Children's use of perceptual cues in product categorization

ABSTRACTThis study examines how young children use perceptual cues in categorizingproducts. Specifically, we examine the notion that young children use perceptualcues in a more informed fashion than is generally assumed by traditional view-points on cognitive development. Data from the three age groups—four- andfive-year-olds, six- and seven-year-olds, and adults—confirm that even youngchildren can use perceptual cues in a manner that is diagnostic of productcategory membership and that this ability increases with age.

Children's Use ofPerceptual Cues inProduct Categorization

Deborah Roedder JohnCarlson School of Management,University of MinnesotaMita SujanSchool of Business, Pennsylvania State University

Understanding how consumers categorize products, and how these cate-gorization schemes affect the processing of product-related information,has emerged as an important topic for consumer research (Alba & Chat-topadhyay, 1985; Cohen & Basu, 1987; Sujan, 1985; Sujan & Bettman,1989; Sujan & Dekleva, 1987). Despite the growing interest in this area,however, little attention has been directed toward the topic of productcategorization in children. Questions pertaining to the development ofproduct categories in children, such as how product categories are formedand whether they differ depending on the age of the child, are certainlyworthy of investigation considering the importance of the children's marketin terms of absolute size, spending power, and purchase influence.

While consumer researchers have largely ignored these questionsregarding product categorization in children, there is an abundant literaturein developmental psychology addressing many of the same issues. Findings

Psychology & Marketing Vol. 7(4): 277-294 (Winter 1990)© 1990 John Wiley & Sons, Inc. CCC 0742-6046/90/040277-018$04.00

CHILDREN'S USE OF PERCEPTUAL CUES

in this area suggest that children's abilities to classify objects emerge at avery young age, as young as 18 to 24 months of age (e.g., Cohen & Younger,1983; Fenson, Vella, & Kennedy, 1989; Sugarman, 1983). Furthermore,evidence points to a gradual development in children's abilities to classifyand categorize from late infancy to late childhood. Of particular interestto marketing, the findings point to a change in the bases children use toform categories between preschool and mid-elementary school (Denney,1974; Markman, 1980; Markman & Callahan, 1983). During this period,children begin to define categories on the basis of underlying or functionalattributes of objects rather than perceptually-salient attributes (Melkman& Deutsch, 1977; Melkman, Tversky, & Baratz, 1981; Olver & Hornsby,1966; Saltz, SoUer, & Sigel, 1972; Tversky, 1985). Children learn to groupobjects according to attributes that suggest taxonomic relationships (e.g.,belts and socks can be worn as items of clothing), attributes that indicatethe relationship of categories to one another (e.g., fruit juices and softdrinks vary on the attribute of naturalness), and attributes inherent to thecore concept of categories (e.g., taste, more than color, is central to thecategory of soft drinks). Prior to the use of attributes such as these, youngchildren typically rely on perceptual attributes that are visually dominant,such as shape, size, or color.

Evidence consistent with these observations has been reported in arecent study of product categorization in children (John & Sujan, 1990).In this study, children were shown triads of products from the beverageor cereal category. One of the items was identified as a target, with theother two items in the triad sharing perceptual and/or underlying featureswith the target. Children were asked to identify which of these two itemswas "most like" the target and were questioned about the basis for theirdecision. Data from children ages four to ten indicated that the use ofperceptually salient cues as a basis for grouping products decreased withage, whereas the use of underlying attributes for categorizing productsincreased with age.

Although these findings converge with those reported by develop-mental psychologists, it is not entirely clear from either line of researchhow this evidence should be interpreted in characterizing children's abilitiesto categorize at different ages. Traditionally, the use of perceptual cues toclassify items has been viewed as a relatively unsophisticated approach,relied upon by young children who were "perceptually bound" to concretefeatures of their environment and unable to reason on a more abstractlevel (e.g., Bruner, Olver, & Greenfield, 1966; Piaget, 1970). However,a current theme of much developmental research is that these broad char-acterizations of the young child may be too harsh and that the young childpossesses some of the same relevant reasoning skills as older children andadults, albeit in more limited degrees (e.g., Flavell, 1985, and Gelman &Baillargeon, 1983). Which of these views best represents the state of young

278 ~ ~ ~ WINTER 1990

CHILDREN'S USE OF PERCEPTUAL CUES

children's abilities has not been assessed with regard to the developmentalshift from the use of perceptual to underlying cues in categorization.

In this paper, we take a closer look at children's use of perceptualcues in product categorization to examine how the use of these cues shouldbe interpreted in terms of young children's classification abilities. Specif-ically, we examine the notion that young children may use perceptual cuesin a somewhat more informed or sophisticated fashion than that generallyassumed by more traditional viewpoints on cognitive development. Evi-dence to this effect would call into question previous characterizations ofyoung children's categorization abilities as well as suggesting a differentinterpretation for the developmental shift from the use of perceptual cuesto underlying cues in product categorization.

In examining these possibilities, we report the results of an empiricalinvestigation of how young children use perceptual cues in categorizingproducts. In presenting this study, the first section of the paper presentsthe conceptual background behind our ideas regarding the locus of agedifferences in categorization. Based on this discussion, the following sectionpresents our hypotheses regarding children's use of perceptual cues incategorizing products. The next two sections describe the experimentalmethodology used to examine these issues and the empirical results ob-tained with our sample. The final section discusses the implications of thesefindings in terms of what mechanisms are responsible for age differencesin the types of cues children use to categorize products and what thesedifferences mean for characterizing young children's abilities to categorizeproducts.

