An Exploratory Investigation of Holistic and Analytic Modes of Product Perception

Hans Baumgartner, Pennsylvania State University
ABSTRACT - A methodology for studying holistic and analytic product perception is described, and some conditions under which products are perceived holistically (i.e., as unitary entities) or analytically (i.e., as composites of individual components) are specified. The results of a study in which subjects have to classify triads of sweaters shows that the proposed procedure is useful for studying holistic versus analytic product perception and that consumers' classification behavior is systematically related to their motivational task set (spontaneous versus meticulous impression formation) and two individual difference variables (need for cognition, style of processing).
[ to cite ]:
Hans Baumgartner (1993) ,"An Exploratory Investigation of Holistic and Analytic Modes of Product Perception", in NA - Advances in Consumer Research Volume 20, eds. Leigh McAlister and Michael L. Rothschild, Provo, UT : Association for Consumer Research, Pages: 673-677.

Advances in Consumer Research Volume 20, 1993      Pages 673-677

AN EXPLORATORY INVESTIGATION OF HOLISTIC AND ANALYTIC MODES OF PRODUCT PERCEPTION

Hans Baumgartner, Pennsylvania State University

ABSTRACT -

A methodology for studying holistic and analytic product perception is described, and some conditions under which products are perceived holistically (i.e., as unitary entities) or analytically (i.e., as composites of individual components) are specified. The results of a study in which subjects have to classify triads of sweaters shows that the proposed procedure is useful for studying holistic versus analytic product perception and that consumers' classification behavior is systematically related to their motivational task set (spontaneous versus meticulous impression formation) and two individual difference variables (need for cognition, style of processing).

INTRODUCTION

When a consumer perceives a product, his or her perceptual experience may consist in a holistic apprehension of the stimulus as a unitary entity, in an analytic registration of various product attributes, or a mixture of both. Later on I will define the terms holistic and analytic perception operationally, but for now it suffices to say that when consumers look at products holistically, they form a global impression of the product as a unitary whole, whereas when consumers look at products analytically, they view the product as a composite of individual components.

How sensory input is encoded (i.e., in a holistic or analytic fashion) depends on aspects of the stimulus (e.g., whether an object is described verbally or presented visually; cf. Holbrook and Moore 1981), on characteristics of the person perceiving the stimulus (both temporary individual differences such as a person's mood state and chronic individual differences such as a consumer's general tendency to structure the environment in a global or differentiated fashion; cf. Murray et al. 1990; Witkin et al. 1962), on the kind of task a person is engaged in (e.g., whether two or more products are considered simultaneously or sequentially; cf. Holbrook and Moore 1981), and on the context in which the task is to be accomplished (e.g., whether or not there is time pressure; cf. Smith and Kemler Nelson 1984).

This paper reports an exploratory investigation of the phenomenon of holistic versus analytic perception of stimuli in the context of product categorization. My objectives are twofold. The primary goal is to explore the applicability of a methodology used successfully by researchers in psychology (e.g., Garner 1974; Smith and Kemler Nelson 1984; Ward 1980, 1983) to distinguish between holistic and analytic modes of product perception. Traditionally, this problem (if addressed at all in consumer research) has been approached by use of regression analysis with interaction effects (e.g., Holbrook and Moore 1981). Because of the robustness of the linear model, this may not be the most effective procedure. On the other hand, existing research in psychology has used rather artificial stimuli (e.g., squares varying in size and brightness), and it is not obvious that the methodology is applicable to more 'real' product stimuli.

A secondary goal is to look at several variables that are expected to influence whether product stimuli are processed holistically or analytically. Evidence for differences in mode of processing as a function of these variables would attest to the usefulness of the proposed methodology and also point to the need for increased attention to whether consumers perceive product stimuli in a holistic or analytic fashion.

PRIOR RESEARCH

Most research on product perception in consumer behavior has explicitly or implicitly espoused an analytical view of the world. Products are often regarded as bundles of attributes, and models of attitudes (Wilkie and Pessemier 1973) and judgment and decision making (Bettman 1979) - the two dominant research streams in the field - are built on this fundamental assumption. Thus, attitudes toward products are assumed to derive from evaluations of product attributes, and product choices are thought to be based on comparing alternative products on their various attributes.

