You are on page 1of 13

721

Measuring and Interpreting Low-Stress Fabric Mechanical


and Surface Properties
Part IV: Subjective Evaluation of Fabric Handle

T. J. MAHAR
1 AND R. POSTLE
School of Fibre Science and Technology. The University of New South Wales.
Kensington, NSW 2033, Australia
ABSTRACT
An investigation is presented of the levels of commonality held by panels of judges
from different countries in assessing fabric handle for both summer and winter men’s
suiting materials. The sets of 214 winter and 156 summer fabrics used to obtain the
subjective assessments and the KESF measurements of low-stress mechanical and
surface properties for men’s suiting fabrics form the basis of this investigation. There
is a common notion of the handle of men’s suiting fabrics, and superimposed on this
notion are subtle national preferences and individual differences within national panels.
In the case of summer fabrics, there are two notions of fabric handle: one held in
common by panels from Japan and the People’s Republic of China, and another held
by national panels from Australia, New Zealand, India, the U.S.A., and Hong Kong
and Taiwan. Individual differences between judges within a panel imply that for winter
fabrics, the mean handle assessment of a national panel can be predicted by the judge
whose assessments agree most strongly with the mean assessment of his national panel
to only a moderate level of precision; i.e., only fabrics with differences in mean handle
rating of greater than 1.6 units on a scale from 0.0 ( for unsatisfactory handle) to 5.0
(excellent handle) can be distinguished reliably. The equivalent value for the summer
fabrics is approximately 3 handle grading units (on a scale of1 to 5), based on the
"Western" notion of summer fabric handle held commonly by the six countries lis-
ted here.

Definition and Concept of Fabric Handle there exists no accepted analytical definition of fabric
handle, as Bona noted [4, 5 ), progress has been made
Fabric handle has been defined [ 21 as &dquo;... the toward understanding the mechanism by which handle
subjective assessment of a textile obtained from the is measured. In Part III of this series of papers ( 17 J,
sense of touch. It is concerned with the subjective
we reported the results of nonlinear optimization of
judgement of roughness, smoothness, harshness, pli- fabric mechanical and surface properties from the
ability, thickness, etc.&dquo; Judgments of fabric handle are handle of suiting materials. In the first two papers of
used as a basis for evaluating quality, and thus for de-
the series [ 7, t4]. we reported the levels of precision
termining fabric value, both within the textile, clothing, obtainable from instrumental measurements of fabric
and related industries and by the ultimate consumer.
low-stress mechanical and surface properties and pre-
Studies of fabric handle may be of major commercial sented the effects on these properties of finishing.
significance if they, for example, assist in explaining pressing, and other treatments.
handle assessment or provide a means of its estimation The work presented in this paper reports the levels
based on objective measurement. of commonality held by panels of judges from different
It is necessary to examine the subjective assessment countries in assessing fabric handle for both summer
of handle before examining its relationship to fabric
and winter men’s suiting materials. Details of this in-
mechanical and surface properties. Despite the fact that ternational survey. of handle assessments are included
in the appendix to this paper. Sources of variability are
isolated for fabric handle assessments made in different
1
Current address: CSIRO Division of Wool Technology, P.O. Box countries, and values are provided for the variability
7, Ryde, NSW 2112, Australia. associated with these subjective assessments.

Downloaded from trj.sagepub.com at LSU Libraries on May 8, 2015


722

SUBJECTIVE ASSESSMENT surface properties, and that the final judgement of


handle is based on the suitability of the mechanical
The subjective assessment of fabric handle has been
and surface properties for the particular end use of the
the subject of scientific study over an extended period.
fabric. This practical approach to handle analysis led
In 1926, Binns [ 3 published a psychologically based
to the formation in 1972 of the Hand Evaluation and
analysis of the assessment of fabric handle. This paper Standardisation committee (HESC) of the Textile
and other .work by him in this field [ 1, 2 ] provided an
Machinery Society of Japan. The HESC has proven to
early analysis of the mechanism by which judges seem be a very active committee, and has provided expert
to assess handle. Binns concluded that the judgment
of a number of experienced persons gives a more re- judges who have assessed the handle of literally thou-

