Changes in Consumer Perceptions: the Impact of Testing Conditions on Perceptions of Branded Products

James McCullough, University of Arizona
Charlene S. Martinsen, University of Washington (Student), University of Washington
Linda Sceurman,
ABSTRACT - Perceptions of products as measured by multidimensional scaling of perceived differences appear to be unsensitive to changes in the stimulus set than had been expected. Selection of elements for presentation to subjects in this type of study does not appear to be a critical source of experimenter bias.
[ to cite ]:
James McCullough, Charlene S. Martinsen, and Linda Sceurman (1979) ,"Changes in Consumer Perceptions: the Impact of Testing Conditions on Perceptions of Branded Products", in NA - Advances in Consumer Research Volume 06, eds. William L. Wilkie, Ann Abor, MI : Association for Consumer Research, Pages: 574-577.

Advances in Consumer Research Volume 6, 1979      Pages 574-577


James McCullough, University of Arizona

Charlene S. Martinsen, University of Washington

Linda Sceurman (Student), University of Washington


Perceptions of products as measured by multidimensional scaling of perceived differences appear to be unsensitive to changes in the stimulus set than had been expected. Selection of elements for presentation to subjects in this type of study does not appear to be a critical source of experimenter bias.


Although the use of multidimensional scaling (MDS) techniques to examine perceptions of sets of products has become of fairly common approach in studies of consumer behavior (Green, 1975), a major difficulty in the use of MDS has been interpretation of the perceptual spaces derived by the technique. The particular problem of dimensional interpretation has been discussed in detail (Shepard, 1974). Studies of simple stimuli indicates dimensional structure is related to the attributes of the stimuli presented (McCullough, Martinsen and Moinpour, 1978). If this is true, then it seems likely that alteration of the set of stimuli presented would change the underlying dimensional structure of the perceptual space derived by MDS.

The method of presentation of stimuli has been shown to influence both the perceptions of the objects (Allison and Uhl, 1964) and the weightings assigned to the perceptual dimensions (Rao, 1972). These researchers have clearly demonstrated that product perceptions and subsequent perceptual spaces are highly sensitive to changes in explicit information presented concerning stimuli at the time of evaluation. Situational effects on perceptual spaces have been discussed by McCullough and Martinsen (1976), particularly when comparing brand evaluations with sensory evaluation of actual products. An excellent taxonomy of the relationship between perceptions and underlying structure can be found in Garner (1976). These concerns have stimulated this study which examines changes in expressed perception induced by changes in the elements of a stimulus set and by alternative methods of stimulus presentation.

Multidimensional scaling (MDS) appears to be a particularly good technique for examining changes in product perception induced by changes in the method of presentation of the products. Moinpour, McCullough and MacLaclin (1975) have demonstrated the usefulness of MDS in measuring longitudinal changes and have shown that measurements of perceptual spaces can be replicated with a high degree of reliability. That study examined the impacts of messages about products when the stimuli (brand names) were presented in a consistent fashion. This study examines the changes in product perceptions induced by alteration of the presentation format and by modification of the stimulus set.


It has been shown that the underlying structure of the perceptual space for a set of product can be related to the characteristics of the elements of the stimulus set (McCullough, Martinsen, and Moinpour, 1978). It seems reasonable from this finding that by changing elements in a set the perceived structure could be altered. It seems likely that these changes would occur: 1) if dissimilar elements were introduced into the stimulus set, or 2) if the set of attributes of the product presented to the subject is changed, since both represent significant alternation of the stimulus set.

The changes measured by MDS could result from three changes in perception by the consumer: 1) The attribution of the dimensions of the space to the stimuli may be changed; that is, the subject may apply the discrimination criteria differently to existing stimuli; 2) the subject may use new discriminating criteria or may reduce the number of criteria used; or 3) the subject may change the relative weighting or importance of the discriminating criteria. If the structure of the perceptual space is more stable than expected and is not altered by changes in the stimulus set, the spatial relationships between the elements of the stimulus set should remain relatively unchanged.

To attempt to measure the effect on the structure of the perceptual space of changes in stimuli three types of change were examined: 1) alteration of the set attributes as perceived by the subjects; 2) substitution of a single stimulus similar in form to existing stimuli in the stimulus set 3) substitution of several stimuli of different form in the stimulus set. It was believed that these treatments would provide a reasonable range of conditions under which changes might be induced.