Only recently have researchers begun to look at alternative conceptualizations. For example, in the area of categorization several researchers have discussed holistic forms of comparisons (based on overall similarity between objects) of a stimulus to category prototypes or exemplars (see Cohen and Basu 1987 for a review). This interest in holistic approaches to categorization has spilled over to the attitude area, where researchers have proposed affect generation mechanisms in which a product is endowed with an evaluation of a product category to which it is similar (cf. Sujan 1985; Wright 1975). In a related vein, in the study of judgment and decision-making, researchers have examined the notion of configural judgments, the basic idea being that overall evaluations of products are more than just the sum of individual attribute evaluations (see, for example, Holbrook and Moore 1981). On balance, however, holistic approaches to product perception have not figured prominently in consumer research.

Despite the dearth of empirical work on the topic, the issue of when consumers process product information holistically or analytically is clearly important. Product perceptions form the basis for subsequent higher-order cognitive processes such as attitude formation and judgment and decision making. As indicated, most models in these areas of research explicitly or implicitly assume that product perceptions are analytic. The present study will provide evidence that analytic processing is restricted to certain situations and that, in the future, consumer researchers will have to concern themselves more with cases in which product perceptions are actually holistic.

Research shows that holistic processing occurs frequently under conditions in which cognitive resources are limited (e.g., when people are not able, not motivated, or do not have the opportunity to engage in analytic processing; see Alba and Hutchinson, 1987, and Cohen and Basu, 1987, for reviews). Many consumer situations are exactly of this type (Alba and Hutchinson 1987). The research described in this paper makes an important first step in outlining an approach to studying holistic and analytic product perception and in delineating some conditions under which each process is most likely to occur.

THEORETICAL DEVELOPMENT

The impetus for the proposed research comes from work in perceptual and cognitive psychology on integral and separable attribute combinations (cf. Garner 1974, Lecture 5). Attributes are called integral if they combine to form an entity that is ordinarily processed in a holistic fashion. Attributes are called separable if they produce an entity that is generally processed analytically. Instances of integral attribute combinations are difficult to come by, the premier example being brightness and saturation (two of the three psychological attributes of colors). Instances of separable attribute combinations are more common, form and size being a frequently cited example.

Research has shown that judgements of similarity are based on overall similarity in the case of integral stimuli and are derived from shared components for separable stimuli. There is also evidence that children treat as integral attribute combinations that are perceived as separable by adults (e.g., Ward 1980). Furthermore, it has been found that under certain conditions - when there is a time constraint, when attention is diverted by a concurrent task, or when people are instructed to respond impressionistically (cf. Smith and Kemler Nelson 1984; Ward 1983) - even adults treat separable stimuli as integral.

In the study described below I will investigate the effects of a consumer's motivational task set (spontaneous versus meticulous impression formation) and two individual difference variables (need for cognition, style of processing) on mode of product perception for product stimuli that are separable. As pointed out by Mervis and Rosch (1975), aspects of stimuli that are normally called attributes are mostly separable, and integral attributes are at a lower level of abstraction that is probably not very relevant to marketing (e.g., brightness and saturation as dimensions of color). Furthermore, if stimuli that are in principle separable are processed as unitary entities under certain circumstances, this would provide fairly strong evidence that products composed of attributes at a level of abstraction useful to marketers are not always perceived analytically and that consumer researchers will have to deal more explicitly with situations in which products are perceived holistically.

The experimental paradigm used to distinguish holistic and analytic processing is based on restricted classifications of triads of stimuli (cf. Garner 1974, Lecture 5). Subjects are presented with sets of three objects and asked to select the two that "go the most together" (cf. Smith and Kemler Nelson 1984). In the simplest case the stimuli are constructed from two underlying dimensions. One pair of objects in a triad is identical on one dimension but substantially different on the other. A second pair is slightly different on both dimensions, but overall the two objects are quite similar. Classifications on the basis of overall similarity are reflective of holistic processing, whereas classifications based on identical values on one dimension are indicative of analytic processing. The third possible classification is considered haphazard.