sands of fabrics [ 10 ] .
liable and sensitive grading of fabrics than is possible
from an individual judge’s grading.
SUMMARY
Howorth and Oliver [ 9 ] made a major contribution
to the analysis of fabric handle assessments. when they The abilities to specify what determines fabric handle
first applied the factor analysis technique developed by and to control or optimize handle for particular mar-
Thurstone [22] in an attempt to identify underlying kets based on objective criteria, i.e., low-stress me-
interrelationships in the handle assessments of a range chanical and surface properties of fabrics, are of major
of suiting, lingerie, and dress fabrics. These analyses technical and commercial significance. But no rigorous
isolated three factors responsible for handle, which the analytical definition exists, a priori, for fabric handle
authors identified as smoothness, stiffness, and bulk, [4, 5 ]. In attempting to develop a system for estimating
the last of these being related to fabric weight and or specifying handle, a textile scientist should recognize
thickness. This represented an important step toward from the general definition of handle [ 21 ] that the psy-
isolating a set of primary fabric hand qualities that chological element in its assessment implies there may
might be used to describe handle preferences. legitimately be differences between individuals. It. is
A number of authors, notably Lundgren [ 13 and therefore necessary to establish the existence of, and
Matsuo et al. [ 16] have attempted to model the as- to quantify if possible, a unified concept of fabric han-
sessment of fabric handle based on psychophysical dle among judges before examining techniques for its
concepts. Lundgren used concepts from both decision estimation, e.g., by studying relationships between
theory and information theory and proposed the ex- handle and mechanical and surface properties. It is ex-
istence of four sensory centers corresponding to the tremely unlikely that one concept of handle will apply
four factors that could be isolated from the fabric anal- for all fabric types, since the requirements of handle
ysis of Howorth and Oliver: (a) smoothness, (b) stiff- will vary according to market demands.
ness, ( c ) bulk properties, and (d) warmth. This theory
is analogous to the Young-Helmholtz theory [ 8 ] , which
proposes three color receptors in the eye, one each for An International Survey
red, green, and blue, for perceiving color.
BACKGROUND
Matsuo [ 16 ] applied to the analysis of fabric handle
a variation of the Weber-Fechner law, the so-called The Hand Evaluation and Standardisation Com-
&dquo;only valid law in psychophysics&dquo; [ 19 ] . Matsuo as- mittee of the Textile Machinery Society of Japan [11]]
sumed that this law held when fabric mechanical prop- has been responsible for an extensive study within Ja-
erties were used as the handle stimuli. This model uses pan relating fabric mechanical and surface properties
a nonlinear combination of mechanical properties to to handle assessments. The availability of the fabrics
explain fabric handle assessments. In practice, the re- used in the HESC study facilitated our extension of
sults obtained from this model depend strongly on the this work to countries outside Japan, in order to es-
values assigned to &dquo;minimum sensibility,&dquo; which a tablish the universality of the definition and concept
judge can discriminate for each mechanical property. of fabric handle held by expert judges from different
Estimation of these critical parameters appears to be countries and different cultural backgrounds, whose
somewhat unreliable [ 16 ] . products must satisfy different climatic conditions. The
Kawabata [ I I ] undertook the analysis of assessments sets of 214 winter and 156 summer fabrics used to
by experts from the Japanese textile industry of the obtain the subjective assessments and the KESF mea-
handle of men’s suiting fabrics. This analysis was based surements of mechanical and surface properties for
on two hypotheses: that handle assessments are based men’s suitings [ 1] ] fonn the basis of the study we report
on tactile sensations caused by fabric mechanical and in this paper.

Downloaded from trj.sagepub.com at LSU Libraries on May 8, 2015


723

ORGANIZATION OF THE SURVEY TABLE II. Mean within-group correlations with the mean fabric
handk assessment of the panels..
An international survey was undertaken of the fabric
handle preferences of judges drawn from the textile,
clothing, and related industries in seven countries, four
of which were represented in the interlaboratory trial
of the KESF instrumentation reported in Part I of this
series of papers [ 14 ] : Japan, Australia, New Zealand,
and the People’s Republic of China (P.R.C.). The re-
maining participating countries were India, the United
States, and Hong Kong/Taiwan.2
Expert judges representing each national group in-
dependently assessed the handle of each fabric accord-
ing to their own definitions of fabric handle and the
grading scale shown in Table I. For the analysis of the
survey results, the gradings were assumed to belong to
a continuous scale; indeed some of the judges graded
National Panels
the handle of some fabrics as &dquo;2.5&dquo; or &dquo;3.5.&dquo; A panel The strong and highly significant correlation coef-
of consumer judges from Australia also assessed the ficients in Table II show that the eight judges within
fabrics. The details of this survey are discussed in the each national panel assessed the handle of the fabrics
appendix to this paper, including a list of the national in a consistent manner. This result indicates that there
coordinators of the survey, the backgrounds of the is a generally agreed concept of handle within each
judging panels, descriptions of the fabrics, and the de- national panel for these fabrics. There appears to be
sign of the survey. more of a consensus between the judges about winter
fabric handle than the handle of the summer fabrics,
despite the far greater number of winter fabrics assessed.
TABLE I. Fabric handle rating scale. This is consistent with the observation made by a
number of the judges that there was greater difficulty
in assessing the summer fabrics.

Consumer Panel
The within-group correlations for Australian
mean
consumer judges are also shown in Table II for com-
parison with the expert judges. While being slightly
lower than for the expert panels, the within-group cor-
WITHIN-GROUP AGREEMENT °
relations for the consumer panel are strongly signifi-
cant : correlation coefficients greater than 0.140 and
The level of agreement on fabric handle assessments 0.175 are significantly different from zero for the winter
within each of the eight (seven expert and one con- and summer fabrics, respectively. The values of the
sumer) panels of judges was established using the fol- correlations in Table II indicate general agreement be-
lowing three-step procedure: (a) the mean fabric handle tween the judges in the consumer group. The lower
assessment was calculated within each panel for each correlation coefficients for the consumer judges com-
fabric, (b) correlation coeflicients were obtained for pared to the expert judges is not unexpected. Binns [ 1 ) ]
the relationship between the handle assessments of each has concluded that untrained judges possess the same
judge and the mean assessment of his (or her) national &dquo;native&dquo; judgment in assessing fabric handle as do ex-
panel, and ( c ) the mean of these correlation coefficients pert judges, but differ from the experts in their lack of
was calculated for each national panel. Mean corre-
appreciation of technical factors in cloth construction.
lation coefficients are shown for each of the eight panels
of judges in Table II for both the winter and summer BETWEEN43ROUP AGREEMENT
fabrics.
In order to establish whether or not there was agree-
2
ment between the seven national panels of judges, we
Owing to the organization of the survey and the patterns of par- calculated correlation coefficients for the relationships
ticipation, Hong Kong and Taiwan are treated as a single national
panel in the analysis of results for this survey. between the mean handle assessments of each pairing

