Cola Preferences: Disguised Taste Vs. Brand Evaluations

Thomas J. Stanley, Georgia State University
ABSTRACT - This study was designed to determine if preferences for colas based on taste evaluations could explain overall cola brand preferences. In 25 of 60 cases studied, subjects' taste preferences were significantly related to their brand preferences. Thus, evidence supports the contention that taste is an important evaluative criterion that underlies consumer cola brand preferences.
[ to cite ]:
Thomas J. Stanley (1978) ,"Cola Preferences: Disguised Taste Vs. Brand Evaluations", in NA - Advances in Consumer Research Volume 05, eds. Kent Hunt, Ann Abor, MI : Association for Consumer Research, Pages: 19-21.

Advances in Consumer Research Volume 5, 1978      Pages 19-21


Thomas J. Stanley, Georgia State University

[The author is indebted to Dr. Ralph L. Day, Professor of Business, Indiana University, for his helpful suggestions.]


This study was designed to determine if preferences for colas based on taste evaluations could explain overall cola brand preferences. In 25 of 60 cases studied, subjects' taste preferences were significantly related to their brand preferences. Thus, evidence supports the contention that taste is an important evaluative criterion that underlies consumer cola brand preferences.


Recent studies (Lane, Zychowski, and Lelli, 1975; Thumin, 1962) have indicated that subjects can distinguish between popular brands of colas; however, these studies have not determined if the taste of colas is an important factor in explaining overall cola preferences. Several scholars (Posner, 1974; and Warne, 1962) suggest that the physical differences of products such as colas are usually unimportant factors in explaining consumer preferences. Social critics conclude that differentiation of cola brands via promotion of taste are nothing more than "foolish puffing." Governmental agencies also frown on what is called puffing ads, which promote in their words "pseudo-differences" (Backman, 1968; Posner, 1974; and Warne, 1962). In actuality, the cola industry has steered away from promoting direct taste comparisons, even though consumers disagree with the idea that product differences between competing brands of most products are insignificant and unimportant to them (Barks-dale and Darden, 1972). Historically, the majority of cola advertisements have not emphasized taste (Block, 1975), but have attempted to develop strong brand images for their products by associating them with emotionally loaded themes such as brotherhood, nostalgia, and reference groups. Critics of advertising argue against the use of these conditioning methods.

Cola producers would have a strong argument as well as motivation for the use of advertising which emphasizes taste differences if they could demonstrate that taste is an important factor in explaining cola preferences. The success of future product as well as promotional developments in this context may hinge upon whether or not taste is an important factor in explaining overall brand preferences for colas (Block, 1975). This paper attempts to determine if taste preference evaluations can explain overall brand preferences for colas.



Sixty students from a state university participated in this study. There were 41 male and 19 female subjects who ranged in age from 19 to 27 years. Eleven of the original 71 volunteers did not participate in the study; five subjects indicated that they were not familiar enough with colas to evaluate them; and the other six withdrew before completing their evaluations.


The stimuli used in this study were seven cola drinks including: (1) Adirondack Cola; (2) Canada Dry Cola; (3) Coca-Cola; (4) Cott Cola; (5) Pepsi Cola; (6) Shasta Cola; and (7) Tab. Adirondack and Cott are regional soft drinks while the others are well known nationally. Shasta and Tab were the only diet colas used in the experiment.

Each subject evaluated two separate sets of stimuli. Subjects were first asked to indicate their preferences for the colas, in a blind, paired comparison taste test. Each subject rated all 21 pairs and indicated a preferred cola in each pair. No information about the identity of the cola was given the subjects prior to or during this part of the experiment. Approximately four ounces of each cola was given to each subject in opaque plastic cups. The colas were stored at approximately 5oC prior to their distribution. Salt free flour crackers were also distributed and subjects were instructed to consume a bit of the crackers after tasting each cola pair. The position of the colas in every pair as well as the order that pairs were presented in was randomized.

Following the completion of the taste evaluations each subject made preference evaluations on a set of 21 randomly ordered pairs of the names of seven cola brands. The brand names were those of colas tasted in the first step of the experiment. The responses to cola names (tastes) were used to construct a brand (taste) preference scale.

The Model

Both the taste and brand preference data were separately analyzed by the MD-PREF scaling algorithm. This program transforms subjects' paired comparison preference data matrices into rank orders or first score matrices. The Eckart-Young factor analytic procedure then decomposes the ranked data and generates a group stimulus space or matrix and termini of subject vectors (ideal points) that best agree in a least squares sense with the matrices of subjects' scaled preference values. In other words, MD-PREF determines the stimulus coordinate values and termini of subject preference vectors from the set of preference judgments that will most closely agree with the preference scale values for each subject (Green and Rao, 1972; and Green and Wind, 1973).

The output from MD-PREF includes the stimuli which are represented as points and subjects as vectors in the same space. A subject's rank preference ordering is depicted as a vector drawn through the origin of the stimulus space to the subject's ideal point. The correlation between the vector and a particular coordinate axis indicates how well a subject's preference ordering can be explained by the positions of the stimuli along that axis. The farther the subject's ideal point is from the origin the better the stimulus configuration can explain a subject's preference ordering (Carroll, 1972).


Today, the average cola consumer must decide what cola to purchase from a large set of offerings. From the results of previous cola taste experiments (Thumin, 1962; Lane, Zychowski, and Lelli, 1975) it is unclear if taste plays a role in explaining cola preferences. In this study subjects were asked to evaluate seven colas given in 21 pairs. This number (seven) may approximate the typical number of cola brands generally offered in the market area where the study was conducted. However, the possibility exists that subjects who tasted 21 successive pairs of colas encountered taste adaptation. Some amount of perceived adaptation may typify the actual market conditions under which consumers must make such decisions. Also adaptation would make it more difficult to reject the null hypothesis that no relation exists between cola preferences based on taste evaluations and those based on brand preference evaluations. Rejection of the null hypothesis under such conditions would provide a stronger argument that taste preferences are generally important in explaining brand preferences.

