Вы находитесь на странице: 1из 15

THE RELATION BETWEEN SENSORY, LIKING AND IMAGE ATTRIBUTES: THE CASE OF SOAP

HOWARD R. MOSKOWITZ

Moskowitz Jacobs Inc. 1025 Westchester Ave., Suite 400 White Plains, New York 106043.508
Received for Publication February 23, 1997

ABSTRACT This paper discusses the relationfor soaps between sensory attributes and both liking and image attributes. A clear relation emerges between sensory attribute level and liking, but no clear relation emerges between sensory attributes and image attributes. There are two possible conclusions to be drawn from these results. One conclusion is that consumers can validly assign ratings to the image attribute of a soap, but that there is no way to trace this image rating back to sensory inputs. This conclusion suggests that more research is needed to understand the meaning of image attributes. The second conclusion is that consumers cannot validly rate the image attributes of a soap, even though they can complete the questionnaire. This second conclusion implies that consumers can validly rate some attributes (e.g., sensory, liking), but not others (e.g., image), and that it may be misleading to collect and attempt to analyze image ratings for health and beauty aids products.

INTRODUCTION Importance of Attributes to Developers


In the applied world of product development knowing what consumers want is quite important because that knowledge guides product developers on the one hand, and product marketers on the other. Product developers need to know what to produce. A product with inappropriate sensory characteristics may not be acceptable, no matter how strong its image and advertising, and in many cases the product will probably fail in the market. Product marketers, in turn, need to understand what specific characteristics to call to the consumers attention, either as primary reinforcers (viz., the sensory attribute is pleasing, and thus increases the chance of product repurchase), or as secondary reinforcers (viz., the sensory
Journal of Sensory Studies 13 (1998) 13-27. All Righrs Reserved. Topyright 1998 by Food & Nutrition Press, Inc., Trumbull, Connecticut.
13

14

H.R. MOSKOWITZ

attribute is neither pleasing nor displeasing, but can be used to signal that the product has other desired properties). Sensory analysts have recognized the importance of descriptive attributes and have expended considerable efforts to train experts in the proper use of terms. Such a recognition exists both for healthmeauty aids as well as for food. In both areas there exist well defined lexicons of terms, primarily for the chemical senses (ASTM 1995), but also for the senses of touch and vision (Szczesniak ef al. 1975. Oftentimes these attributes are linked directly with reference stimuli in order to make the developers job easier, because the developer is given a physical representation of the sensory attribute. It then becomes the developers job to identify the physical stimuli or process variables which can generate that sensory attribute. For image attributes, there are no such physical referents, and the task of description becomes much more difficult, and, as will be shown in this paper, much less actionable. Although many researchers working with descriptive analysis confine their efforts to trained consumers and/or experts, untrained consumers also use sensory attributes in a reliable and valid fashion. In some cases the data show that consumers can do quite well in describing the product, almost as well as do experts (Moskowitz 1995). This ability of consumers to scale the intensity of an attribute does not surprise experimental psychologists, especially psychophysicists who study the relation between physical stimulus properties and perceived sensory (or hedonic) properties. Psychophysicists have repeatedly demonstrated in the scientific literature that a well instructed consumer can rate the sensory intensity of many different types of stimuli, and that if these stimuli are systematically varied, then the ratings will correlate highly with the physical intensities (Stevens 1975). Such correlations abound in the scientific literature, mainly with sensory attributes that are unambiguous (e.g., brightness, sweetness, fragrance intensity, etc.), and where the attribute can be clearly linked to the physical level of the stimuli (viz., brightness vs luminance of light; sweetness vs concentration of sweetener, fragrance intensity vs concentration of the odor molecule in air or in a liquid, etc.). Such correlations emerge even when the attribute is the inverse of what would normally be the case (viz. fluidity vs the apparent physical viscosity of the stimulus (Steven and Guirao 1964)).