Downloaded from trj.sagepub.com at LSU Libraries on May 8, 2015


724

of the seven national panels. The values of these cor- Summer Fabrics
relations are shown in Table III for winter fabrics and
The situation is quite different when comparing the
Table IV for summer fabrics. The means of these cor-
relations for each panel with each of the other six na-
between-group correlation coefficients for the summer
fabric handle assessments. In this case, there were high
tional panels are shown in Table V.
correlation coefficient values (0.74 to 0.82) for com-
parisons between the Australian, New Zealand, Indian,
TABLE Ill. Correlations between mean handle assessments of and U.S. judges. If we add the Hong Kong/Taiwan
national panels for 214 winter fabrics. panel of judges to this group, there are moderate cor-
relation values for this panel with other panels (from
0.55 to 0.72).
the between-group correlation coefficients involving
the Japanese and Chinese panels on the one hand, and
each of the other national panels on the other, averaged
-0.28 and -0.19 for the Japanese and Chinese panels,
respectively. Values of these correlations ranged from
a low O.13 to negative values as high as -0.40. The

between-group correlation for the Japanese and


Chinese panels was 0.75.
TABLE IV. Correlations between mean fabric handle assessments The analysis of summer fabric handle assessments
of national panels for 156 summer fabrics. indicates that at least two different, and somewhat op-
posite, assessments were made of the handle of men’s
summer fabrics as indicated by the size and signs of
the correlation coefficients in Table IV. One consistent
handle assessment was made in Australia, New Zea-
land, India, the U.S.A. and, to a slighty lesser extent,
in Hong Kong and Taiwan. A second, different, but
again consistent handle assessment was made in Japan
and the P.R.C. There was also a significant level of
direct disagreement between the Japanese/Chinese as-
sessments and those of the other five national panels.
TABLE V. Mean values’ of the between-group correlation
It is unclear whether these two different types of fab-
coefficients for the handle assessments of each national panel with
the other six panels for the winter and summer fabrics. ric handle preference are based on cultural or climatic
differences or some combination of both. Certainly Ja-
pan and parts of China experience annual periods of
hot ( >30°C ) and very humid (>90% relative humidity)
weather. Similar conditions also apply in some coastal
areas of Australia, India, the U.S.A., and Taiwan, as
well as Hong Kong. One might also argue that of the
seven countries investigated, Japan, the P.R.C.,
Hong
*
Kong/Taiwan, and India have maintained their strong
Mean values taken only for the cases where the sign is constant. cultural identities, despite the obvious influence of
Western culture, e.g., the wearing of Western-style suits.
We will investigate further the reasons underlying the
Winter Fabrics ,
different fabric handle preferences for summer suitings
There was a good level of agreement between the in Part V of this series.
mean handle assessments of the national panels for the
Summarizing, the analyses of the between-group
winter fabrics as shown by the high correlation coef- agreement among the seven national judging panels
ficients in Table III. Values of the correlation coeffi- strongly support the notion of an internationally agreed
cients for the Chinese ( P.R.C.) panel (from 0.52 to concept of fabric handle for winter suiting fabrics.
0.71 ) are only moderate compared to the higher values Conversely, the analyses for the summer fabrics show
(from 0.76 to 0.91 ) for all other pairings of the national that there is definitely not one agreed-upon concept of
panels. handle among the national panels for these fabrics. Two

Downloaded from trj.sagepub.com at LSU Libraries on May 8, 2015


725
different and somewhat opposed concepts of fabric underlying interrelationships in both the national pref-
handle have been isolated. erences and those of the individual judges.

Patterns of Agreement Between Judges and


WINTER FABRICS
Between National Panels Table VI shows the factor matrices resulting from a
PRINCIPAL COMPONENT ANALYSIS principal component analysis of the mean handle as-
sessments of the seven national panels for both the
We used principal component analyses [6] sepa- winter and summer fabrics. Each of the national panels
rately on the sets of 214 winter and 156 summer fabrics had a very high factor score ( >0.91 ) on the only factor
in order to investigate patterns in the fabric handle isolated in this analysis for the winter fabrics; the ex-
assessments of the expert judges. In this sort of analysis,
ception was the Chinese panel (factor score 0.75 ). This
a linear transformation is made of a set of data, the
result implies that the mean fabric handle assessments
data matrix, in terms of a new set of variables, called of the seven national panels are mostly determined by
the principal components, such that these principal one common underlying factor.
components are orthogonal. The principal compo-
nents, or factors, are typically ordered in terms of the
decreasing proportion of the variability in the data ma-
trix accounted for by each principal component.

TABLE Vi. Factor scores from principal component analyses of


mean handle assessments of national panels for winter and summer
The value of this sort of analysis depends on whether fabrics.
the majority of the variance in the data is accounted
for by a small number of truly &dquo;principal&dquo; components.
If a small number of principal component factors (in
relation to the number of original variables) accounts
for a large proportion of the variance in the original
variables, these components constitute a valid and
much simplified description of the variability in the
original data, i.e., the original data could be expressed
as a smaller number of independent factors. The anal-
ysis produces a matrix of weightings, called factor
scores, of each parameter on each of the principal
component factors isolated. These weightings can be
interpreted as correlation coefficients between the pa-
rameter and the principal component.
There are two major limitations of principal com-
ponent analysis. Unfortunately, this technique does not Tables VII and VIII summarize the varimax rotated
provide a physical description of the principal com- factor matrices [ 6 J for an analysis of the handle as-
ponents isolated during the analysis. One must interpret sessments of all 56 (=7 panels x 8 judges/pane!) :n-
the pattern of factor scores that constitute a principal dividual judges for both winter fabrics (Table VII ) and
component in order to infer a description of that com- summer fabrics (Table VIII ). For the sake of clarity,
ponent. Second, no acceptable technique is available only the numbers of judges from each national panel
to provide an estimate of the errors applying in this who had factor scores equal to or greater than 0.30 are
sort of analysis. Despite these shortcomings, we have shown in the table, along with their range of factor
been able to use this analysis to isolate underlying re- scores. A more detailed listing of these factor scores is