Both the taste and brand preference data were separately analyzed by the MD-PREF multidimensional scaling algorithm. The resulting preference orderings (taste preference and brand preference scales) for each individual were correlated to determine the extent to which taste evaluations could explain evaluations based on brand names. The multidimensional scaling method was used instead of more traditional scaling methods for several reasons. An important aspect of multidimensional scaling methods lies in their ability to accurately reproduce, in the form of a multiple dimension geometric space, preference as well as perceptual configurations (Dubois, 1973; Green and Carmone, 1970; and Stanley, 1976, a, b).


Tables 1 and 2 illustrate the stimulus spaces from the MD-PREF analyses of taste and brand preference evaluations. Two factors explained 70 percent of the subjects' taste judgments and 78 percent of the preference evaluations based on brands. Factor one explained slightly more than 50 percent in both cases. In both cases the third factor added little to the explanation of preference judgments and was therefore excluded from further analysis.

Tables 1 and 2 demonstrate that in regard to factor one high negative scores are associated with the diet colas, namely Shasta and Tab, while high positive scores are noted for Pepsi and Coca-Cola. The direction of subjects' ideal points suggest that positive values are indicative of more preferred colas than those with negative values.





Factor one in both preference spaces indicates that subjects consistently evaluated Coke and Pepsi as distinctly more preferred over the diet colas. Therefore, the evaluations of Coke and Pepsi and the diet colas would make a substantial contribution to high positive correlations that may be found between taste and brand preferences. Without the assistance of other measures such as perceptual and corresponding property scales, no attempt to label the factors generated from an internal preference analysis should be made.

Subject Correlations

Two rank orders (individual taste and brand preferences) of the stimuli were computed from the two factor MD-PREF outputs for each of the 60 subjects along their preference vectors. In turn, 60 Spearman rank correlations were computed between the ranks assigned by each subject to colas on the basis of taste and brand evaluations (see Table 3).

Twenty five of thc 60 correlations were positive and significantly greater than zero, at the p < .05 level. The chance expectation at the p < .05 level is that only three of the 60 subjects will have a significant rank order correlation. As determined by the Poisson approximation to the binomial distribution, the probability of finding 25 or more of the 60 significant correlations where no actual relationship exists between taste and brand preference for any of the subjects is less than .001, where n = 60, p = .05, and M = np = 3.0.


The results of this study indicate that a significant relationship exists between cola preferences based on taste evaluations and brand preferences. Therefore, a possibility exists that cola brand preferences could be predicted from taste evaluations. Thus, cola producers may wish to consider placing more emphasis on promoting and developing unique flavor characteristics for their products. By demonstrating that variations in brand preference can be explained by taste evaluations cola producers can counter social critics and government agencies who condemn the use of product differentiation themes in cola promotions.

Multidimensional scaling techniques appear to be a useful tool in measuring consumer taste as well as brand preferences. Future studies should investigate the perceptual properties that explain taste and corresponding brand preferences. By being able to understand the variations in the saliences consumers attached to the dimension of taste, cola producers will be in an excellent position to develop products which are more closely tailored to the needs of various market segments.




Jules Backman, "Is Advertising Wasteful?" Journal of Marketing, 32 (January, 1968), 2-8.

Hiram C. Barksdale and William R. Darden, "Consumer Attitudes Toward Marketing and Consumerism," Journal of Marketing, 36 (October, 1972), 28-35.

Maurine Block, "Pepsi Dares Coke in Texas Taste Text," Advertising Age, (April 14, 1975)

J. D. Carroll, "Individual Differences and Multidimensional Sealing," in R. N. Shepard, A. K. Romney, and S. B. Nerlove, (eds.), Multidimensional Scaling Volume I (New York: Seminar Press, 1972), 105-155.

B. Dubois, "Ideal Point Versus Attribute Models of Brand Preference: A Comparison of Predictive Validity," in Mary Jan Schlinger (ed.), Advances in Consumer Research, Vol. 2 (Chicago: Association for Consumer Research, 1975), 321-333.

Paul E. Green and Vithala Rao, Applied Multidimensional Scaling (New York: Holt, Rinehart and Winston, Inc. 1972), 7-8, 212-214.

Paul E. Green and Yoram Wind, Multiattribute Decisions in Marketing (Hinsdale, Illinois: The Dryden Press, 1973), 108-109.

Paul E. Green and Frank J. Carmone, Multidimensional Scaling (Boston: Allyn and Bacon, 1970), 82-84.

Sam H. Lane, James Zychowski, and Kenneth Lelli, "Cola and Diet Cola Identification and Level of Cola Consumption,'' Journal of Applied Psychology, 60 (1975), 278-279.

Richard Posner, "Advertising and Product Differentiation" in J. S. Wright and J. E. Mertes (eds.), Advertising's Role in Society (St. Paul, Minnesota: West Publishing Co., 1974), 44-46.

Thomas J. Stanley and Murphy A. Sewall, "Image Inputs to a Probabilistic Model: Predicting Retail Potential," Journal of Marketing, 40 (July, 1976), 48-53. (a)

Thomas J. Stanley, "Predicting Retail Trade via Chain Preference Measures," a paper presented at the 1976 Conference of the American Psychological Association, September 6, 1976, Washington, D. C. (b)

F. J. Thumin, "Identification of Cola Beverages," Journal of Applied Psychology, 46 (1962), 358-360.

Colston E. Warne, "Advertising--A Critic's View," Journal of Marketing, 26 (October, 1962), 10-14.