EXPLAIIVIJVG AGE DIFFERENCES INCATEGORIZATION BASES

Classical Views

A variety of explanations have been advanced to answer the questionof why young children initially rely on perceptual cues to classify objectsbut then shift to using more underlying or functional features as they growolder. Of these, the earliest explanation of why children rely on perceptualcues centered on the idea that young children are "stimulus bound" or"perceptually bound" (Bruner et al., 1966; Piaget, 1970; Olver & Hornsby,1966; Saltz et al., 1972). According to this view, young children are lockedin a concrete immediate environment, unable to think at a more abstractlevel. Being locked into their immediate environment, young children areunable to suppress highly salient perceptual cues in favor of more validunderlying cues in classifying objects.

PSYCHOLOGY & MARKETING/VOL. 7, NO. 4 279~

CHILDREN'S USE OF PERCEPTUAL CUES

Although highly salient visual cues can attract the attention of youngchildren, several recent investigations cast doubt on the notion of "per-ceptual boundness" as an explanation for age differences in categorization.For example, Tversky (1985) compared the use of perceptual and under-lying cues for categorizing objects which were presented to children orallyor pictorially. Presumably, mentioning the names of the objects orally,without the visual cues typically supplied by pictures, would alleviate anydifficulties associated with interference from highly salient perceptual cues.The findings point to no difference in the extent to which perceptual cueswere used by children in either presentation condition, suggesting thatsomething other than simple perceptual salience is responsible for thereliance on perceptual cues in categorization. On quite a different front,Gelman and her colleagues (Gelman, 1988; Gelman & Markman, 1986)have demonstrated quite conclusively that younger children readily ignoreperceptual cues in favor of category membership information when drawinginferences about objects. When children are shown an object that is per-ceptually similar to another, but is described as belonging to a differentcategory, even preschoolers draw conclusions about the target object morein line with the underlying information provided by the category label thanwith the highly salient perceptual information.

Contemporary Views

In light of such evidence, more recent explanations for the shift fromperceptual to underlying cues in categorization have tended to concentrateon a quite different factor than those proposed by earlier research. Inparticular, a number of explanations related to children's developingknowledge bases have emerged as the primary focus for understandingshifts in category organization. Perhaps the simplest of these is the notionthat children begin to use underlying cues for classifying objects only whenthey have acquired the necessary functional information about objects thatcan serve as a basis for grouping. Before this point in time, younger childrenare forced to rely on perceptual attributes that are highly available andvisually salient simply because they lack knowledge of what underlyingattributes might characterize the objects being classified.

The idea that simple knowledge of facts about underlying features ofobjects is responsible for shifts in categorization does not appear to beconsistent with several lines of developmental evidence. Although it iscertainly true that unfamiliar stimuli can have a detrimental effect onyoung children's performance in a wide variety of situations (Gelman &Baillargeon, 1983), it is also the case that young children have far moreinformation at their disposal than they effectively use in many types oftasks. In categorization tasks, simple instructions to use a particular basisfor grouping can evoke the use of underlying cues in many children for

280 WINTER 1990

CHILDREN'S USE OF PERCEPTUAL CUES

whom perceptual cues appear to be the preferred basis for categorizing(John & Sujan, 1990). Even young children who group items on the basisof something other than underlying cues can justify classifications basedon underlying attributes (Smiley & Brown, 1979). And, in retrieval tasks,young children who do not spontaneously group according to underlyingfeatures can be induced to use such groupings to facilitate recall (Moely,1977). Taken as a whole, these studies undermine the argument that youngchildren do not use underlying or functional information because they donot know the basic information needed for grouping on this basis.

A more viable explanation is that children have much of the basicinformation they need to categorize on the basis of underlying features,but that this information is not well-integrated into their existing knowledgestructures (Melkman et al., 1981) or is not salient enough in the knowledgestructure to be readily activated when needed in a particular processingsituation (Bjorklund & deMarchena, 1984; Bjorklund & Thompson, 1983).In effect, the argument is that underlying or functional features are notcentral to the representations of object categories held by young children,even though much of the basic information is present in the knowledgestructure. With age and experience, underlying and functional featuresbecome more integrated into the knowledge base, making their activationrelatively automatic (Bjorklund & deMarchena, 1984).

It would appear that this last explanation has much to offer in re-conciling the fact that the use of underlying cues in categorization certainlylags the acquisition of basic knowledge in this area. What this approachdoes not offer, however, is an explanation for the process whereby un-derlying or functional features become better integrated into children'sknowledge structures as they grow older. Similarly lacking is an explanationfor why perceptual cues remain so salient in children's knowledge structuresfor such a long time.

We believe that the process of integrating different types of cues,both perceptual and underlying cues, into children's knowledge structuresregarding categories can best be understood by considering a "theory based"explanation for the use of perceptual and underlying cues. In general, a"theory based" explanation proposes that individuals develop theories ofhow different cues or features relate to the meaning of concepts and cat-egories (Medin & Smith, 1984; Murphy & Medin, 1985). In turn, thesetheories guide what particular cues are accessed when interpreting themeaning of concepts and using concepts to classify new stimuli. It is theexistence of these theories, then, which provides the mechanism wherebydifferent types of cues become more or less integrated into children'sknowledge structures at different ages.

