measuring the involvement construct

Measuring the Involvement ConstructJUDITH LYNNE ZAICHKOWSKY**A bipolar a<^ective scale, ttie Personal Involvement Inventory (PU). was developedto capture the concept of involvement for products. The scale successfully metstandards for internal relisUiility, reliability over time, cwitent validity, critwion-relatedvalidity, and construct validity. Tests of consfruct validjty demonstrated ttiat thescores were positively related to perceived differences among brands, brand pref-erences, interest in gafiiering infomiation about ite product category, and comparisonof product attributes among chers of cotisumer behavior have historicallydeveloped a number of complex theories in theattempt to explain and predict the behavior of the con-sumer (e.g., Bettman 1979; Enge!, KoUat, and Blackwell1978; Howard and Sheth 1969). These theories proposethat consumers actively search for and use informationto make informed choices. This implies that the con-sumer is an intelligent, rational, thinking, and problem-solving organism, who stores and evaluates sensory in-puts to make a reasoned decision (Markin and Naray-ana 1975).However, a great deal of consumer behavior does notinvolve extensive search for information or a compre-hensive evaluation of the choice alternatives, even forthe purchase of major items (Olshavsky and Granbois1979). The average consumer makes dozens of mun-dane decisions each day, few of which tnay be of im-portance. For such decisions, it may be inappropriateto assume an active information processor (Kassarjian1978, 1981). This idea has led theorists to view con-sumer behavior in terms of a two-fold dichotomy; lowinvolvement consumer behavior and high involvementconsumer behavior (Engei and Blackwell 1982).THE PROBLEMAlthough researchers agree that the study of low ver-sus high involvement states is interesting and important,there is currently little agreement about how to bestdefine, and hence measure, the construct of involve-ment (Cohen 1983). The reasons for the diverse defi-nitions and measures of involvement are perhaps due'This article was a finalist in the 1984 Robert Ferber Award forConsumer Research competition for the best interdisciphnary articlebased on a recent doctoral dissertation. The award is cosponsored bythe Association for Consumer Research and the Journal of ConsumerResearch.**Judilh Lynne Zaichkowsky is Assistant Professor of Marketing,Faculty of Business Administration, Simon Fraser University, Bur-naby, B.C., Canada V5A 1S6. The author wishes to thank Hal Kas-sarjian, who chaired the dissertation on which this article is based atthe University of California, Los Angeles, and the Canada Councilfor financial support. A special thanks goes to The American Uni-versity and Nanette Brown for help in word processing the manuscript.34 ito the different applications of the term "involvement."The literature suggests that a person can be involvedwith advertisements (Krugman 1962, 1965, 1967,1977), with products (Howard and Sheth 1969; Hupferand Gardner 1971), or with purchase decisions (Clarkeand Belk 1978). Involvement with these different objectsleads to different responses. For example, involvementwith ads leads one to give more counterarguments tothe ad (Wright 1974). Involvement with products hasbeen hypothesized to lead to greater perception of at-tribute differences, perception of greater product im-portance, and greater commitment to brand choice(Howard and Sheth 1969). Involvement with purchasesleads one to search for more information and spendmore time searching for the right selection (Clarke andBelk 1978). Therefore, each area might have its ownidiosyncratic result of the state of being involved withthe chers generally use the resulting behaviors asindicators of the level of involvement. Previous researchhas examined involvement with advertisements via afive-point scale that measures the degree of attentionto the ad (Wright 1973, 1974). Involvement with prod-ucts has been measured by several methods: rank-or-dering products (Sheth and Venkatesen 1968), rating aseries of products on an eight-point concentric scale asto their importance in the subject's life (Hupfer andGardner 1971), asking how important it is to get a par-ticular brand (Cohen and Goldberg 1970), or findingthe total times that subjects report "don't know" for aseries of brands (Ray 1973). On a broader level, in-volvement has been measured by administering Likertstatements that were thought to tap the underlying con-cept—e.g., the product means a lot to me, it matters tome, or the product is important to me (Lastovicka andGardner i978a; Traylor 1981).These diverse measures pose several problems for re-searchers. If conflicting results are obtained, we do notknow if the discrepancy is due to different measures orto different behaviors. Second, many scales are single-item measures and may not capture the total involve-ment concept. Finally, single-item measures have lowreliability, and current multiple-item measures have not© JOURNAL OF CONSUMER RESEARCH • Vol. 12 • December 1985