Consumers, Image Attributes and Market Research


In contrast to sensory analysts, market researchers use attributes to describe the image of a product. The image is more nebulous. An image attribute such as masculine-feminine brings into play a complex of impressions, with the underlying sensory attributes combining in different ways. Unlike sensory attributes which often can be defined in terms of a reference standard, image attributes are assumed to be idiosyncratic and personal. Image attributes span a

SENSORY. LIKING AND IMAGE ATTRIBUTES

15

wide range, from those which consumers agree on to those which consumers disagree on. Image attributes may comprise descriptions of the product, ratings of appropriateness for end use, ratings of fit to a concept, etc. These critical differences between sensory attributes and image attributes can be summarized below:
( 1 ) Image Attributes are Complex. They call into play different sensory attributes, in ways that are not understood. For instance masculine-feminine may call into play the evaluation of intensity, sweetness, etc. ( 2 ) Image Attributes Show Great Inter-individual Variability. For instance, asking ten people to define masculine-feminine or unique will call forth ten different definitions. Each person has his or her own criteria. These criteria may be based upon experience, but are not easily defined or explicated. The panelist may feel very strongly that one product is more masculine than another product, but cannot really tell the researcher why that is so. (3) Image Attributes May Vary as a Function of Current Advertising. When asked to describe a product (e.g., a beverage) using image attributes, a consumer will often parrot back what has been advertised about that product. This is advertising language. The consumer may use the term sophisticated when describing a coffee, because the consumer has heard this term used in conjunction with the product. Two year later, when another term is used the consumer may then use that term from advertising, instead of sophisticated. Although image attributes appear to be far less tractable than sensory attributes, nonetheless image attributes are critical to the market researcher, at least in an applied setting. Market researchers use image attribute ratings as diagnostics - to understand why a product is accepted or rejected. The market researcher instructs the panelist to evaluate a product (often at home), and rate the characteristics of the product on one or another scale. Most market research studies use questionnaires containing many of these image attributes for a single product. From the marketers perspective the image attribute is the key for the marketing process, despite its apparent difficulties for the research scientist. For instance, a marketer may be interested in a cosmeticoriented facial soap. The questionnaire must deal with the degree to which the soap fits this image. The image attributes in the questionnaire may extend far beyond appropriate for face, to a variety of related end uses and images. The extensive use of image attributes in market research is no accident. To the marketer the sensory characteristics are less relevant than the degree of correspondence between the product and the concept. Many market research questionnaires contain row after row of these image attributes, inserted into the questionnaire so that the researcher can quantify the fit between the actual p&luct itself and the image that the product is going to convey.

16

H . R . MOSKOWITZ

Operationalizing Image Attributes - The Use Of Functional Relations One way to make sense of image attributes is to relate the attribute either to physical variables or to sensory variables. In either case the relation shows how the image attribute co-varies with factors better understood by the developer or researcher. If the relation consists of image attribute versus formula/process variable, then the developer understands what to do physically to achieve a given level of the image attribute. If the relation consists of image attribute versus sensory attribute, then the developer understands how a well-understood sensory attribute drives the image attribute. This second case is more prevalent in practice, since the consumer researcher can easily acquire ratings both of the image attribute and the sensory attribute, for the same set of products. A key organizing principle for analysis comes from the finding that as a sensory attribute increases, liking first increases, peaks and then drops down. The prototypical relation follows the shape of an inverted U. This relation has been used to understand how sensory attributes drive overall liking (Moskowitz 1981), as well as to segment consumers on the basis of the pattern relating sensory attribute level to liking (Moskowitz and Rabino 1994). This paper makes further use of that organizing principle to compare likmg, performance and image characteristics. If this organizing principle can be extended to image attributes then the researcher will be able to understand what drives an image attribute. If this organizing principle cannot be extended to image attributes (even though it works for liking versus sensory attributes), then we will have clear evidence that image attributes obey quite different rules as compared to overall liking. Scope of This Paper This paper investigates the relation between a set of four sensory attributes of soap and a set of evaluative attributes (liking, image, performance). Evaluative attributes are those wherein the panelist goes beyond describing the sensory characteristics of the stimulus, and judges the product (e.g., in terms of acceptance, in terms of fitting an image, in terms of doing its job, etc.). The objective of the analysis is to determine whether or not the different attributes describe lawful, easy to understand relations, when they are plotted against sensory stimuli.