Using this procedure, the effects of the following variables on mode of perception are investigated. First, the way in which consumers form an impression of the product stimuli should be related to classification behavior. Specifically, it is expected that subjects who are told to classify without much thinking and to rely on first impressions (spontaneous motivational task set) will provide more holistic (less analytic) classifications than subjects who are told to be careful and to take all the time needed (meticulous motivational task set). This prediction follows from the work of Smith and Kemler Nelson (1984, Experiment 5) and Ward (1983).

In addition, some subjects will not be given instructions as to how to classify the stimuli (control group). For these subjects (and possibly for subjects in one of the other two conditions whose task set resembles the one for control subjects), it is hypothesized that need for cognition (a person's preference for engaging in and enjoying thinking; cf. Cacioppo and Petty 1982) and style of processing (a person's preference for engaging in visual versus verbal processing of information; cf. Childers, Houston, and Heckler 1985) will be related to mode of classification. Specifically, it is expected that need for cognition will be positively correlated with the extent to which classifications are analytic, and it is also hypothesized that being a visualizer (verbalizer) will encourage more holistic (analytic) classification behavior. The rationale for the former hypothesis comes from the fact that need for cognition assesses enjoyment of "effortful analytic activity" (Petty and Cacioppo 1986, p. 151, emphasis added) and that, as discussed by Cacioppo and Petty (1982), need for cognition is related to field independence/dependence (the tendency of a person to structure the environment in a differentiated or global fashion, cf. Witkin et al. 1962), an individual difference variable which has been found to determine preference for analytic and holistic processing. The rationale for the latter hypothesis is that preference for engaging in visual processing of information involves imagery, which is holistic (MacInnis and Price 1987), and that imagery processing should lead to more holistic classifications than verbal processing (Holbrook and Moore 1981).

METHOD

Subjects

A total of 48 undergraduate students (25 females, 23 males) in two sections of a consumer behavior class participated in the study during class time. Subjects were assigned randomly to one of three treatment conditions: spontaneous motivational task set, meticulous motivational task set, and control group. Sixteen subjects participated in each of the three conditions.

Stimuli

Pictorial sweater designs similar to those used by Holbrook and Moore (1981) were developed which varied two attributes on four levels: size of the sweater (size) and density of the dot pattern (pattern). Convenience samples of students were used to calibrate the spacing of the four attribute levels.

Triads of sweaters were constructed such that two of the sweaters in each triad were identical on one dimension but were maximally different on the other dimension, whereas another pair of sweaters within each triad differed by only one scale step on both dimensions so that overall the two sweaters were quite similar although the two sweaters differed on both dimensions. With four levels on each of two attributes, this procedure yields 24 triads of sweaters; in 12 of these, two sweaters were identical on the size dimension, and in the other 12 there was an identity on the pattern dimension.

Stimulus booklets were then constructed which contained the 24 triads of sweaters, each on a separate page, with size and pattern identities in alternate order. The arrangement of the three sweaters in each triad was such that holistic, analytic, and haphazard classifications involved the left and middle, left and right, and middle and right sweaters equally often.

Procedure

Subjects received a booklet that contained all experimental tasks. They were told that they would be shown three objects at a time side by side (arranged in the order 1-2-3) and that they had to decide which two of the three objects went the most together. They were to indicate their decision by putting down 12, 13, or 23, depending on whether they thought the left and middle, left and right, or middle and right sweaters went the most together. It was stressed that there were no right or wrong answers, and subjects were also told that at the beginning and at the end of the classification task they would be required to mark down the time (to the nearest minute) from a big clock placed on a table in front of the classroom.

Immediately before subjects started the classification task, their motivational task set was manipulated. In the spontaneous motivational task set condition, subjects were told, "When grouping the three objects, make your classification without thinking about it. Just give your first impression, just let whatever happens happen." In the meticulous motivational task set condition, subjects were told, "When grouping the three objects, carefully decide which two objects should go together. Be meticulous and careful, taking all the time you need." In the control condition, no instructions were given to subjects.

Subjects then put down the time to the nearest minute and classified the 24 sweaters. When they were done, they again put down the time to the nearest minute, and they indicated how important they thought the dimensions of size and pattern had been in their classifications of the sweaters (11-point scale from -5 to +5, with endpoints of 'size much more important' and 'pattern much more important').

