0 and 22.0 N/m, respectively; and e1 and e2 are the elongation distances (mm) closest to 2.0 and 22.0 N/m, respectively (ISO 12625-4 2022).
The difference between tensile stiffness and TM is shown in Fig. 3. Tensile stiffness is the initial slope from the origin, whereas TM is the slope between two specified points (e1 and e2). The Young’s modulus can be calculated according to Eq. 3.
Young’s modulus = Tensile stiffness / Thickness (3)
Fig. 3. Tensile stiffness and tensile modulus in a tensile force–elongation curve
As the tensile force–elongation curve of tissue products is not linear, it is extremely difficult, if not impossible, to determine tensile stiffness. Tensile strength (TS) is defined as the maximum tensile force per unit width that a test piece can withstand before breaking (ISO 12625-4 2022). The tensile force at the elongation at break is calculated according to Eq. 4, as follows,
where TS is the tensile strength (N/m); 
is the maximum tensile force (N); and w is the initial width (mm) of the test piece (ISO 12625-4 2022).
Surface Roughness and Friction for the Surface Softness Determination
A stylus-contact type surface tester has been used to determine both surface roughness and friction of bathroom tissue products (Spendel 1990; Ampulski et al. 1991; Dwiggins et al. 2003; Harper et al. 2002; Yokura et al. 2004; Beuther et al. 2012; Park 2017; Ko et al. 2018; Moon 2021; Lee et al. 2023). ISO 12625-18 (2022) presents the determination of the friction of bathroom tissue products. However, to date, few works on the softness of facial tissue products have been available.
Among commercially available surface testers, only the Kawabata surface tester (Model: KES-SESRU, Kato Tech, Kyoto, Japan) can simultaneously measure both the surface roughness and friction from the same scan lines (Kato Tech 2018a,b). Recently, this tester has been successfully used for bathroom tissue products (Ko et al. 2017a,b; 2018; Lee et al. 2017b; Park 2017; Moon 2021; Lee et al. 2023). As such, the Kawabata surface tester was used to determine both the surface roughness and friction of facial tissue samples in this study.
Surface Roughness Determination
Figure 4 shows the surface roughness profiles of a facial tissue sample (FT2) in the machine direction (MD) and cross-to-the machine direction (CD), as obtained using the Kawabata KES surface tester (Model: KES-SESRU, Kato Tech Co., Ltd., Kyoto, Japan), as an illustration.
Fig. 4. Surface roughness profiles of a facial tissue sample (FT2) in the (a) MD and (b) CD
From such a surface roughness profile in Fig. 4, the average of surface roughness (Ra) and the mean absolute deviation (RMAD) from Ra are calculated according to Eqs. 5 and 6, respectively:
(5)
(6)
where Ra is the roughness average (μm); Ri is the roughness (μm) at scanning point i; and N is the number of data points in the scan length.
Surface Friction Determination
Figure 5 shows the surface friction profiles of a facial tissue sample (FT2) in the MD and CD, which were obtained simultaneously with the roughness profile in Fig. 4.
Fig. 5. Surface friction profiles of a facial tissue sample (FT2) in the (a) MD and (b) CD
A recently published ISO standard (ISO 12625-18 2022) outlines the determination of the friction of tissue products. The average coefficient of friction (COF, ) and the mean absolute deviation (FMAD) from are calculated according to Eqs. 7 and 8, respectively (ISO 12625-18 2022):
(7)
(8)
where 
is the average COF; N is the number of data points from the scan length; 
is the COF at point i; and FMAD is the mean absolute deviation from .
EXPERIMENTAL
Subjective Softness Evaluation
To obtain the subjective softness of seven facial tissue samples in an interval scale, the previously mentioned subgrouping pair-comparison method was used (Ko et al. 2015; 2017a; 2018). In the present study, it is referred to as the Ko-method for convenience.
Prior to grouping the seven samples into two subgroups, a ranking method was used to select the anchor sample. As a result, FT4 was ranked in the middle and selected as the anchor sample. Table 3 shows the subgrouping pairs of SG1 and SG2. After subgrouping, each pair-comparison test was performed by 100 untrained panelists (Ko et al. 2017a; 2018; Park 2017).
Table 3. The Subgrouping Pairs
Facial Tissue Samples
Table 4 lists seven commercial two-ply facial tissue samples, and all physical properties were measured for not one-ply, but two-ply samples. The samples were conditioned at a temperature of 23 ± 1 °C and relative humidity of 50 ± 2% for more than 48 h, according to ISO standard 187 (1990).