Viewed from this perspective, children's initial reliance on perceptualcues to categorize objects can be traced to the theories they hold aboutcorrelations between perceptual features and category membership. At

PSYCHOLOGY & MARKETING/VOL. 7, NO. 4 28?"

CHILDRENS USE OF PERCEPTUAL CUES

very early ages, we would argue that children develop and hold theories,albeit primitive or naive, that relate perceptual cues to the meaning ofcategories. For example, young children might hold implicit theories suchas "purple drinks are sweet" and "drinks in cans taste good." These the-ories of covariation, though sometimes misleading, are often diagnosticbecause they reflect knowledge of the stable correlational structures in theworld (e.g., juice is purple but milk is not; sodas come in cans). Theimportant point is that, although children form their early concepts basedon perceptual rather than underlying similarity ("things that look alike—are purple—go together"), some knowledge of what these attributes meanand how these attributes relate to other attributes underlie and constrainthe use of perceptual cues from an early age. As children grow older, thesetheories become more complex, weeding out perceptual attributes that arenondiagnostic and adding new connections between underlying attributesand concept meaning. These more "sophisticated" theories provide thestrong ties between underlying features of objects and category member-ship evidenced by older children and adults.

HYPOTHESES

Adopting a theory based account for the existence of age differencesin categorization bases suggests a new direction for interpreting the use ofperceptual attributes in young children. We believe that at least two pre-dictions are forthcoming from our previous discussion of how children'simplicit theories about concepts guide their use of perceptual cues in ca-tegorizing objects. First, one can hypothesize that even young children willuse perceptual cues in a diagnostic fashion. Diagnostic use of these attri-butes implies that judgments will be based on those pieces of informationthat are most relevant for making informed or appropriate decisions.What this suggests is that, faced with different pieces of perceptual infor-mation about an object, young children will use those perceptual attributesthat are more likely to be related to the underlying meaning of categories,thereby making categorization decisions that are, at least for the most part,normatively right. Note that the alternative hypothesis is that young chil-dren's use of perceptual attributes is guided purely by perceptual salienceand hence is not diagnostic of "true" category structure. Evidence consis-tent with the notion that young children use perceptual cues in a diagnosticfashion would suggest a characterization of young children's abilities inexcess of what classical views, and even some contemporary views, ofdevelopmental differences in this area would predict.

Second, a theory based account for the use of perceptual cues suggeststhat, as the theories underlying perceptual attributes mature with age, thediagnostic use of perceptual attributes will increase. In fact, there is some

"282 WINTER 1990

CHILDREN'S USE OF PERCEPTUAL CUES

evidence that perceptual attributes continue to be used even into adulthoodfor categorization decisions. For example, Rosch and her colleagues arguethat basic level categories (such as chairs or animals) derive their "basic-ness" from shared perceptual similarity (such as shape) and that theseshared perceptual features allow for faster and more accurate categoriza-tion decisions. However, adults, unlike young children, constrain their useof perceptual similarity by knowledge of underlying attribute structuresthat define concepts. Even when underlying attributes are known to youngchildren, they may not employ these attributes to constrain the use ofperceptual cues (e.g., Carey, 1982). Thus, as theories relating perceptualcues to other attributes mature with age, such that false correlations be-tween perceptual cues and category membership are actively rejected andtrue correlations are accepted, the diagnostic use of perceptual attributesis likely to improve.

Thus, two predictions regarding the use of perceptual cues in cate-gorizing products were tested—first, that even young children use percep-tual cues diagnostically (HI), and second, that the diagnostic use of per-ceptual cues increases with age (H2).

METHOD

Subjects

A sample of young children and adults was recruited to test ourhypotheses. Although the focus was on children's use of perceptual cuesin categorizing products, an adult sample was deemed necessary to val-idate our experimental manipulations of perceptual cues that were diag-nostic versus those that were nondiagnostic. As such, the adult samplefunctioned as a baseline against which young children's responses could becompared.

ChildrenOne hundred and seventeen children were recruited by a local mar-

keting research firm in a large midwestern city. Parents received a smallmonetary incentive for their child to participate. An approximately equalnumber of boys and girls were recruited from each of two age groups: four-and five-year-olds and six- and seven-year-olds. In recruiting, precautionswere taken to ensure children's familiarity with the product classes usedin the study. Parents were asked whether their child consumed a numberof items in each product class, and recruitment was terminated in thosecases where usage was not indicated for at least half of the specific itemsin each product class. As a further check for familiarity, parents filled outa questionnaire at the research site, thereby confirming that children of all

PSYCHOLOGY & MARKETING/VOL. 7, NO. 4 283~

CHILDREN'S USE OF PERCEPTUAL CUES

ages were equally familiar with the specific products and package typesused in the study (for tests of age differences across all items, all p's >.20).

Of those recruited, four children were dropped from the study dueto problems in understanding the experimental instructions. The final sam-ple, then, contained 113 children: 59 in the four- and five-year-old groupand 54 children in the six- and seven-year-old group.

AdultsFifty college-age adults were recruited from a small midwestern city

to participate in the study. Students participated as a part of their regularclass activities and received no incentives to participate in the study.