Finally, subjects completed the 34-item need for cognition scale (Cacioppo and Petty 1982) and the 22-item style of processing scale (Childers, Houston, and Heckler 1985). Higher scores on the two instruments are indicative of greater need for cognition and a preference for visual rather than verbal processing of information.

RESULTS

Manipulation check

The time taken to perform the classification task was used to check whether the manipulation of motivational task set had the intended effect. Statistical tests of mean differences among conditions were performed using dummy variable regression (with levels of significance referring to one-sided tests because of the directional nature of all the tests).

The overall regression was significant (F(2,44)=5.99, p<0.01), and as expected, subjects who were told to classify spontaneously completed the task significantly faster than subjects who were asked to be meticulous (means of 4.56 min and 6.73 min, respectively, t44=3.36, p<0.01). Subjects in the control condition (mean of 6.06 min) also took significantly longer than spontaneous subjects (t44=2.36, p<0.05), and their classification time did not differ significantly from that of meticulous subjects (t44=1.04, n.s.). In the context of this experiment, the 'natural' classification strategy adopted by subjects in terms of time taken was thus to be meticulous.

Effect of motivational task set on mode of classification

In classifying a given triad of sweaters, there are three possible outcomes: holistic, analytic, or haphazard. Before looking at the hypotheses of substantive interest, it is necessary to check that (1) the incidence of haphazard classifications is low and that (2) a spontaneous motivational task set does not primarily lead to more haphazard classifications.

Overall, the mean number of haphazard responses was 2.67 out of 24 (11 percent), which is low. Furthermore, there were no significant differences in the mean number of haphazard responses by condition (means of 2.50, 2.75, and 2.75 for spontaneous, meticulous, and control subjects, respectively).

An index reflecting mode of classification for each subject was constructed as the ratio of number of analytic classifications to holistic plus analytic classifications (summed over all 24 triads of sweaters and excluding haphazard responses). The resulting index ranges from 0 to 1, with 0 indicating entirely holistic processing and 1 indicating entirely analytic processing. Across all subjects, the mean on the classification index was 0.36, with a range from 0 to .96. Thus, on average 36 percent (64 percent) of the classifications were analytic (holistic). Since the stimuli were presented in visual form in all conditions, this may have resulted in relatively holistic classifications (Holbrook and Moore 1981). However, it is not the overall classification behavior across conditions that is of interest here, but the relative proportion of holistic and analytic classifications by condition.

Since control subjects took just as long to classify the sweaters as meticulous subjects, regression analysis with contrast coding was used to test the following two hypotheses (which involve orthogonal contrasts): control and meticulous subjects will be equally holistic/analytic in their classifications, and control and meticulous subjects combined will provide significantly more analytic (less holistic) classifications than spontaneous subjects.

Although the overall regression was not significant, the specific effects tended to confirm this prediction. Control and meticulous subjects did not differ significantly in their classification behavior (means of .40 and .38, respectively, t45=.23, n.s.), and control and meticulous subjects classified in a more analytic/less holistic fashion than spontaneous subjects (means of .39 for control/meticulous condition and .29 for spontaneous condition, t45=1.53, p<0.10). It was also checked whether pattern or size identities led to more analytic classifications, but there were no differences.

The pattern of the means is clear-cut, although the statistical significance of the result is admittedly not very strong. In part this is probably due to the small sample size. Another reason may be that the manipulation of motivational task set was too weak.

Effect of classification time on mode of classification

To get further insights into subjects' classification behavior, an internal analysis was conducted on the relationship between classification time and mode of classification. The correlation between the two variables was .25 (t45=1.73, p<0.05), indicating that the longer a person took to classify the sweaters, the more analytic/less holistic his or her responses tended to be. There were no differences in the strength of this relationship by condition.

The effects of need for cognition and style of processing on mode of classification

Regressions of the mode of classification index on the dummy-variable (or contrast) coded manipulations, need for cognition (or style of processing) and their interactions were conducted to test whether the two individual difference variables have an effect on classification behavior and whether the strength of the relationship differs by condition.