lationships in our fabric handle assessment data. given elsewhere [ 15 . The analysis for the winter fabrics
Statistical programs are available for performing the isolated six factors from the original 56 judges. which
necessary transformations of the data. For the analyses collectively account for 70.4% of the variance in the
reported here, we used the Statistical Packages for the data.
Social Sciences (SPSS) [ 20 ] suite of programs. The pattern of factor scores for the factor matrix of
There were two separate analyses for each set of fab- individual judges is similar to that shown in Table VI
rics : one was based on the means of the panels and the for the mean assessments of the panels. Most of the
other on the individual judges from each national variance, 52.9%, is explained by one factor on which
panel. This was done to ensure an examination for each judge has a significant ( >0.14 ) factor score, but

Downloaded from trj.sagepub.com at LSU Libraries on May 8, 2015


726

TABLE VII. Range of factor scores for each of the seven national panels on the rotated factor matrix for the handle assessments
of the 214 winter fabrics: 6 factors with eigenvalue > 1.0; factor scores > 0.30.

extra information is available from the analysis of the SUMMER FABRICS


assessments of the individual judges over and above
that obtained from the mean assessments of the na- Inspection of the factor matrix in Table VI reveals
tional panels. This information concerns the pattern two results concerning the handle
mean assessments
of factor scores in factors 2 to 6, inclusive. On an in- of the seven national panels for the summer fabrics.
dividual basis, these five factors account for levels of First, two factors are isolated that explain over 80% of
variance ranging from 8% for factor 2 to 2.0% for the variance in the data. This result is not unexpected
factor 6. from the analysis of the correlation data we discussed
There are clearly secondary or, in the case of the earlier in this paper. The second important and more
Chinese judges, primary fabric handle preferences in- unexpected result is that each of the national panels,
dependent of the major factor (factor 1 ) for some of including the Japanese and P.R.C. judges, has a sig-
the national panels. The pattern of these preferences nificant factor score on factor 1. There are negative
is evident from an inspection of Table VII. The main factor scores for the Japanese and P.R.C. judges, while
features of this pattern are summarized for the fabric scores for the other judges are positive, ranging from
handle assessments of different panels as follows: the 0.71 to 0.92. Although all the national panels consider
eight judges in the Chinese panel are related strongly the same factor, factor 1, in assessing fabric handle,
to factor 2 - mean factor score, 0.72; the eight Japanese the Japanese and P.R.C. judges place an opposite
judges have moderately high scores on factor 3 - mean weighting on the factor compared to the other five
factor score, 0.59; seven of the eight U.S. judges have panels of judges. For factor 2, only the Japanese,
moderate scores on factor 4 - mean factor score, 0.45; Chinese, and Hong Kong/Taiwanese panels had sig-
and five of the eight Hong Kong/Taiwanese judges nificant factor scores (0.74, 0.86, and 0.46, respec-
have moderately high scores on factor 5 - mean, 0.58. tively).

Downloaded from trj.sagepub.com at LSU Libraries on May 8, 2015


727

TABLE VIII. Range of factor scores for each of the seven national panels on the rotated factor matrix for the fabric handle assessments
of the I 56 summer fabrics: !0 factors with eigenvalue > 1.0; factor scores > 0.30.


Proportion of variance explained.

The pattern of factor scores given in Table VIII for handle assessment of a panel of expert judges from a
the fabric handle assessments of the individual judges given country to represent this common notion of
is similar to the pattern we reported above for the mean handle in this study. Some national preferences for
assessments shown in Table VI. Factors 1 and 2 ac- winter fabric handle have been isolated but not de-
count for 32.0% and 14.1 % of the variance in the data, scribed by the principal component analysis procedure.
respectively. The &dquo;secondary&dquo; factors, i.e., factors 3 to Differences between individuals are to be expected
10 inclusive, explain a significant proportion, 20.8%, based on their definition of fabric handle, though the
of the variance in the data. In this case the pattern of extent of these differences varies in the countries
factor scores is not related to the fabric handle pref- studied.
erences of the national panels. Absolute values of the In the case of summer fabrics, we have isolated two
scores in these eight factors are also generally low different bases for assessing fabric handle. Variation in
( <0.40 ) . The less well defined pattern of factor scores the assessments of individual judges within a national
than is the case for the winter fabrics is consistent with panel is greater for the summer fabrics than for the
the lower within-group (national panel) correlation winter fabrics. The subtle preferences in the handle of
coefficients previously noted for these fabrics compared the winter fabrics observed for some national panels
with the winter fabrics. (see Table VII ) are not evident for the summer fabrics.

SUMMARY Variability of Fabric Handle Assessments


commonly held concept
There appears to be a single, The analyses of winter and summer fabrics show
among expert judges of fabric handle for winter suiting that there is no perfect agreement about the handle of
fabrics. Superimposed over this notion of winter fabric these men’s suitings. In this section, we first investigate
handle are differences imposed by national and indi- whether particular fabrics are responsible for very high
vidual preferences. We have chosen the mean fabric or very low variability in the handle assessments of the

Downloaded from trj.sagepub.com at LSU Libraries on May 8, 2015


728

judges. Second, we use the results of an analysis of data given in the appendix and the primary hand value
variance to determine sources of variability within the assessments [ 11 ] .
data and levels of precision in handle assessments.