METHOD
Stimuli-Selection Procedure The stimuli comprised 59 soaps, which had been chosen from a larger set of

SENSORY, LIKING AND IMAGE A'ITRIBUTES

17

180 in-market soaps by profiling. A university-based panel of experienced, trained individuals evaluated each of the 180 soaps on appearance, fragrance, and touch characteristics, using a set of 46 attributes. These attributes comprised sensory descriptions of appearance, fragrance, touch for the dry soaps. The data matrix comprised mean ratings from 10 panelists in the expert panel, each of whom evaluated every one of the soaps (one judgment per soap per panelist), over a five day period. The matrix of 180 soaps x 46 attributes was then subjected to a factor analysis of the attributes, to generate a reduced matrix of 11 factors x 180 soaps. The soaps were then clustered by K-Means clustering (Systat 1994) to a reduced set of 59. The set of 59 was selected to be the most dissimilar to each other, and judged to contain within its domain most of the key sensory aspects of the soap market. The selection procedure was done prior to the consumer research. The total number of 59 soaps was selected in advance to accord with the objective of testing a wide number of soaps within a limited budget. The requirement was to obtain a minimum of 40 ratings per soap. A base size of 50 has been found to yield stable mean data, but would have necessitated testing fewer total soaps than the 59 chosen.

Home Use Evaluation by Consumers


The consumer panel comprised 21 1 females, ages 18-64, approximately equally divided across the entire age range (18-65 years old). The consumers were users of either bath or facial soap respectively, on a regular basis (at least five times per week). Panelists were pre-recruited to participate, qualified by telephone and invited to participate for the full test session (for which they would be paid). The panelists were evenly divided over six markets in the United States (New York, Chicago, Los Angeles, Jacksonville, Dallas, and Minneapolis, respectively) with approximately an equal number of panelists in each market. Each panelist evaluated 12 of the 59 soaps over a six week period. The panelists were required to use each soap for three days, and then rate the soap on a set of sensory, liking and image attributes. The data matrix thus generated an average of 42 respondent ratings per soap. Panelists rated the attributes using anchored 0-100 point scales on liking, sensory, performance, and image attributes. The sensory attributes covered appearance, fragrance and touchhexture. The image attributes dealt with uniqueness, appropriateness for body versus face, appropriateness for end uses, etc. The anchoring of scales at both ends reduced ambiguity. Anchored scales of this type have been used extensively in other studies because they are easy to use, promote discrimination, and reduce ambiguity (Moskowitz 1985).

18

H.R. MOSKOWITZ

Attributes
Table 1 presents the attributes used in the study. Prior to the actual home use portion of the study, the panelists were oriented in the meaning of each attribute by an interviewer through a one-on-one session lasting about 5 min. The objective of the orientation was to ensure that the panelist understood what was expected of her, understood each of the scales and the anchor points, and knew the full procedure for returning the soaps after one week to pick up the next set.
TABLE 1. ATTRIBUTES USED BY CONSUMERS IN THE HOME USE EVALUATION O F SOAPS (Attributes in each section ranked by decreasing standard deviation) Attribute SD Range Anribute SD Range

Liking
Purchase Shape FragrancelWet FragrancelSkin Lather Size
SenS0I-y

Image
15
14 13 11 10 10 51 62 48 48 49 48

Unique Body-Face Daily -Special YoungerlOlder

10 8 7 4

40
57 26 23

Performance
13 13 12 12 11 9 9 9

Creamy Large FragrancelWet Smooth FragrancelSkin Heavy Lather SliPperY

60 60
50
64

Comforts Skin Make Skin Supple Refreshes Softens Skin Moisturizes

12 8 8 8 8
6

54 35 34 4 0 37 21 25

41 42 47 48

Cleans
Rinses Off

SD = standard deviation of the 59 mean ratings Range = maximum mean (across 59 products) - minimum mean