Overall, the correlation between proportion of holistic/analytic classifications and need for cognition was .24 (t46=1.71, p<.05). Furthermore, the correlations by condition showed an interesting pattern: r=-.04 for spontaneous condition (t14=-.17, n.s.), r=.30 for meticulous condition (t14=1.18, n.s.), and r=.73 for control condition (t14=3.95, p<0.01). Based on the results of an overall regression analysis (F(5,42)=2.45, p<0.05), the difference in the magnitude of the correlations between meticulous and control groups was of borderline significance (t42=1.51, p<0.10), while the difference between control and spontaneous groups was clearly significant (t42=2.36, p<0.05). The relationship between classification behavior and need for cognition in the combined meticulous and control groups (r=.44) was significantly higher than the correlation for the spontaneous group (t42=2.18, p<0.05).

The relationship between proportion of holistic/analytic classifications and style of processing for the entire sample was not significant (r=-.14), but by condition there was again an interesting pattern: r=.38 (t13=1.47, p<.10) for spontaneous condition, r=-.28 for meticulous condition (t13=-1.11, n.s.), and r=-.44 for control condition (t13=-1.78, p<0.05). Based on the results of an overall regression analysis (F(5,40)=1.63, n.s.), the strength of the relationship between classification behavior and style of processing was not significantly different for meticulous and control subjects (t40=-.07, n.s.). However, the correlation of -.35 for both groups combined indicates that visualizers were more holistic in their classifications than verbalizers (t29=-2.03, p<.05). Subjects in the combined meticulous and control groups differed significantly from subjects in the spontaneous group (t40=2.27, p<0.05).

Mode of classification and dimensional importance

Subjects were asked to indicate how important they thought the dimensions of size and pattern had been in their classification behavior, relative to each other. It was expected that subjects who indicated that they had relied strongly on one of the two dimensions in their classifications would show more analytic responses. Since the importance scale had endpoints of 'size much more important' and 'pattern much more important,' the hypothesis implies a U-shaped relationship between classification behavior and self-ratings of importance.

The predicted result was obtained. In a regression of proportion of holistic/analytic responses on self-rated importance (F(2,45)=4.53, p<0.05), both the linear term (t45=-2.35, p<0.05) and the quadratic term (t45=2.07, p<0.05) were significant in the expected direction. The minimum of the function was at 1.38 on a -5 to +5 scale. The hypothesis can also be tested by correlating the absolute value of the importance scale with classification behavior; the correlation was significant (r=.30, t46=2.16, p<0.05).

DISCUSSION AND CONCLUSION

The substantive findings of this study can be summarized as follows. First, subjects who were instructed to classify without much thinking and to rely on first impressions categorized product stimuli in a more holistic (less analytic) fashion than subjects who were told to be careful and to take all the time they needed. Subjects who were not given any instructions provided responses similar to meticulous subjects. In addition, an internal analysis of the data showed that the time spent on classifying the stimuli was positively related to the extent to which classifications were analytic. These findings suggest that, as hypothesized by Alba and Hutchinson (1987) and Cohen and Basu (1987) on the basis of research in psychology, product perceptions will be less analytic in situations in which cognitive resources are limited (i.e., when consumers are not able, not willing, or do not have the opportunity to expend the time and effort necessary to form a thorough impression of a stimulus).

Second, two individual difference variables C need for cognition and style of processing C were found to have an effect on how product stimuli were perceived. Specifically, consumers high in need for cognition and verbalizers tended to view products more analytically (less holistically) than consumers low in need for cognition and visualizers. These relationships existed most clearly when subjects were allowed to follow their natural inclinations in classifying the stimuli (control group) or when subjects were told to be meticulous (which tended to be the natural tendency in the context of this study, in the sense that the classification behavior of control subjects, whose motivational task set was not manipulated, was similar to that of meticulous subjects). When subjects were instructed to rely on first impressions, need for cognition and style of processing had no effects on mode of product perception.

In addition to these substantive findings, the present study indicates that the triad classification task is a useful methodology for distinguishing between holistic and analytic modes of product perception and thus for investigating which factors influence whether consumers perceive products holistically or analytically. With the help of this procedure, researchers should be able to examine a variety of other factors that are expected to have a bearing on whether products are perceived holistically or analytically.

The basic argument of this research is that whether products are perceived holistically or analytically depends on a variety of factors, and the study reported in this paper makes an important step in delineating some conditions under which the two forms of processing are likely to occur. Some authors have suggested that the conditions that favor holistic processing occur frequently in everyday consumer behavior (e.g., Alba and Hutchinson 1987). The implications for marketing are clear: In the future, researchers will have to concern themselves more with the case in which product perceptions are holistic and develop models of attitude formation and judgment and decision making that are applicable in this situation.