STANDARD DEVIATIONS WITHIN PANELS TABLE X. Frequency of a fabric occurring in the lowest five or
highest five standard deviations for the handle assessments of the
We have calculated standard deviations of the handle seven national panels.
assessments within each panel for each of the 214 win-
ter and 156 summer fabrics used in the international
survey. The means, minima, and maxima of these
standard deviations for each national panel of judges
are shown in Table IXa for the 214 winter fabrics and
IXb for the 156 summer fabrics. The grand average
standard deviations of all the national panels are 0.77
for the winter fabrics and 0.88 for the summer fabrics,
There was one case where four very smooth, soft,
reflecting the greater within-group agreement previ- flexible fabrics
yet extremely occurred three, four, and
ously noted for the winter fabrics compared to the five times in this analysis. Each of these fabrics, which
summer fabrics.
contain cashmere, is among both the lowest and highest
standard deviations of different panels of judges. A
.
number of judges, particularly from the New Zealand
TABLE IX. The standard deviations of fabric handle assessments
for seven national panels of expert judges.
panel, assessed these fabrics as being of poor handle,
while the majority of judges assessed their handle as
&dquo;above average&dquo; or &dquo;excellent.&dquo; Conflict between some
judges concerning the handle assessments of these fab-
rics may derive from a difference in interpretation of
the term &dquo;handle&dquo; as applied to suitings. There can be
little doubt that these fabrics feel exceptionally smooth,
soft, and flexible compared to the majority in the sur-
vey. Some judges may feel that the poorer expected
wearlife of these fabrics, warrants a &dquo;poor&dquo; handle as-
sessment, whereas other judges accept that these char-
acteristics result in an &dquo;excellent&dquo; fabric handle. How-
ever, Mahar [ 15 showed that fabric durability is not
related to fabric handle assessments in other work. In
Part V, we will provide further discussion of the rela-
tionship between handle preferences and fabric dura-
bility.
Other evidence exists to support the notion that par-
ticular fabrics are responsible for relatively high levels
of disagreement between judges. Kawabata [12] re-
The occurrence of a particular fabric among the ported on 214 winter suiting fabrics as being &dquo;selected
lowest standard deviations of a national panel is an for more precise judgement of (fabric handle) from
indication that this fabric is responsible for good agree- the original 500 samples.&dquo; He also mentioned an in-
ment within that national panel. The reverse applies dependent set of 40 winter suiting fabrics [ 12 ] assessed
for the occurrence of a fabric among the highest stan- for handle by five of the eight Australian expert judges.
dard deviations of a national panel. The number of In this case, there was a lower standard deviation (0.49)
times a given fabric appears among the lowest or highest of handle assessments by the judges than was the case
5 standard deviations of the fabric handle assessments for the 214 fabrics, even allowing for the smaller num-
for the seven national panels is shown in Table X. We ber of judges. The fabrics used in this second handle
investigated the occurrence of particular sets of fabrics assessment survey were chosen based on two charac-
or fabric types being responsible for either the teristics : the full range of handle gradings was repre-
high or
the low standard deviations within a national panel of sented, and there was an excellent agreement amongst
judges. This analysis was based on the fabric description the Japanese panel as to the handle assessments of these

Downloaded from trj.sagepub.com at LSU Libraries on May 8, 2015


729

fabrics. The Japanese handle assessments also agreed There were two analysis of variance studies: one for
strongly with the ratings based on measurements of the winter and the other for the summer fabrics. The
mechanical and surface properties [ 12, 177 ]. A full de- design of the studies included seven national panels
scription of these fabrics, which were published as for the winter fabrics and five panels for the summer
standards of winter fabric handle [ 12 ] , can be obtained fabrics (Japanese and Chinese panels not included),
elsewhere [ 15 . 214 winter and 156 summer fabrics, and two estimates
Despite the occurrence of the four cashmere-style of each national handle assessment. Thus the data space
fabrics among the very high standard deviations for for the winter fabric analysis was
the New Zealand judges, there was no general pattern
of particular fabrics or fabric types responsible for the
(7 national panels) X ( 214 fabrics) X (2 estimates)
high variability of the judges’ handle assessments for and for the summer fabric analysis
either the winter or summer fabrics.
There is a relatively high level of variability within
(5 national panels ) X ( 156 fabrics) X (2 estimate)
the handle assessments of each of the two fabric series We also assumed that there was no systematic re-

surveyed in this analysis. The grand average standard lationship between the standard deviation and the
deviations of handle assessments are 0.77 (winter fab- mean of the fabric handle assessment within each na-

rics) and 0.88 (summer fabrics) on a grading with full tional panel. This assumption is supported by the low
scale variation from 1 to 5.0. correlation (absolute value < 0.33) listed in Table XI
between the mean and the standard deviation of fabric
handle assessments from each national panel. The only
ANALYSIS OAF VARIANCE
exception was the Indian panel’s assessments of the
Model and Assumptions winter fabrics (correlation coefficient -0.67). We =