Experimental Stimuli

Three categories of beverage items were used in constructing theexperimental stimuli: milk, juice, and soft drinks. These categories wereselected for several reasons. First, all three product categories are familiarto children and are frequently served at home, school, and in restaurants.Second, beverages in these categories come in a variety of sizes, colors,and package types, providing a number of perceptual cues that might beconsidered in testing our hypotheses. Third, the perceptual cues that dif-ferentiate beverage items offer a number that are diagnostic and nondi-agnostic in distinguishing different categories of beverages from one an-other. For example, soft drinks do not come packaged in small boxes,whereas juices do (diagnostic cue). However, both soft drinks and juicescome packaged in small glass bottles (nondiagnostic cue).

Using these beverage categories, two types of stimuli were developedto assess subjects' use of perceptual cues in categorizing products. The firstset of stimuli consisted of five beverage triads, with three beverage itemsper triad. Two of the items were real products that were highly familiarto children (e.g., apple juice, chocolate milk. Coke). These products weredisplayed in their original packaging. The third items in the beverage triadswere hypothetical products designed especially for the study. These items,referred to as "secret drinks," were designed in such a way that they sharedone perceptual feature in common with each of the two real drinks. How-ever, only one of the shared perceptual features was truly diagnostic ofcategory membership. Such a configuration allowed us to discern whethersubjects were using diagnostic or nondiagnostic perceptual cues when askedto select the real product that would be most similar to the secret drink.

For example, in one triad, the two real products were a black boxof Minute Maid Grape Juice and a red can of Coke. The secret drink wasa plain unlabeled red box. The secret drink could be matched with thegrape juice on the basis of being packaged in boxes; the secret drink could

284 WINTER 1990

CHILDREN'S USE OF PERCEPTUAL CUES

be matched with Coke on the basis of having red packages. In this example,package type (box) is more diagnostic of category membership (juice versussoft drink) than package color (red), because both juices and sodas comein red packages, but only juices come in small boxes. Thus, diagnostic useof perceptual cues would result in the secret drink being grouped with thegrape juice, whereas nondiagnostic use of perceptual cues would result inthe secret drink being grouped with Coke.

Five beverage triads similar to the one described above were con-structed for the study, one for a warm-up exercise and four for the actualexperiment. All five triads were designed so that package type was thediagnostic perceptual cue and package color or size was the nondiagnosticperceptual cue. Package types were selected as the diagnostic cues by virtueof the fact that they provide the strongest perceptual cues for distinguishingdifferent beverage categories (e.g., 12-ounce cans for soft drinks, smallcartons for milk). Multiple triads, with different package types as the di-agnostic perceptual cue, were developed to avoid the problem that, forany one triad, package type might be more or less perceptually salient thanthe nondiagnostic cue of package color or size.

The second set of stimuli was designed to directly assess subjects'knowledge of the correlation between diagnostic perceptual attributes andcategory membership. To complement data obtained from the beveragetriads described above, the correlations tested were those between packagetypes and category membership (milk, juice, soft drinks). A series of sevencards was developed, with each card containing two black and white linedrawings of various types of beverage packages. The illustrations weredrawn in such a way that package type was the only available perceptualcue, leaving out other cues such as colors or symbols. For example, onecard contained a picture of a small five-sided carton and a picture of asmall can. Illustrations were placed on the cards so that one picture (e.g.,small carton) would depict a suitable package type for one of the beveragecategories (e.g., milk), whereas the other picture (e.g., small can) wouldnot depict a suitable package type. In this way, subjects' responses to thefocal question ("Which of these does milk come in?") allowed us to assessthe extent to which package types were correctly correlated with beveragecategory membership.

Procedure

Children and adults were taken through the same basic procedure,with only minor modifications. Children were interviewed individually dueto the necessity of having an interviewer record children's responses forthem on the questionnaire form. Adults completed the procedure in aclassroom setting in two large groups with a self-administered questionnaire

PSYCHOLOGY & MARKETING/VOL. 7, NO. 4 285~

CHILDREN'S USE OF PERCEPTUAL CUES

form. Otherwise, the procedure for children described below was alsofollowed for the adult sample.

Upon arriving at the research site, children and their parents weregreeted by a field supervisor and introduced to one of several interviewers.While the interviewer engaged the child in a brief conversation, parentswere given a questionnaire to fill out, assessing how frequently their childconsumed each of the beverages used in the study as well as how frequentlydifferent beverage categories (milk, juice, soft drinks) were purchased indifferent types of packages (cans, boxes, 2-litre bottles). As parents startedto fill out the questionnaire, the child was led to one of four interviewingrooms. Each room was set up with six stations partitioned off from oneanother so that the child's attention could be directed towards one stationat a time. The first five stations contained the beverage triads, with thelast station containing the booklet of various line drawings of beverageitems.

After entering the interviewing room, interviewers told children thatthey were going to play some games with different kinds of drinks. Childrenwere led to the first station, which served as a warm-up task for the beveragetriad task, and were given specific instructions for the game. Interviewersfirst identified each of the three beverage items by name. Two of the itemswere real products that were highly famiUar to children (e.g., apple juice,coke, chocolate milk). Interviewers referred to these items by their genericnames (e.g., "orange juice") rather than their specific brand names (e.g.,"Minute Maid orange juice") to facilitate children's identification of andfamiliarity with the items. However, Coke and Pepsi were referred to bytheir brand names due to the fact that children were more familiar withthese brand names than the generic "cola" label. The third item in thebeverage triad, a hypothetical product designed especially for the study,was identified by interviewers as the "secret" drink. Children were askedto select which of the "real" drinks was most like the "secret" drink andwere asked to indicate their reason for its selection. Following the warm-up task, children proceeded to play four more games just like this one withdifferent displays of beverage items.

