I__. - International Military Testing Association

think that difference especially disc!oncerting, while others
thinkit’s monumental - ‘and those viewpointscanbeirrespective
of whether there is an understanding of the
underlying distribution or not.
These issues led us to the literature to see what had
been reported on rating scale distributions in respect to
product acceptability.
The literature showed that, in recent years, a new
objective measure for determining level of product acceptability
labelled the “acceptor set” has been described
in marketing research literature, especially that of the food
industry (Gordon and Norback, 1985). The measure has
been used in conjunction with food product optimization
techniques and market positioning (Lagrange and Norback,
1987).
While the acceptor set canbe determined by a simple
binary method (dichotomous question), both Gordon
(198s) and Choi and Kosikowski (1985) described creating
an “acceptor set” from scaled data by splitting the
sample group into two percentages, the percentage who
found a product acceptable and those who did not. For
example, respondents to our 7-point scale (l=“dislike
verymuch”107=“like verymuch”)couldbesplit intotwo
groups - tither the S-7 group, or the l-4 group, with the 5-
7 group constituting the acceptor set. Product optimization
then means finding methods to increase the acceptor
set percentages (however derived) as measures ofproduct
improvement.
Given those findings, we decided to look at the acceptorset
conceptinrespect tooursurveydatabasetoseel~ow
we could add to the definition ofproduct acceptability for
both the manager and researcher.
Methodology
Data base description
The rating scale data were obtained on questionnaires
administered to approximately 7,500 combat arms soldiers
who rated products such as field rations, protective
clothing, tents, and airdrop equipment. Data collectors
went to the survey sites after these soldiers had returned
from major training exercises where they had used one or
more of the products. Entire units were tasked to participate
in the surveys; soldiers in these units could refuse to
fill out questionnaires if they chose, but few did. The
sample size at each site r‘anged from 200 to 400. The
soldier population was male, with over 90% between the
agcsofbetween 19and 23 andservingintheenlistedranks
E-2 to E-4.
The verbal rating scales were either 7-point or Y-point
scales. The 9-point scale, which has also been called a
hedonic scale (Peryam and Pilgrim, 1957), has been a
242
scale traditionally used in military and civilian food research
formore than 30 years (Maller and Cardello, 1984).
Itwasonlyusedforrating the taste ofspecificration(food)
items. Acceptability ratings for ration attributes other
than taste (e.g. acceptability of portion sizes) were obtained
on 7-point scales.
The verbal anchors for the 7-point scales were: goodbad,
satisfied-dissatisfied, easy-difficult, comfortableuncomfortable
andlike-dislike. The 9-point scale anchors
werelike-dislike. Eachscale had adverbmodifiers forthe
anchors that graduated in intensity. For example, the 7point
good-bad scale was: .
VERY MODERATELY SLIGHILY NEtTtLER MAD SLKWTLY MUDEKATELY VERY
BAD BAD BAD NOR CQOD CXWD GOOD GOOD
1 2 3 4 5 6 7
Each of the scales had a neutraI point and the positive
verbal anchors for the scales were at the high ends, i.e. 5-
7 for the 7-point scale and 6-9 for the g-point scale. The
product acceptnnce issues covered a wide range of variablessuchasdurability,
appearance,comfort, taste, weight,
compatibility (with other pieces of equipment), weatherproofing,
warmth, and “overall” acceptability.
Analysis
We based our ‘analysis on randomly selected mean
ratings from our survey data. The number of means
selected was 155 for the 7-point scale and 21 for the 9point.
The largest sample contributing to any particular
mean numbered 347, and the smallest 34. The lowest
mean rating on the 7-point scale w;1s 2.94 and the highest
was 6.53; the lowest for the 9-point was 3.01 ‘and the
highest 6.35. The mean of the means obtained on the 7point
scales was 4.71 (SD=.75), while the mean of the
means obtained on the 9-point scale was 4.59 (SD=.9 1).
The distributions for all the selected means were unimodal.
We explored the relationships of the means to the size
of the acceptor sets through regression analyses. The
acceptor set definition was the percent of ratings falling in
the entire positive range for either scale, i.e. S-7 for the 7point
and 6-9 for the 9-point scale.
Results
The results show extremely good fits with linear
regression models for both 7- and 9-point scales. Figure
1 shows the scatter plot for the relationship of the means
to the acceptor sets for the 7-point scale. The R* in this c,ase
is .97 with a regression equation of:
y = - 54.45 + 24.13x.