did not see any obvious nonlinear relationships in the


We made an analysis of variance on the handle as- plots between these means and standard deviations.
sessment data similar to the analysis of the KESF’ data
reported in Part I of this series [ 14 ] in order to establish TABLE Xi. Simple comelation coefficients between the means and
sources of variation and measures of repeatability and standard deviations of the mean fatxic handk assessments of seven
national panels.
reproducibility. This analysis uses the concept of a
mean handle assessment for each national panel rather
than the assessments of a series of individual expert
judges.
The model is based on the results obtained earlier
in this paper. We have assumed that a separate national
assessment exists for each of the seven countries in-
vestigated. We have further assumed that this assess-
ment can be calculated for each country as the mean
assessment of the eight expert judges previously de-
scribed. For the sake of this analysis, we considered Results
the fabric handle assessments of each judge in a na- We found differences in handle assessment signifi-
tional panel as replicates of the assessments of that cant at the 5% level between the seven panels for the
panel, i.e., replicates in this context refer to handle winter fabrics and the five panels for the summer fab-
assessments of the same sample by one of the expert rics. The ( panel X fabric) interaction effects were also
judges from within a national panel, not to assessments significant as we can see from Table XII, which lists a
of the same sample by the same judge at different times. breakdown of the components of variance into within-
Only two handle ratings could be used in this analysis panel variation, ( panel X fabr~c ) interaction, and be-
( i.e., one replicate ) because of the limitations of the tween-panel variation for both the winter and summer
available computing facilities. The two ratings we chose fabrics. The table also includes the grand means and
were the mean handle ratings for each national panel estimates of the 95% repeatability and reproducibility
and the ratings of the judge whose assessments corre- intervals [t 8] for the handle assessments.
lated most highly with the mean assessment of his The greatest source of variation in handle assessment
panel. The handle ratings of this judge provided the occurred between each national panel mean and the
most reliable estimate of the mean assessment of his judge from within that panel as indicated by the rela-
national panel. tively large values of within-panel variation for both

Downloaded from trj.sagepub.com at LSU Libraries on May 8, 2015


730

TABLE XII. Analysis of variance for fabric handle assessments Table XIII presents calculated values P w of the per-
’I ’.
7~ by national panels of judges. cent accuracy (or maximum error) for estimating the
mean handle assessment of one, two, four, and eight
individuals judges from within a panel. The table also
gives similar values Po for estimating panel mean as-
sessments from judges drawn from other panels. These
values are calculated on the same basis as we used in
Part I of this series [ 14 ] for objectively measured fabric
low-stress mechanical and surface property data. Based
on the results for the winter fabrics, eight judges from
a panel are able to estimate the panel mean to an ac-

curacy of 12% at the 95% confidence level. Similarly,


four judges from a panel can estimate the mean as-
sessment of other panels to an accuracy of 17%. For
the summer fabrics, the accuracy is 30% for both the
judges’ panel and other four national panels.

sets of fabrics. This variation represents a significant


proportion, 9.7%, of the grand mean for the winter TABLE XIII. Percentage maximum error in the estimation of fabric
fabrics, and a very large proportion, 46%, for the sum- handle assessments by an increasing number of expert judges.
mer fabrics. It indicates that estimates of the mean as-
sessment of a national panel, even from the judge whose
assessments correlated most strongly
with the panel
mean, are subject to a great deal of variation, especially
for the summer fabrics. By comparison, equivalent
values are smaller for the ( panel X fabric) interaction
variance-7.5%a for winter and$6.5% for summer fab-
rics-and for the between-panel variation-1.8% for
winter and 6.5% for summer fabrics.
The large variation is highlighted in the assessment We also madea second and similar analysis of vari-
of handle for a given fabric between judges within the ance using as replicates the handle assessments of the
national panels when examining values for the re- two judges whose decisions correlated most strongly

peatability of these assessments as shown at the bottom with the mean assessment of their national panel. The
of Table XII. For the winter fabrics, the judge whose results of this analysis, given in Table XIV, support
assessments correlate most highly with the mean as- the breakdown of the sources of variance found in the
sessment of his national panel is able to agree with the
mean assessment of the national panel to within 1.6
TABLE XIV. Analysis of variance for fabric handle assessments-
handle grading units in 95% of cases. The equivalent international panels of judges (using two judges as replicates).
value for the summer fabrics is~proximately 3 handle
grading units (on a scale of 1 to 5), based on the
&dquo;Western&dquo; notion of summer fabric handle held com-
monly by panels from Australia, New Zealand, India,
the U.S.A., and Hong Kong/Taiwan.
These results imply that for winter fabrics, a single
judge in isolation can reliably distinguish between
&dquo;poor&dquo; fabric handle on the one hand and &dquo;above av-
erage&dquo; or &dquo;excellent&dquo; handle on the other, between
&dquo;below average&dquo; and &dquo;excellent&dquo;, and between &dquo;av-
erage&dquo; handle on the one hand and either &dquo;excellent&dquo;
or &dquo;poor&dquo; handle on the other. For the summer
fabrics,
a single judge in isolation can reliably differentiate be-
tween &dquo;poor&dquo; and &dquo;excellent&dquo; fabric handle.