SENSORY, LIKING AND IMAGE ATTRIBUTES

19

RESULTS
Analysis 1-The Distribution of 59 Soaps on Each Attribute
The first analysis computes the range and the standard deviation for the 59 means (one per soap), on each attribute. If the range is small, or if the standard deviation is small (on a relative basis), then the panelists do not perceive many differences among the soaps on the attribute. Either the attribute is irrelevant to the soap, or the attribute is relevant but the panelists have no clear idea of what the attribute means. If the panelists do not understand the attribute then they will tend to assign a random number to the soap on the attribute. The 59 soaps should lie close together. Table 1 shows the range and the standard deviation for the mean ratings of the products, on an attribute by attribute basis. Each soap generates a mean rating on each attribute, based upon the 42 panelists who evaluated it. Each attribute, in turn, has its own range of mean ratings and standard deviation across the 59 soaps tested. Table 1 suggests the following:
(1) Panelists know what they like and what they dont like. The 59 soaps show a wide range of liking. ( 2 ) One cannot conclude that panelists differentiate among products on one class of attributes (e.g. sensory attributes), but do not differentiate among products on another class of attributes (e.g., image). The specific attribute itself is critical. (3) The image attributes show smaller ranges of means than do the sensory, liking or performance attributes.

Analysis 2-Univariate Models for Sensory Attributes Versus Evaluative Attributes


The objective here is to create simple univariate models relating a sensory attribute to an evaluative attribute, and calculate the multiple R2. Although there is no absolute criterion for a valid univariate model, one can compare the univariate models created for image attributes (versus sensory attributes) to similar models created for liking and performance attributes (versus the same sensory attributes). Figure 1A-1B shows the three evaluative attributes (image-body vs face; acceptance-purchase intent; performance-moisturizes) plotted against two different sensory attributes (fragrance strength when wet, amount of lather, respectively). The curves were created by fitting the following model to the data:

20

H.R. MOSKOWITZ

Evaluative Attribute = k, = k, (Sensory Attribute)

+ k2 (Sensory Attribute)*

The fitted curves give the researcher an idea of how the sensory attribute drives the evaluative attribute. Fragrance-Intensity vs Rating
Purchase-In MoisNriza

Body-vsFace

10

20

30

40

50

60

70

80

90

100

Fragrance-Intensity
FIG.1A. HOW PERCEIVED FRAGRANCE INTENSITY DRIVES THREE EVALUATIVE RATINGS: IMAGE (BODY VERSUS FACE), ACCEPTANCE (PURCHASE INTENT), AND PERFORMANCE (MOISTURIZES)

Amount-of-Lather vs Rating
Body-vr-

Purchase-In

Moisturizes

10

20

30

40

50

60

70

80

90

100

Amount-of-Lather
FIG. 1B. HOW PERCEIVED AMOUNT O F LATHER DRIVES THREE EVALUATIVE RATINGS: IMAGE (BODY VERSUS FACE), ACCEPTANCE (PURCHASE INTENT), AND PERFORMANCE (MOISTURIZES)

SENSORY. LIKING AND IMAGE ATTRIBUTES

21

Although one can fit the quadratic function to the data, the goodness-of-fit differs, as shown by the multiple R2 values for the equations. Table 2 shows this goodness-of-fit statistic for the full set of liking, image and performance attributes separately regressed against each of four key sensory attributes.

TABLE 2. MULTIPLE R2 FOR FUNCTION RELATING SENSORY ATTRIBUTE (TOP ROW) TO RATING ATTRIBUTE (COLUMN): ATTRIBUTE = k,) + k, (SENSORY ATTRIBUTE) + k, (SENSORY ATTRIBUTE), Sensory Creaminess Sensory Fragrance Intensity Sensory Heaviness Sensory Amount Of Lather

Image
Facial vs. Body Ordinary vs. Special Unique Older vs. Younger 0.09 0.10
0.00