Following this, children performed a second task to provide furtherevidence of their use of perceptual cues. The purpose of this game was todirectly assess children's knowledge of the correlation between diagnosticperceptual attributes and category membership. Children were shown aseries of seven cards, with two line drawings of different types of beveragepackages per card (e.g., a small milk carton and a soda can). Subjects wereasked to judge which of the two items depicted in the picture belonged toa specified category (e.g., "Which of these does milk come in?"). Subjects'selections for each card were recorded.

After completing the last task, children were returned to their par-

~286 WINTER 1990

CHILDREN'S USE OF PERCEPTUAL CUES

ents. Parents and children were debriefed and paid. On average, the pro-cedure took 5 to 10 minutes to complete.

Independent Variable

Three age groups, four- and five-year-olds, six- and seven-year-olds,and college-aged adults, served as the levels for the independent variableof interest, age. Children in the age range from 4 to 7 years were selectedby virtue of the fact that they often rely upon perceptual features to ca-tegorize products (John & Sujan, 1990). Age differences within this rangewere examined by comparing four- and five-year-olds with six- and seven-year-olds, as these groups differ with respect to their reliance on perceptualcues (John & Sujan, 1990) and their basic processing skills (Roedder, 1981).College students were selected as the baseline group based on the beliefthat they would exhibit categorization skills representative of the adultpopulation.

Dependent Variables

Predictions regarding the use of perceptual cues, that even youngchildren would use them diagnostically and that the diagnostic use of per-ceptual cues would increase with age, were tested using multiple dependentmeasures. First, subjects' responses to the beverage triad task, in whichthey had to group one "real" product with a "secret" drink, were examinedacross all four beverage triads. The number of correct selections, or thoseselections presumably based on the diagnostic perceptual cue, constitutedthe dependent measure of interest here. Thus, subjects' responses on thismeasure ranged from 0 to 4.

Second, subjects' stated reasons for making their selections in thebeverage triad tasks were examined. Subjects' reasons were coded intofive main perceptual cue categories: package (e.g., "both cartons," "bothplastic bottles"), size (e.g., "both large"), color (e.g., "both red"), symbol(e.g., "both have fruit pictures on them"), and weight (e.g., "both lookheavy"). In addition, a category for underlying cues (e.g., "both good foryou," "both taste sweet") and a category for "other" reasons were in-cluded. Using a coding scheme developed along these lines (see Appendix),two independent judges who were blind to the hypotheses coded subjects'responses (inter-rater reliability was 87%). Disagreements were resolvedby discussion and checked by a third judge.

Using this scheme, at least one reason could be coded as a perceptualfeature for 98% of the classifications (more than one reason could be givenper classification). The three largest categories of perceptual features men-tioned by subjects, namely package (62%), color (23%), and size (13%),

PSYCHOLOGY & MARKETING/VOL 7, NO. 4 287~

CHILDREN'S USE OF PERCEPTUAL CUES

served as the dependent measures for subjects' responses. Mentions ofpackage style were considered as evidence of using diagnostic perceptualcues, whereas mentions in the other two categories were not. In additionto perceptual cues of this type, underlying cues were mentioned in 19%of the classifications. While not of direct importance to this study, it isinteresting to note that this proportion is quite low for older children andadults when compared to previous studies, quite likely because the targetdrink was hypothetical and underlying attributes for the product could notbe known, but only inferred.

The last dependent measure was constructed from subjects' responsesto the correlation task. The number of correct responses to the sevencorrelation tasks for each subject, ranging from 0 to 7, was noted as themeasure of interest.

RESULTS

Hypothesis 1

Our first hypothesis stated that even young children could use per-ceptual cues in a diagnostic way to categorize products. Evidence to thiseffect was obtained by assessing how well young children performed ontwo tasks, one being the beverage triad task in which they were asked toclassify the "secret" drink with one of two real products and the secondbeing the correlation task in which they were asked to judge which of twoperceptual cues (package types) was indicative of a specific beverage cat-egory. Means and standard deviations for these measures, grouped by age,are presented in Table 1.

The data indicate that young children in both age groups tested hereperformed quite well, in that their use of diagnostic cues was significantlyabove chance levels. Examining the evidence for four- and five-year-olds,one finds that they correctly classified on the basis of diagnostic perceptualcues about 70% of the time in the beverage triad task (an average of 2.8correct classifications out of 4 triads), a performance level significantlyabove that which could be expected by chance (x(l) = 8.7, p < 0.01).Additional evidence regarding children's use of perceptual cues in this agegroup, obtained by looking at their verbalized reasons for classifications,indicates that diagnostic package cues were explicitly mentioned for 32%of the classifications. Color cues received about the same number of men-tions (39%), whereas size cues received relatively few mentions (14%).Although these data point to the use of diagnostic cues among youngerchildren, the evidence is clearly weaker than that provided by the classi-fication task. The most likely explanation for this difference in findings ischildren's difficulty in verbalizing the rationale for their classifications, with

288 WINTER 1990

CHILDREN'S USE OF PERCEPTUAL CUES

TABLE 1Means and Standard Deviations for Dependent Measures

Measure

Number of correct classifications(maximum = 4)

Number of correct correlation judgments(maximum = 7)