Downloaded from trj.sagepub.com at LSU Libraries on May 8, 2015


731

initial analysis. Greater variability was evident between sider the handle preferences for the summer fabrics.
the assessments of the two judges than between the In this case the analysis discerned two different, and
assessments of one judge and the mean assessment of in some sense opposing, notions of fabric handle: one
the panel. identified in common by the Australian, New Zealand,
Indian, U.S., and Hong Kong/Taiwanese panels, and
SUMMARY a second handle preference for the Japanese and P.R.C.
panels.
Despite their simplifying assumptions, these analyses We conclude that there is a common notion of the
of variance have highlighted the great unreliability of
handle of men’s winter suiting fabrics and that, super-
using the fabric handle assessments of one expert judge imposed over this common notion, are subtle national
as a predictor of the assessments of other judges. We
differences and individual differences within national
have quantified the highest expected improvement in
panels. For the summer fabrics there are two notions
accuracy (lowest reduction in percentage error) for us-
of fabric handle: one held in common by the Japanese
ing the mean assessment of selected numbers of judges and P.R.C.
rather than an individual judge, based on the judge panels and another held by the other five
Individual differences between judges within a
whose assessments are most highly correlated to his panels.
national panel are again superimposed over these two
panel mean. For the winter fabrics, relatively low values notions of handle for summer fabrics.
of percentage error can be achieved by including more
The individual differences between judges within a
than four judges in a panel, e.g., 12% maximum error
for the winter fabrics that the mean handle
for eight judges. For the summer fabrics, even using panel imply
assessment of a national panel can be predicted by one
eight judges results in an expected maximum error of expert judge with only a moderate level of
30%. The low between-panel variation and the small precision.
i.e., only fabrics with differences ininean handle rating
increase in the value of reproducibility over repeata-
of greater than 1.6 units on a scale from 1.0 to 5.0 can
bility indicate that prediction is very good from one be
distinguished reliably. This
panel mean to another panel mean. This good predic- reasonable levels, ±0.37 units, reliability
improves to
when the assessments
tion between national panels will not apply for all the
of eight judges are considered for the winter fabrics.
winter fabrics because of the significant ( panel X fabric)
Further useful information about fabric character-
interaction effects we observed in this analysis.
istics, such as smoothness, softness, stiffness, flexibility,
etc., can be obtained based on assessments of these
Conclusions properties. These characteristics may also be used to
A prerequisite for applying low-stress mechanical explain more fully the notions of fabric handle devel-
and surface properties of fabrics to the estimation or oped in this analysis. The next paper in this series con-
prediction of handle is to establish clearly what is meant siders these characteristics or quality attributes in some
by the term &dquo;fabric handle.&dquo; No accepted analytical detail, and a later paper will present an objective clas-
definition exists. The general definition of handle dis- sification of the winter and summer fabrics based en-
cussed in this paper includes an element of psychology, tirely on instrumental measurements of low-stress me-
implying that there may legitimately be differences in chanical and surface properties.
assessments between individuals.
The investigations we present in this paper have
ACKNOWLEDGMENTS
shown that there exists a common notion of the handle
of men’s suiting fabrics within national panels of expert We are greatly indebted to the many judges from
judges drawn from the textile and clothing industries the textile and clothing industries of eight countries for
of Japan, Australia, New Zealand, India, the U.S.A., their invaluable cooperation and participation in the
the P.R.C., and Hong Kong/Taiwan. The strength of international fabric handle survey reported in this pa-
this common notion is greater, i.e., there are fewer in- per. We also acknowledge with sincere thanks the con-
dividual differences, for the handle of the winter fabrics tributions made by the national coordinators of the
than for the summer fabrics. fabric handle survey listed in the appendix to this paper.
There is a strong commonality of handle preferences We would like to thank Dr. R. C. Dhingra for his help-
for the winter fabrics as demonstrated by the seven ful suggestions and Mr. J. Galea for his valuable tech-
national panels. Subtle underlying national preference nical assistance. We also acknowledge the financial
have also been isolated in the handle assessments of support provided for this work by the Australian Wool
these fabrics. The situation is different when we con- Corporation from the Wool Research Trust Fund.

Downloaded from trj.sagepub.com at LSU Libraries on May 8, 2015


732

.
Appendix how close to their ultimate consumers the expert judges
were in their assessments of fabric handle.
DETAILS OF 1NTERNATIONAL FABRIC
The selection of a final panel of judges to represent
DANDLE SURVEY
the handle preferences of a particular country was made
Aims
d
of Survey on the basis of the correlation of each judge’s handle

There were two particular aims of the international assessments with those of the mean assessments of his
national panel. In this manner, seven national panels
survey: to establish whether or not there exists an
were assembled, each consisting of the eight judges
agreed-upoh concept, or concepts, of fabric handle for
men’s suitings, or more specifically, whether panels of whose handle preferences most closely matched the
mean assessments of their national groups.
judges assessed the handle of two series of men’s suit-
ings, one of winter fabrics and the other of summer Fabrics
fabrics, in a similar manner; and to ascertain whether
The 214 winter and 156 summer fabrics used in the
panels ofjudges with experience in the textile or cloth-
ing industries, from different countries and with dif- trial were all produced commercially. These two sets
ferent cultural backgrounds, rank fabric handle in a of fabrics were chosen to be representative of suiting
similar manner. fabrics available on the market, and to be of such a
number that each fabric in a set could be assessed by
National Coordinators each judge.
We obtained the cooperation in the nominated Table AI provides a brief summary of the two sets
countries outside Australia from people well placed to of fabrics used in the winter and summer surveys. A
perform the survey in a rigorous scientific manner. complete description of the fabrics is provided else-
These national coordinators, to whom we are extremely where [ 15 ] . Inspection of Table AI highlights the pre-
grateful are Professor S. Kawabata, Kyoto University, dominately pure wool worsted and wool/synthetic
Japan: Dr. G. A. Camaby, Wool Research Organisation worsted nature of the fabrics. The great diversities of
of New Zealand; Professor V. B. Gupta, Indian Institute fabric structure and surface finish in the winter samples
of Technology, Dehli; Dr. R. C. Dhingra, Department are also evident within that set. -

of Textile Technology, University of New South Wales


(India); Dr. P. Brown, University of Minnesota TABLE AI. Commercially produced fabrics used in handle survey.
(U.S.A.); and Professor Chen Wen Xiang, China Tex-
tile University, Shanghai; Professor Kung Fan Chao,
Tianjin College of Engineering, Tianjin, People’s Re-
public of China (P.R.C.).
Judging Panels
Independent sets of expert judges were selected from
a variety of areas within the textile and clothing in-
dustries to represent each of the seven nominated
countries. The 17 expert Australian judges were drawn
from the following areas of these industries: fabric de-
signing (seven judges, including two with marketing
and two with quality control responsibilities ), finishing
(two judges), clothing manufacture (three judges, in-
cluding one with retailing responsibilities), merchan-
dising (two judges), marketing (one judge), and fabric
research and development (two judges). The six other
*
sets of expert judges had a similar mix of background Includes pure wool, pure cashmere, cashmere/wool, and cash-
b
and experience, with the following total numbers: 12 mere/synthetic blends. Barathea and knitted constructions.