The optical photographs of the samples were taken using Optitopo surface deviation (OSD, L&W, Sweden). As all products were two plies, there was no pattern difference in the optical photographs between the top and bottom layers of each sample. Accordingly, only the top layer of the sample was used to determine the surface roughness and friction. Figure 6 shows the top layers of the seven facial tissue samples with the arrows indicating the MD.
Table 4. Properties of the Facial Tissue Samples
Fig. 6. Optical photographs of the facial tissue samples
Tensile Modulus Determination
A tensile tester from MTS, Eden Prairie, Minneapolis, USA (Criterion® Model 41) was used under the following conditions according to ISO 12625-4 (2022): sample length was 150 mm; sample width was 50 mm; load cell was 50 N; span length was 100 mm; and strain rate was 50%/min. TM was calculated according to Eq. 2. The samples were measured 10 times along the MD and CD.
Surface Roughness and Friction Testing
The surface roughness and friction of the samples were simultaneously measured using a Kawabata surface tester (Model: KES-SESRU, Kato Tech, Kyoto, Japan). Figure 7 shows the configuration of the Kawabata surface tester with a U-tube stylus.
Fig. 7. Configuration of the Kawabata surface tester with a U-shaped stylus
The sample plate, instead of the stylus, moves during measurement. The testing conditions were as follows: contact force was 5 g (force); scan length was 20 mm; scan speed was 1 mm/s; and data acquisition rate was 10 Hz (or 10 point/s). The gravitational force (gf) is the mass in grams multiplied by the gravitational constant, g. For each sample, 10 measurements were taken in the MD and CD.
RESULTS AND DISCUSSION
The Subjective Softness Values (ISV)
Table 5 shows the ISV Softness of SG1 and SG2 and final ISV Softness obtained by linking two subgroups with FT4 as the anchor. The anchor sample, FT4, has the softness ISV of 0.00 in SG1, but 0.53 in SG2 and final. The final ISV Softness was used for developing the facial tissue softness models. Among the seven facial tissue samples, FT1 has the highest final ISV of 1.69, indicating it is rated softer than FT7 by 95% and FT4 by 88%, as shown in Fig. 2.
Table 5. ISV Softness of Seven Facial Tissue Samples
Tensile Properties
Table 6 summarizes the tensile properties of the seven facial tissue samples. TM was calculated according to Eq. 2, and TS and elongation at break % were determined according to ISO 12625-4 (2022). The geometric mean (GM) was calculated as the square root of the product of the MD and CD values.
The Facial Tissue Softness Models
Table 7 summarizes the ISV softness and physical properties used for developing the softness models. Ra, RMAD, , and FMAD were calculated according to Eqs. 5 through 8, respectively.
After intensive review, Ramasubramanian (2002) concluded that a power-law model may be applicable to the softness models for tissue products, that is,
(9)
where X is the bulk softness; Y is the surface softness; and c, m, and n are the curve fitting coefficients. The power-law model can be linearized by taking the logarithm of each value (Ko et al. 2018). Consequently, a series of facial tissue softness models have been developed. For three-parameter (3-P) Models, Eq. 10 was used,
(10)
where C, m, n, and l are the curve fitting coefficients (Ko et al. 2018). For two-parameter (2-P) Models, either Eq. 11 or Eq. 12 was used.
(11)
(12)
Table 6. Tensile Properties of the Seven Facial Tissue Samples
Table 7. Softness ISV and Physical Properties of the Seven Facial Tissue Samples
The results obtained by the regression analysis were as follows:
Discussions of Facial Tissue Softness Models
Once a physical softness model is developed, a Predicted ISV Softness can be calculated by substituting the physical parameters to the model. It is noted that all the above models (Eq. 13 and 15 to 21) have constants with negative signs for each parameter. This indicates that the subjective softness should be inversely related to the TM, surface roughness, and friction. As such, in addition to the low R2 values, Eq. 14 should be excluded from the models.
Thus, in developing the softness models, it is most important that they should be technically sound. Simply making a judgment based on the R2 value only can lead to a faulty conclusion. Equations 13, 15, 16, and 18 have shown relatively high correlation with the ISV Softness. Table 8 shows the results of the ISV Softness and Predicted ISV Softness calculated from these equations.
Table 8. ISV Softness and Predicted ISV Softness using the 3-P (TM, RMAD, and FMAD) and 2-P (TM and RMAD) Models