Subjects and Training

Predominately female groups of approximately 20 subjects between the ages of 18 and 35 were chosen to participate in the various phases of this study. The subjects were familiar with sensory evaluation techniques but had no prior experience with MDS procedures. They were told that they were taking part in a test of a new marketing research tool, that they would be asked to evaluate a set of soft drink brands and that they were to use any criteria they felt were appropriate. Subjects then participated in a tasting session in which they were asked to taste and evaluate a set of soft drink samples. Subjects were given no information concerning the brands presented for sensory evaluation until the scoring had been completed.

Stimulus Presentation

Previous studies had employed sets of commonly consumed and readily available soft drinks selected to represent the variety found in the marketplace. The single house brand in the stimulus set was selected as the element to be substituted since it was likely to have less impact on perceptual structure than the national brands. For the multiple substitution treatment, diet analogs of brands were substituted into the stimulus set while attempting to maintain the analogous product clusters which had been previously found in the standard set.

All products used in sensory evaluation were purchased in 12 ounce cans at local supermarkets and were refrigerated prior to serving. One ounce samples of soft drinks were presented to the subjects in small glasses for sensory evaluation. The soft drink brands used in the study are shown in Table 1.

Subjects completed a questionnaire requiring similarity-dissimilarity ratings on nine point scales for all possible pairs of products or brand names of the products under study. The pairs of stimuli were evaluated in randomized order, and each subject responded to the same order of pairs. Most individuals required 15-30 minutes to make the required pairwise evaluations.

Proximity ratings provided by each subject submitted to the INDSCAL individual differences scaling algorithm (Carroll and Chang, 1970). The resulting stimuli spaces were compared using the C-MATCH algorithm (Cliff, 1966). Individual saliences obtained from INDSCAL were compared across groups using the MANOVA algorithm where appropriate (Cooley and Lohnes, 1971).


Subject dissimilarity ratings were submitted to the INDSCAL algorithm to identify the brand position in the perceptual space most common to the subjects. The solution obtained for the standard set is shown in Figure 1. Perceptual spaces obtained for the various treatment conditions, were compared using the C-MATCH algorithm on the coordinates of all stimuli common to the spaces being compared. That is, only coordinates of stimuli present in both spaces were used in making spatial comparisons. The degree of the shift in the eight stimulus spaces induced by substituting a single element was measured by comparing the positions of the remaining seven stimuli, while comparison of the standard set with the high diet set was based on comparisons of the five common elements. These analysis were performed for both sensory evaluation of products and for evaluation of brand names. The results of these analyses are shown in Tables 2 and 3.

Sensory Evaluation

For common non-diet soft drink brands, a C-MATCH goodness of fit of .96 and a distance vector correlation of .95 had been reported between brand perceptions and sensory evaluations. In that study McCullough and Martinsen (1976) had reported significant changes in dimensional salience despite high correlations between spatial maps. Based on those findings it was suggested that verbal surrogates should not be employed in marketing research applications aimed at measuring actual product perceptions because perceptions in the two presentation modes likely would differ. However, they were unable to demonstrate any significant difference in the perceptual spaces between brand perceptions and sensory evaluations of soft drinks, in spite of significant differences in dimensional salience between the two modes. Their conclusions only served to raise questions concerning the methodology.

In the present study, the perceptual maps were so clearly different that the MANOVA test for differences in salience was unnecessary and seemed inappropriate since the actual underlying dimensions had obviously changed. This strengthens the previous findings that suggested verbal representations of actual soft drink products should be employed with caution in marketing research situations.







In attempting to explain the reasons for the significant differences noted in this study, it was believed that color represented a significant discriminating variable in the sensory evaluation of the soft drink samples that might be influencing perceptions. To test the impact of color, subjects were presented samples in covered containers which hid the color of the samples. There was no significant difference between the configurations and the goodness of fit was .95. These results are shown in Table 2. Apparently the differences in sensory perceptions are more complex than had been hypothesized.

Brand Evaluations

Substitution of a single element in the stimulus set was not expected to produce any significant change in the perceptual maps. The goodness of fit was .98 and the correlation of distance vectors was .95 based on the relative positions of the seven constant elements. The result shown in Table 3 indicate a clear consistency between the two spaces and no significant alteration of the space was found.

Substitution of three diet brands for three conventional soft drinks was expected to induce a shift in spatial dimensionality to include a diet/non-diet dimension altering the relative positions of the five constant elements which included Tab, also a diet drink. The goodness of fit of .99 and the vector correlation of .98 indicate that such a shift did not occur.

Examination of the mean subject correlations with the INDSCAL perceptual map for the various spaces indicates a possible, slight alteration of the underlying perceptual structure as elements are altered. The single brand substitution improved the mean square correlation from .76 to .83 while the substitution of diet brands reduced the correlation to .73. These shifts are slight and not statistically significant indicative of a tendency toward higher dimensionality.