Japanese, 13 New Zealanders, 15 Indians, 15 U.S.; 16


Chinese (P.R.C.); and 12 Hong Kong/Taiwanese.
In this trial, we also used another set of judges con- Survey Design
sisting of eight Australians with no particular experi- In the execution of the survey each judge was asked
ence in these industries (consumer judges) to establish to assess 20 X 10 cm samples of each fabric. They were

Downloaded from trj.sagepub.com at LSU Libraries on May 8, 2015


733

told to judge the fabrics following their country’s base Multiple Factor Analysis to the Assessment of Fabric
and their own definitions of &dquo;good&dquo; handle, but the Handle, J. Textile Inst. 49, T540 (1958).
effect of fashion, color, and pattern was excluded so 10. Kawabata, S., "The Standardisation and Analysis of
that only fabric handle was judged. The grading of Hand Evaluation," The Textile Machinery Society of
&dquo;good&dquo; handle is as shown in the rating scale set out Japan, Osaka, 1975.
in Table I of this paper. 11. Kawabata, S., "The Standardisation and Analysis of
Fabric Hand," 2nd ed., The Textile Machinery Society
The large number of fabrics to be assessed and the
of Japan, Osaka. 1980.
requirement that each judge had to assess each fabric 12. Kawabata, S. ed., "Standard of Hand Evaluation," vol.
necessitated that a rating scale be used as the means III, Hand Evaluation and Standardisation Committee,
of assessing fabric handle, rather than a paired com- Textile Machinery Society of Japan, Osaka, 1982.
parison technique as recommended by Howorth and 13. Lundgren, H. P., New Concepts in Evaluating Fabric
Oliver. The rating scale given in Table I is of the type Handle, Textile Chem. Color. 1 (1), 35(1969).
recommended by Sheppard [ 19] on the basis of psy- 14. Mahar,T. J., Dhingra, R. C., and Postle, R., Measuring
chological investigations. and InterpretingLow-Stress Fabric Mechanical and Sur-
face Properties. Part I: Precision of Measurement, Textile
Res. J. 57, 357 (1987).
Literature Cited 15. Mahar, T. J., The Measurement of Low Stress Mechan-
1. Binns, H., A Tactile Comparison of the Cloth Qualities ical and Surface Properties of Fabrics and Their Appli-
of Continental and Noble-combed Materials, J. Textile .cation to Fabric Handle and Fabric Tailorability. Doc-
Inst. 25, T157 (1934). toral thesis, University of New South Wales. 1989.
2. Binns, H., The Discrimination of Wool Fabrics by Sense 16. Matsuo, T.. Nasu, N., and Saito, M., Study of Hand.
of Touch,
B. J. Psychol. 16, 237 (1926). Part 2: The Method of Measuring Hand, J. Textile Mach.
3. Binns, H., A Comparison Between the Judgements of Soc.
17
(3), 92 Jpn.
(1971).
Individuals Skilled in the Textile Trade and the Natural 17. Postle, R., and Dhingra, R. C.. Measuring and Inter-
Judgements of Untrained Adults and Children, J. Textile preting Low-Stress Fabric Mechanical and Surface
Inst. 17, T615 (1926). Properties, Part III : Optimization of Fabric Properties
4. Bona, M., Essai sur la Determination Statistique de la for Men’s Suiting Materials, Textile Res. J. 59, 448
Qualite des Tissus, Doctoral thesis, University Haute Al- (1989).
sace, 1984. 18. Precision of Test Methods&mdash;Determination of Repeat-
5. Bona, M., A Contribution to the Statistical Definition ability and Reproducibility by Interlaboratory Tests. ISO
of Fabric Quality, I.W.T.O. Report 2, Tokyo, May 1984. 5725, International Standards Organisation, Switzeriand,
6. Cooley, W. W., and Lohnes, P. R., "Multivariate Data 1981.
Analysis," J. Wiley & Sons, New York, ch. 4 and 5, 19. Sheppard, D., Subjective Assessment of Scientific Mea-
1971. surements, Lab. Pract. 488 ( Sept. 1953).
7. Dhingra, R. C., Liu, D., and Postle, R., Measuring and 20. Statistical Packages for the Social Scientist, "User’s
Interpreting Low-Stress Fabric Mechanical and Surface Guide." SPSS Inc., Mc Graw-Hiil Book Co., New York,
Properties, Part II: Application to Finishing, Drycleaning, 1983.
and Photodegradation of Wool Fabrics, Textile Res. J. 21. The Textile Institute, "Textile Terms and Definitions."
59,
357 (1989). 8th ed., Manchester, United Kingdom. 1986.
8. Hilgard, E. R., and Atkinson, R. C., "Introduction to 22. Thurlstone, L. L., "Multiple Factor Analysis," University
Psychology," 4th ed., Harcourt, Brace and World Inc., of Chicago Press, Chicago,1947.
New York, 1967.
9. Howorth, W. S., and Oliver, P. H., The Application of Manuscript received January 19. /989. awepted M&oelig;dt 16, 1989.

Downloaded from trj.sagepub.com at LSU Libraries on May 8, 2015

You might also like