0.30 0.11 0.16 0.20

0.13 0.00

0.09 0.12 0.02 0.04

0.03
0.05

0.05

Liking
Purchase Intent Fragrance Lather Shape Size
0.55

0.21 0.38 0.26 0.06 0.10

0.03
0.02 0.03 0.01

0.58 0.54 0.91 0.23 0.28

0.69

0.77
0.24 0.28

0.05

Performance
Cleans Moisturizes Refreshes Softens 0.52 0.65 0.59
0.62

0.30 0.24 0.29 0.29

0.03 0.03

0.59 0.58 0.65

0.03

0.03

0.50

From this initial analysis, therefore, it appears that although one can create sensory-image curves, the goodness-of-fit of these curves is so low that one should not believe the relation. None of the R2 values exceeds 0.20. On the other hand, for liking and performance attributes the story is quite different. One can relate liking and performance to sensory attributes with varying degrees of success, but there are at least high correlations in some cases (R2>0.5, which is highly significant with 59 products). There are a sufficient number of curves with high goodness-of-fit values to suggest that liking and performance attributes can be validly related to sensory attributes (whereas image attributes cannot).

22

H.R. MOSKOWITZ

This second analysis suggests, therefore, that image attributes are not as tractable as liking and performance attributes. There are no strong-fitting sensory-image curves in which the researcher can place confidence when guiding product development. Thus, this second analysis again suggests that consumers do not have a clear idea of the meaning of the image attributes. This second analysis also suggests that sensory attributes are not equal to each other in driving attribute ratings. For instance, the heaviness of the soap does not drive any attribute, whereas the creaminess of the soap does drive all attributes except image attributes. Analysis 3-Multivariate, Factor Based Models The objective of this analysis is to create a multivariate model relating sensory attributes to each evaluative attribute. The univariate models shown in Fig. 1A and lB, and the goodness-of-fit statistics shown in Table 2 indicate that the image attributes cannot be easily modeled as a function of a single sensory attribute. However, it is possible that the image attributes are rather complex, requiring a multivariate model. One could develop a simple linear equation relating each image (or liking or performance) attribute to a set of relatively uncorrelated sensory attributes as follows: Image Attribute =

+ k,(Appearance) +k,(Fragrance) + k,(Touch). ...

Although the foregoing equation is attractive because of its simplicity, it ignores the fact that the sensory attributes (appearance, fragrance, touch) may be correlated with each other (redundancy, multi-collinearity), and that by including only a linear term there is no possibility of discovering quadratic effects. A more reasonable approach first reduces the set of sensory attributes to basic factor scores by means of principal components analysis (PCA). PCA generates a set of vectors which are statistically independent of each other, and limited in number. Each of the 59 soaps can be located on each of the PCA vectors. The locations, or factor scores for the soaps, then represent the levels of the soaps on a new series of independent variables. The PCA vectors then become predictors in the new equation. Furthermore, the PCA vectors can enter the equations in linear form, and in square form, as well as in cross-term form, allowing the researcher to take into account nonlinearities in the relation between sensory attribute level and evaluative attribute level. Table 3 shows the three major PCA vectors or dimensions emerging. Factor 1 appears to combine two sensory inputs, fragrance and the combination of lather and creamy. Factor 2 is an appearance/tactile attribute (heavy, large), and Factor 3 is a tactile attribute (smooth, slippery). The fragrance and latherkreamy attributes appear to be confounded with each other because they do not separate, even when the eigen value is reduced to 0.8.

SENSORY, LIKING AND IMAGE ATTRIBUTES

23

TABLE 3 . FACTOR LOADINGS BASED UPON PRINCIPAL COMPONENTS ANALYSIS Factor 1 Attribute FragranceBkin FragrancdWet Lather Creamy Heavy Large
Smooth