Proportion of classifications for whichstated reason relates to package cues

Proportion of classifications for whichstates reason relates to color cues

Proportion of classifications for whichstates reason relates to size cues'*

Age of Subject

Four- and Five-Year-Olds

2.8(1.0)"4.3

(2.8)32%(.35)39%(.33)14%(.21)

Six- and Seven-Year-Olds

3.4(0.8)5.2

(2.7)62%(.39)20%(.25)16%(.22)

Adults

3.9(0.3)6.9

(0.4)89%(.14)11%(.16)10%(.20)

" Numbers in parentheses represent standard deviations.'' Proportions of package cues, color cues, and size cues total to more than 100% in some cases due tomultiple reasons given for a particular classification.

some children finding it difficult to describe package types as readily assimple color or size cues and some children (10%) finding it difficult todescribe anything beyond a general perceptual cue (e.g., "it looks thesame") or a nonresponse (e,g,, "don't know"), A performance level morein line with the classification task can be found in children's responses tothe correlation task, with children correctly determining correlations be-tween diagnostic perceptual attributes and category membership 62% ofthe time in the correlation task (an average of 4,3 correct out of 7), Thisperformance level is marginally above the performance level which mightbe expected purely by chance (x(l) = 2,8, p < 0,09),

Data for the six- and seven-year-old age group suggests even betterperformance, with an even greater use of diagnostic perceptual cues. Chil-dren in this age group correctly classified by diagnostic perceptual cues in85% of the cases, a performance level significantly above what one mightexpect purely by chance (x(l) = 25,60, p < 0,01), Direct evidence of theiruse of diagnostic perceptual cues, obtained by looking at children's ver-balized reasons for their classifications, confirms that diagnostic packagecues were used most frequently, mentioned about 62% of the time. Men-tions of this type were far more frequent than mentions of either colorcues (20% of the time) or size cues (16% of the time). Furthermore,children in this age group correctly determined correlations between per-ceptual cues and category membership about 74% of the time, significantlyabove what might be expected by chance (x(l) = 10,39, p < 0,01),

Thus, there is support for the first hypothesis in that children in bothage groups appear to be using perceptual cues in a diagnostic fashion in

PSYCHOLOGY & MARKETING/VOL 7, NO. 4 289~

CHILDREN'S USE OF PERCEPTUAL CUES

the beverage triad task and, to a somewhat lesser degree, in the correlationtask.

Hypothesis 2

The second hypothesis predicted that the diagnostic use of perceptualcues would increase with age. Support for this hypothesis was found byexamining data for both experimental tasks. Looking first at the beveragetriad task, there was a significant age trend in the number of correct clas-sifications. Six- and seven-year-olds classified more correctly than four-and five-year-olds (3,4 versus 2,8, (̂160) = 3,9, p < 0,01), and adultsperformed better than six- and seven-year-olds (3,9 versus 3,4, r(160) =3,2,p<0,01), Furthermore, the verbalization data provide direct evidenceof age trends in the use of diagnostic package cues. Adults mentionedpackage cues more frequently as a reason for their classifications thandid six- and seven-year-olds (89% versus 62% of the time, f(160) = 4,4,p < 0,01), and six- and seven-year-olds mentioned package cues morefrequently than four- and five-year-olds (62% versus 32% of the time/(160) = 4,9, p < 0,01),

Similar age differences were evident for the correlation tasks, Six-and seven-year-olds identified the correct perceptual cues more often thanfour- and five-year-olds (5,2 versus 4,3, r(160) = 2,0, p < 0,05), and adultsidentified the correct perceptual cue more often than the six- and seven-year-olds (6,9 versus 5,2, /(160) = 3,7, p < 0,01), Thus, there is substantialsupport for the second hypothesis in that age differences in the use ofdiagnostic perceptual cues were evident in both the beverage triad taskand the correlation task.

DISCUSSION

Our findings indicate that young children use perceptual cues in adiagnostic fashion to categorize products. Even children as young as fourand five years of age were able to classify products based on diagnosticperceptual cues, while ignoring nondiagnostic perceptual cues, at levelssubstantially above chance levels. These children were also able, to a some-what lesser degree, to discern the correlation between diagnostic perceptualcues and product category membership. Taken as a whole, these datasuggest that young children should be viewed as possessing some of thesame basic categorization skills as older children and adults, even thoughyoung children are less skilled in using perceptual attributes than theirolder counterparts.

Our findings also suggest a different explanation for how children'scategorization abilities develop than those offered to date. We argue that

290 ' WINTER 1990

CHILDREN'S USE OF PERCEPTUAL CUES

developmental trends in categorization, such as the shift from using non-diagnostic to diagnostic perceptual cues and the shift from using perceptualto underlying cues, can be traced to differences in the "theories" aboutcategories that children hold. Young children's theories, while relativelynaive, tie perceptual attributes to the meaning of concepts and categories(Murphy & Medin, 1985). Thus, perceptual attributes are not used in atotally random fashion by these children, but are used in a way that isdiagnostic of category membership according to their limited theories ofthe world. As children grow older, these theories become more complex,focusing only on those perceptual attributes that are truly diagnostic andadding new connections between underlying attributes and category mean-ing. Thus, rather than assuming a developmental shift from perceptuallybased concepts to more underlying or theoretically based concepts, perhapsthe more accurate conclusion is that all concepts are integrated with the-ories and it is simply a question of degree in how sophisticated thesetheories are.