When the perceptual spaces are not structurally dissimilar, as occurred between sensory evaluations and brand perceptions of diet soft drinks, a MANOVA test for salience dissimilarity would seem to be appropriate. The utilization of the dimensional structure by the subjects as indicated by dimensional salience may change even when the perceptual maps appear identical, and examination of possible variation in dimensional salience using MANOVA can provide additional insight concerning product perceptions. In this case, however, the MANOVA analysis failed to indicate any significant differences between the groups tested. Group mean saliences for the groups tested are shown in Table 4. The only significant differences are between the standard sets in sensory and brand evaluations as had been previously reported (McCullough and Martinsen, 1976).

Comparison of Sensory and Brand Evaluations

Earlier research had shown that configurations derived from brand evaluations were quite similar to those derived from sensory evaluation. This was not the case for the multiple substitution set. The goodness of fit for these two perceptual spaces was only .57. This may indicate that for conventional soft drinks brand perceptions are much closer to the actual consumption experience than for diet drinks with which the consumers may have had less experience. This is clearly only a hypothesis for which additional experimental support is necessary.




The results of this study would seem to indicate that the selection of elements for inclusion in stimulus sets for evaluation by MDS may not be a critical source of experimenter bias, and that subjects are able to maintain a stable perceptual space even when elements are substituted. Since one would expect that the selection of particular stimuli would play a large role in the determination of the underlying dimensions of the perceptual space developed by individual subjects, this raises a question about the basis for dimensional determination.

One common technique identifying the axis of MDS solutions involves use of maximum "r" regressions against attribute ratings for individual stimuli. It is generally assumed, however, that the subject is free to select any criteria deemed appropriate in developing the perceptual space, and the resulting space contains the actual perceived distances between stimuli in that space. If alteration of the stimulus set fails to produce changes in the perceived relationships between unaltered elements, as occurred in this experiment when additional diet drinks were substituted, then the dimensions of the space may not be linear axes in the space but might more correctly be viewed as areas or domains as Shepard (1974) suggested. This question should be explored in more detail.

At a practical level, it would appear that alterations of a stimulus set which do not change the basic perceived character of the set can be incorporated. The experiment involving a single element change clearly indicates that is true. The failure of the multiple substitution to change the space supports this, since extensive changes can apparently be made and the perceived relationships between unaltered stimuli will be retained.

This study has not attempted to thoroughly catalog changes which can or cannot be made. The effects of large and small changes have been examined and characterized for a single product category. Further experimentation in this area is necessary to adequately characterize effects for stimuli which are not so familiar as soft drinks.


R. I. Allison and K. P. Uhl, "Influence of Beer Brand Identification on Taste Perception." Journal of Marketing Research, 1964, 1, 36-39.

J. D. Carroll and J. J. Chang, "Analysis of Individual Differences in Multidimensional Scaling Via an N-Way Generalization of Eckart-Young Decomposition." Psychometrika, 1970, 35, 283-319.

N. Cliff, "Orthogonal Rotation to Congruence." Psychometrika, 1966, 31, 31-42.

W.W. Cooley and P. R. Lohnes, Multivariate Data Analysis. New York: John Wiley and Sons, 1971, 223-242.

W.R. Garner, "Interaction of Stimulus Dimensions in Concept and Choice Processes." Cognitive Psychology, 1976, 8, 98-123.

P. E. Green, "Marketing Applications of MDS: Assessment and Outlook." Journal of Marketing, 1975, 24, 410-420.

J. M. McCullough and C. S. Martinsen, "Differences in Stimuli Evoked Responses Between Physical Stimuli and Verbal Surrogates." Proceedings of the American Marketing Association Educators' Conference, 1976, 45-48.

J. M. McCullough, C. S. Martinsen and R. Moinpour, "Application of Multidimensional Scaling to the Analysis of Sensory Evaluations of Stimuli with Known Attribute Structures." Journal of Applied Psychology (in press).

R. Moinpour, J. M. McCullough and D. L. MacLachlan, "Time Charges in Perception: A Longitudinal Application of Multidimensional Scaling." Journal of Marketing Research, 1976, 13, 245-53.

V. R. Rao, "Changes in Explicit Information and Brand Perceptions." Journal of Marketing Research, 1972, 9, 209-213.

R.N. Shepard, R.N. "Representation of Structure in Similarity Data: Problems and Prospects." Psychometrika, 1974, 39, 373-421.