Factor 2

Factor 3

0.89 0.82
0.78 0.71

0.25

-0.11 -0.10 0.36

0.30
0.02
-0.16
0.91

0.52

0.20 0.23
0.30
-0.06

0.02
0.16

0.89 0.06

0.88 0.82

Slippery

0.40
24 %

%Variance Accounted For

34 %

24 %

The next step in the analysis creates models relating the sensory attributes to each rating attribute. The modeling approach has been discussed at length elsewhere (Moskowitz 1994). The factor scores (Fl,F2, F3) are used to create models. The models relate each attribute in turn, one at a time, to the factor scores by means of an equation. The equation comprises an additive constant, the linear terms, (3 variables), the quadratic terms (3 variables) and any additional cross terms which add to the prediction. There are three possible cross terms (FI*F2,Fl*F3,F2*F3). A cross term enters the equation if it explains additional variability left unexplained by the linear and square terms. Each cross term is correlated with the residuals from the equation. The cross term showing the highest absolute correlation (with the residuals) is entered into the equation, if the absolute correlation exceeds 0.15. A new equation is then developed, incorporating all linear, square, and the just added cross terms. New residuals are then calculated. This process of adding additional cross terms continues until no cross term shows a correlation of 0.15. At that point the modeling stops. The key aspect of the modeling exercise of interest here is whether or not one can create models for image attributes with the same degree of predictability and goodness-of-fit that one creates for liking, sensory, and performance attributes. Two indices of goodness-of-fit suggest no, as Table 4 shows. The first index is the multiple R2, or the percent of variability accounted for by the equation. This value is consistently lower for the image attributes than it is for the sensory, liking, and performance attributes. [It makes sense for the sensory attributes to have a higher R2, because the factors were obtained from these attributes in the first place, but the liking and performance attributes had nothing to do with the

24

H.R. MOSKOWITZ

principal components analysis]. The second index is the standard error of the regression, which shows the variability around the line of best fit. The standard error of regression generally shows more variability around the regression models for image and liking, and less for sensory attributes. Both goodness-of-fit statistics thus suggest that the image attributes fare poorly when the researcher attempts to model them.

DISCUSSION Concept Formation and Product Evaluation Product developers often create products on the basis of concepts which score well in concept screening. [Concept screening is a well accepted technique in which panelists are exposed to descriptions of the product, viz., concepts, and asked to rate their interest in the concept based upon what they read]. The concepts which win in the study are then delivered to the R&D staff, who are instructed to create products fitting the concept. Quite often these concepts comprise long descriptions of the product couched in image attributes. The product developer attempts to understand what is meant by image-laden product concept, and then formulates the prototypes to fulfill the concept. The next step in development comprises a concept-product test to measure how well the prototype product delivers against the promises made by a concept. Yet, as this study shows, for simple concepts, represented by phrases (e.g., for facial use), the developer cannot easily create a sensory profile corresponding to that concept, either in univariate modeling (concept attribute versus single sensory attribute), or in multivariate modeling (concept attribute versus a set of factor scores). If there were a statistically strong relation between image attributes and sensory attributes, then we would feel more comfortable using image attributes in developmental research. How does the lack of groundedness for image attributes affect the market researcher who uses concept-product fit tests to help determine whether or not a product lives up to the concept? The answer is simply that the consumer data may have to be believed (primarily because those data are the only valid evaluations), but at the same time there may be no way to anchor the consumer image ratings to any sensory ratings of the same product. On the Potential Incorrectness of Staged Product Development One of the key implications of this paper is that staged product development in the commercial realm may be proceeding incorrectly. A lot of product development proceeds in a carefully staged fashion, first with concept screening,

SENSORY, LIKING AND IMAGE ATTRIBUTES

25

TABLE 4. GOODNESS-OF-FIT OF MODELS RELATING ATTRIBUTES TO SENSORY-BASED SOAP FACTORS Attribute Multiple R' Standard Error Of Regression

Sensory
Smooth Heavy Large FragranceISkin Slippery Lather Creamy FragranceIWet Image Facial vs Body Ordinary - Special Uses Unique Older vs Young Liking Lather FragranceISkin Fragrance/Wet Purchase Intent Size Shape Performance Refreshes Moistens Skin Enhances Skin Feel Softens Skin Cleans Feels Comfortable During Use Rises Off Completely 0.73 0.72 0.71 0.68 0.65 0.49 0.35 4.38 4.45 4.87 4.94 4.09 9.03 4.43 0.81 0.74 0.73 0.59
0.50