The data presented here are consistent with this developmental se-quence, in that young children used diagnostic perceptual cues at a levelsubstantially above what might be expected by chance, suggesting the ex-istence of a theory or approach quite different from simple perceptualsalience. Furthermore, the fact that the use of diagnostic cues increasedwith age is consistent with the idea that the underlying approach or theoryused to categorize objects develops with age. However, more direct evi-dence of such theories remains a topic for further research. Several meth-ods, perhaps direct verbalizations of theories or more extensive tests forknowledge of what types of attributes covary with category membership,may be required to actually document the existence and form of children'stheories in this area. Evidence to this effect should be pursued for a varietyof perceptual cues, such as color and size, to generalize the findings beyondthe type of packaging cues examined here.

In addition to this focus, future research might also examine whyyoung children's theories are initially naive and why they become moresophisticated with age. One possibility is that young children overestimatethe degree of covariation between perceptual attributes and category mem-bership, thereby producing inexact theories about categories. Young chil-dren may overgeneralize from a few instances and hold onto those gen-eralizations by failing to update their theories with newly acquired evidenceto the contrary. Viewed from this perspective, the errors we see in youngchildren's theories can be linked to deficits in basic reasoning skills aboutdetecting and explaining covariation (Kuhn, Amsel, & O'Loughlin, 1988,for related arguments).

A second possibility for the existence of naive theories in youngchildren is that some perceptual dimensions, other than those most diag-nostic of category membership, dominate processing due to their inherent

PSYCHOLOGY & MARKETING/VOL. 7, NO 4 291

CHILDREN'S USE OF PERCEPTUAL CUES

properties. There is some evidence, for example, that nominal dimensions,containing values that cannot be logically ordered (e.g., shape and color),dominate continuous dimensions, containing ordered values such as size,in making categorization decisions (e.g.. Garner, 1974), possibly becauseof their perceptual salience. This explanation is consistent with our datathat indicates young children's use of both color and package shape inmaking categorization decisions. It may be that the bias toward usingnominal dimensions in categorization is particularly strong in younger chil-dren, leading to a reliance on such perceptual dimensions in children'stheories about concepts, regardless of their diagnostic ability.

A final explanation for why young children's theories are relativelynaive is that young children lack certain processing skills or strategiesneeded for encoding, organizing, and retrieving information. Thus, eventhough information needed to build more sophisticated theories is availableto young children, it is not in an easily accessible form or easily activatedwhen processing occurs (Bjorklund & Jacobs, 1985). Viewed in this way,young children's less developed theories are a consequence of deficits inmeta-cognitive skills or processing strategies needed to interpret and usedomain-specific knowledge most effectively. Indirect support for this ex-planation exists in that there is a strong correlation between the ages atwhich children shift to using more diagnostic cues to categorize objects(e.g., underlying cues) and the ages at which a number of informationprocessing strategies for encoding and retrieving information emerge.

Clearly, some of these possibilities are closely related. For example,deficiencies in processing strategies may limit access to certain types ofinformation, leading to the overutilization of less diagnostic information.Problems experienced by young children in estimating covariation betweenattributes and category membership may be largely governed by an inabilityto process complex stimulus configurations. Understanding the possibleconnections between these developmental phenomena awaits further evi-dence about the general development of children's theories of conceptsand categories.

APPENDIX

Coding Scheme

Perceptual cues (98% of classifications). These are concrete attributesand the presence (or absence) of perceptual cues can be assessed basedon direct, visual observation. Perceptual cues include: (1) package cuesthat relate to package type ("both cartons"), package shape ("bothsquare"), package material ("both plastic"), and package details or

"292 ' WINTER 1990

CHILDREN'S USE OF PERCEPTUAL CUES

components ("same type of cap"); (2) color ("both black"); (3) size("both large"); (4) symbol ("both have fruit pictures on them"); and(5) weight ("both heavy").

Underlying cues (19% of classifications). These are more abstract at-tributes and convey a deeper understanding of the category or product.These attributes generally cannot be assessed by visual inspection andneed to be learned either through direct experience or socialization.Examples of such cues include product taste ^'both sweet") and prod-uct benefits ("both nutritious").

Other cues (4% of classifications). These are attributes that cannot beeasily classified under the other two categories and include mostlymiscellaneous responses ("I don't know").

REFEREIVCES

Alba. J. W., & Chattopadhyay, A. (1985). The effects of context and part-category cues onthe recall of competing brands. Journal of Marketing Research, 22, 340-349.

Bjorklund, D. R., & deMarchena, M. R. (1984). Developmental shifts in the basis of or-ganization in memory: The role of associative versus categorical relatedness in chil-dren's free recall. Child Development, 55, 952-962.

Bjorklund, D. R, & Jacobs, J. W., Ill (1985). Associative and categorical processes inchildren's memory: The role of automaticity in the development of organization infree recall. Journal of Experimental Child Psychology, 39, 599-617.

Bjorklund, D. F., & Thompson, B. E. (1983). Category typicality effects in children's memoryperformance: Qualitative and quantitative differences in the processing of categoryinformation. Journal of Experimental Child Psychology, 35, 329-344.

Bruner, J. S., Olver, R. R., & Greenfield, P. M. (1966). Studies in cognitive growth. NewYork: Wiley.

Carey, S. (1982). Semantic development: The state of the art. In E. Wanner & L. R. Gleitman(Eds.), Language acquisition: The state of the art (pp. 347-389). Cambridge, England:Cambridge University Press.