As always, certain limitations of the study should be kept in mind when interpreting the findings. First, the difference in classification behavior between spontaneous and meticulous subjects was not particularly strong. One possible reason for this result is that the manipulation of motivational task set was not strong enough. Although subjects in the two conditions differed significantly in terms of the time taken to classify the stimuli, the absolute magnitude of these differences was not very large. Another reason for the borderline significance of the statistical test is that the sample size was too small. Second, the research was conducted in a classroom setting, and it would be desirable to replicate the study in a more controlled experimental environment. Third, the research should be extended using other stimuli and other subject populations. Despite these shortcomings, the results thus far look encouraging and the implications of the difference between holistic and analytic perception for attitude formation and judgment and decision making are profound so that further investigations of the topic seem warranted.

REFERENCES

Alba, Joseph W. and J. Wesley Hutchinson (1987), "Dimensions of Consumer Expertise," Journal of Consumer Research, 13 (March), 411-454.

Bettman, James R. (1979), An Information Processing Theory of Consumer Choice. Reading, MA: Addison-Wesley.

Cacioppo, John T. and Richard E. Petty (1982), "The Need for Cognition," Journal of Personality and Social Psychology, 42 (January), 116-131.

Childers, Terry L., Michael J. Houston, and Susan E. Heckler (1985), "Measurement of Individual Differences in Visual Versus Verbal Information Processing," Journal of Consumer Research, 12 (September), 125-134.

Cohen, Joel B. and Kunal Basu (1987), "Alternative Models of Categorization: Toward a Contingent Processing Framework," Journal of Consumer Research, 13 (March), 455-472.

Garner, Wendell R. (1974), The Processing of Information and Structure. Potomac, MD: Erlbaum.

Holbrook, Morris B. and William L. Moore (1981), "Feature Interactions in Consumer Judgments of Verbal Versus Pictorial Presentations," Journal of Consumer Research, 8 (June), 103-113.

MacInnis, Deborah J. and Linda L Price (1987), "The Role of Imagery in Information Processing: Review and Extensions," Journal of Consumer Research, 13 (March), 473-491.

Mervis, Carolyn B. and Eleanor Rosch (1981), "Categorization of Natural Objects," Annual Review of Psychology, 32, 89-115.

Murray, Noel, Harish Sujan, Edward R. Hirt and Mita Sujan (1990), "The Effects of Mood on Categorization: A Cognitive Flexibility Hypothesis," Journal of Personality and Social Psychology, 58 (September), 411-425.

Petty, Richard E. and John T. Cacioppo (1986), "The Elaboration Likelihood Model of Persuasion," Advances in Experimental Social Psychology, 19, 123-205.

Smith J. David and Deborah G. Kemler Nelson (1984), "Overall Similarity in Adults' Classification: The Child in All of Us," Journal of Experimental Psychology: General, 113 (January), 137-159.

Sujan, Mita (1985), "Consumer Knowledge: Effects on Evaluation Strategies Mediating Consumer Judgments," Journal of Consumer Research, 12 (June), 31-46.

Ward, Thomas B. (1980), "Separable and Integral Responding by Children and Adults to the Dimensions of Length and Density," Child Development, 51, 676-684.

Ward, Thomas B. (1983), "Response Tempo and Separable-Integral Responding: Evidence for an Integral-to-Separable Processing Sequence in Visual Perception," Journal of Experimental Psychology: Human Perception and Performance, 9 (1), 103-112.

Wilkie, William L. and Edgar A. Pessemier (1973), "Issues in Marketing's Use of Multiple-Attribute Attitude Models," Journal of Marketing Research, 10 (November), 428-441.

Witkin, H.A., R.B. Dyk, H.F. Faterson, D.R. Goodenough and S.A. Karp (1962), Psychological Differentiation: Studies of Development, New York: Wiley.

Wright, Peter L. (1975), "Consumer Choice Strategies: Simplifying Versus Optimizing," Journal of Marketing Research, 11 (February), 60-67.

----------------------------------------