0.91
0.90

3.93 3.23 4.59 4.02 3.85 4.37 5.85 5.46

0.89 0.88 0.84 0.80 0.80 0.80

0.35 0.27 0.17 0.17

6.99 6.32 9.39 3.78

4.54 6.13 7.03 10.51 7.82 10.95

0.48

26

H.R. MOSKOWITZ

then with concept optimization, followed by concept-product fit, and then the product launch. All too often these stages proceed sequentially, and do not overlap. The potential problem, as shown here, and also in other papers with condiments (Moskowitz and Grander 1997) is that there may be no clear relation between the image of the product and the actual delivery of the product. That is, the concept may promise one thing, but on a sensory basis the consumer may want something entirely different. Since there is no clear link between image attributes (which can be considered to be mini-concepts) and sensory attributes, the developer cannot anchor his work. In the case of condiments there was a modest link between image attributes and sensory attributes. The sensory expectation on the basis of the image oriented concept was shown to be more mainstream and far more constrained for the developer than what was found when the developer did the actual product study. Therefore, when the product developer began with results from concept evaluation (comprising images), the direction was much narrower and far less optimal than it could have been. The problem is far worse in the case of soap as discussed here. The image of the soap product may actually have little to do with the product itself. The formulator may have to create one type of product for sensory acceptance, and the marketer may have to select a completely different concept (or image profile) for concept acceptance. This paradox goes right to the heart of current day product development procedures which begin with concepts, and narrow down these concepts to promising ones. So far so good. However, the winning concepts are then given over to the product developer to realize in an actual formulation. At this point the lack of linkage between sensory attributes and image attributes strands the product developer in a trackless void, with only intuition to carry him through to the actual formulation. Next Steps for a Sensory-Image Lexicon Image terminology appears to be at the same stage of development as descriptive language was five decades ago. There are a host of image terms, used in different ways by consumers, without any clear references to these terms in the physical world. The next research steps may involve a collection of these terms, and the development of a lexicon with reference standards. What will make matters more difficult is that these terms may differ from country to country, necessitating cross cultural research, and a three dimensional grid (term x country x reference standard).

SENSORY, LIKING AND IMAGE ATTRIBUTES

21

REFERENCES ASTM. 1995. FlavLex, Version 1.15 (G.V. Civille & B.G. Lyon; copyright Softex, Lancaster), American Society for Testing and Materials, West Conshohocken. MOSKOWITZ, H.R. 1981. Relative importance of perceptual factors to consumer acceptance: Linear versus quadratic analysis. J. Food Sci. 46,244248. MOSKOWITZ, H.R. 1985. New Directions in Product Testing and Sensory f Analysis o Foods. Food & Nutrition Press, Trumbull, CT. MOSKOWITZ, H.R. 1994. Product testing 2: Modeling versus mapping and their integration. J. Sensory Studies 9, 323-336. MOSKOWITZ, H.R. 1995. Experts versus consumers. J. Sensory Studies 11, 19-35. MOSKOWITZ, H.R. and GANDLER, A. 1997. Segmenting markets by communication versus sensory preferences: The tongue has its reasons that the mind does not know. Presented at the 1997 Meeting o the Market Res. f Soc., Birmingham. MOSKOWITZ, H.R. and RABINO, S. 1994. Sensory segmentation: An organizing principle for international product concept generation. J, Global Marketing 8, 73-93. STEVENS, S .S. 1975. Psychophysics: An Introduction to Its Perceptual, Neural and Social Prospects. John Wiley, New York. STEVENS, S.S. and GUIRAO, M. 1964. The scaling of apparent viscosity. Science 145, 1157-1158. SYSTAT. 1994. SYSTAT, The System for Statistical Analysis. Evanston. SZCZESNIAK, A.S., LOEW, B.J. and SKINNER, E.Z. 1975. Consumer texture profile technique. J. Food Sci. 40, 1253-1256.

Вам также может понравиться