Cohen, J. B., & Basu, K. (1987). Alternative models of categorization: Toward a contingentprocessing framework. Journal of Consumer Research, 13, 455-472.

Cohen, L. B., & Younger, B. A. (1983). Perceptual categorization in the infant. In E.Scholnick (Ed.), New trends in conceptual representation: Challenge to Piaget's theory(pp. 197-220). Hillsdale, NJ: Eribaum.

Denney, N. W. (1974). Evidence for developmental changes in categorization criteria forchildren and adults. Human Development, 17, 41-53.

Fenson, L., Vella, D., & Kennedy, M. (1989). Children's knowledge of thematic and tax-onomic relations at two years of age. Child Development, 60, 911-919.

Flavell, J. H. (1985). Cognitive Development. Englewood Cliffs, NJ: Prentice-Hall.Garner, W. R. (1974). The processing of information and structure. New York: Wiley.Gelman, R., & Baillargeon, R. (1983). A review of some piagetian concepts. In J. H. Flavell

& E. M. Markman (Eds.), Handbook of child psychology: Vol. 3. Cognitive Devel-opment (pp. 167-230). New York: Wiley.

Gelman, S. A. (1988). The development of induction within natural kind and artifact cate-gories. Cognitive Psychology, 20, 65-95.

Gelman, S. A., & Markman, E. M. (1986). Categories and induction in young children.Cognition, 23, 183-209.

John, D. R., & Sujan, M. (1990). Age differences in product categorization. Journal ofConsumer Research, 16, 452-460.

PSYCHOLOGY & MARKETING/VOL. 7, NO. 4 293

CHILDREN'S USE OF PERCEPTUAL CUES

Kuhn, D., Amsel, E., & O'Loughlin, M. (1988). The development of scientific thinking skills.San Diego, CA: Academic Press.

Markman, E. M. (1980). The acquisition and hierarchical organization in categories by chil-dren. In R. Spiro et al. (Eds.), Theoretical issues in reading comprehension, (pp. 371-406). Hillsdale, NJ: Erlbaum,

Markman, E. M., & Callahan, M. A. (1983). An analysis of hierarchical classification. InR. Sternberg (Ed.), Advances in the psychology of human intelligence (Vol. 2, pp.325-365). Hillsdale, NJ: Erlbaum.

Medin, D. L., & Smith, E. E. (1984). Concepts and concept formation. Annual Review ofPsychology, 35, 113-138.

Melkman, R., & Deutsch, C. (1977). Memory functioning as related to developmental changesin bases of organization. Journal of Experimental Child Psychology, 23, 84-97.

Melkman, R., Tversky, B,, & Baratz, D. (1981). Developmental trends in the use of per-ceptual and conceptual attributes in grouping, clustering, and retrieval. Journal ofExperimental Child Psychology, 31, 470-486.

Moely, B. E. (1977). Organizational factors in the development of memory. In R. V. Kail,Jr. and J. W. Hagen (Eds.), Perspectives on the development of memory and cognition.Hillsdale, NJ: Erlbaum.

Murphy, G. L., & Medin, D. L. (1985). The role of theories in conceptual coherence.Psychological Review, 92, 289-316.

Olver, R. R., & Hornsby, J. R. (1966). On equivalence. In J. S. Bruner, R. R. Olver, & P.M. Greenfield (Eds.), Studies in cognitive growth. New York: Wiley.

Piaget, J. (1970). Piaget's theory. In P. H. Mussen (Ed.), Carmichael's manual of childpsychology (Vol. 1). New York: Wiley.

Roedder, D. L. (1981). Age differences in children's responses to television advertising: Aninformation processing approach. Journal of Consumer Research, 8, 144-152.

Rosch, E., Mervis, C. B., Gray, W. D., Johnson, D. M., & Boyes-Braem, K. (1976). Basicobjects in natural categories. Cognitive Psychology, 8, 382-439.

Saltz, E., Soller, E., & Sigel, I. E. (1972). The development of natural language concepts.Child Development, 43, 1191-1202.

Smiley, S. S., & Brown, A, L. (1979). Conceptual preference for thematic or taxonomicrelations: A nonmonotonic age trend from preschool to old age. Journal of Experi-mental Child Psychology, 28, 249-257.

Sugarman, S. (1983). Children's eariy thought: Developments in classification. New York:Cambridge University Press.

Sujan, M. (1985). Consumer knowledge: Effects on evaluation strategies mediating consumerjudgments. Journal of Consumer Research, 12, 31-46.

Sujan, M., & Bettman, J. R. (1989). The effects of brand positioning strategies on consumers'brand and category perceptions: Some insights from schema research. Jourrml ofMarketing Research, 26, 454-467.

Sujan, M., & Dekleva, C. (1987). Product categorization and inference making: Some im-plications for comparative advertising. Journal of Consumer Research, 14, 141-154.

Tversky, B. (1985). Development of taxonomic organization of named and pictured cate-gories. Developmental Psychology, 21, 1111-1119.

The authors wish to thank Yvonne Karsten for her assistance in stimuli developmentand data collection. This project was supported by the Graduate School and theCarlson School of Management at the University of Minnesota,

Requests for reprints should be addressed to Dehorah Roedder John, CarlsonSchool of Management, University of Minnesota, 271 19th Avenue South, Min-neapolis, MN 55455,

294 WINTER 1990