342been tested for internal reliability, stability, or a standardized, general, valid, and multiple-itemmeasure of involvement should be MEASURING INVOLVEMENTBACKGROUND AND CRITERIAhavior that results from involvement—would allow theA measure of involvement—independent of the be-researcher to use the same measure across various re-search studies. This measure should also be sensitive tothe proposed areas that affect a person's involvementlevel. These areas might be classified into three cate-gories (Bloch and Richins 1983; Houston and Roths-child 1978):1. Personal—inherent interests, values, or needs thatmotivate one toward the object2. Physical—characteristics of the object that cause dif-ferentiation and increase interest3. Situational—something that temporarily increasesrelevance or interest toward the objectIn Houston and Rothschild's (1978) framework, differ-ent situations and different people are two factors thatlead to various levels of involvement. Houston andRothschild integrate physical characteristics of theproduct as part of

Bloch and Richins (1983), the present article separatesthe situational factor. Coinciding withthe physical from the situational and allows the samephysical object to be subjected to different levels of in-volvement given different , or situational—that influence the consumer'sThe evidence for the three factors—physical, per-level of involvement or response to products, advertis-ing, and purchase decisions is found in the example, Wright (1974) found that variation in thetype of media—print versus audio—influenced the re-sponse given to the same message (physical). Lastovickaand Gardner (1978a) demonstrated that the same prod-uct has different involvement levels across people (per-sonal), and Clarke and Belk (1978) demonstrated thatdifferent purchase situations for the same productscause differences in search and evaluation or raise thelevel of involvement (situational). Based on this priorreasoning, a measure of involvement might be devel-jects, and that would pick up differences across people, ob-ing a measurement approach that seemed to be gener-Different types of scales were pretested before select-alizable across all product categories. First, a series ofvignettes was developed to represent involvement. Thevignettes were similar to scenarios found in Lastovickaand Gardner (1978b). Problems arose with developingenough generalizable scenarios for a reliable scale. Lik-ert scale items proposed a problem because items thatseemed to be appropriate for frequently purchasedTHE JOURNAL OF CONSUMER RESEARCHgoods did not seem to apply to durable goods and to be a semantic differential type (Osgood, Suci,The most effective and generalizable tyjre of scale ap-and Tannenbaum 1957). The Semantic Differentialconsists of a series of bipolar items, each measured ona seven-point rating scale. It is easy to administer andscore, takes only a few minutes to complete, and is ap-plicable to a wide array of objects. The descriptors orphrases easily relate across product categories and canbe appropriate to other domains, such as purchase de-cisions or adverti^ments. (However, the main focus ofthis article and scale development is involvement withproducts.) The steps taken to develop the measure were:1. Define the construct to be measured.2. Generate items that pertain to the construct.3. Judge the content validity of generated items (itemreduction).4. Determine the internal reliabiiity of items judged tohave content validity (item reduction).5. Determine the stability of internally reliable itemsover time (item reduction).6. Measure the content validity of the 20 selected itemsas a whole.7. Measure the criterion-related validity, which is theability of the scale to discriminate among differentproducts for the same people and different situationsfor the same product and same people.8. Test the construct validity or theoretical value of thescale by gathering data and testing whether the scalediscriminates on self-reported NG THE CONSTRUCTment that focuses on personal relevance (GreenwaldThis article will adopt the general view of involve-and Leavitt 1984; Krugman 1967; Mitchell 1979;Rothschild 1984). In the advertising domain, involve-ment is manipulated by making the ad "relevant:" thereceiver is pereonally affected, and hence motivated, torespond to the ad (e.g.. Petty and Cacciopo 1981). Inproduct class research, the concern is with the relevanceof

In purchase decision research, the concern is that thethe product to the needs and values of the on is relevant, and hence that the consumer willbe motivated to make a careful purchase decision (e.g.,Clarke and Belk 1978). Although each is a different do-main of research, in general, high involvement meanspersonal relevance (Greenwald and Leavitt 1984).the purposes of scale development was:In this study, the definition of involvement used forA person's perceived relevance of the object based oninherent needs, values, and definition recognized past definitions of involve-ment (e.g., Engel and Blackwell 1982; Krugman 1967;

MEASURING INVOLVEMENTMitchell 1979). Judging from previous writings, thisdefinition may be applied to advertisements, products,or purchase decisions. Early work by Krugman (1962,Wright (1974) defined involvement with advertising as1967) in advertising focused on persona! receiver's perception of the relevancy of the ad con-tent to some pending problem. In the area of productclass involvement, Howard and Sheth (1969) used theterms importance of purchase and involvement inter-changeably; they defined involvement in terms of aperson's needs or values. Hupfer and Gardner (1971)defined involvement as a general level of interest in orconcern about an issue without reference to a specificposition. Finally, Houston and Rothschild (1978) re-ferred to response involvement and defined it as a func-tion of enduring involvement or a need derived from avalue in the individual's hierarchy of GENERATION ANDCONTENT VALIDITYbased on the eaTlier definition of involvement. Thus, aA semantic differential scale was to be developedlist of 168 word pairs was generated to represent thisconcept of involvement. Examples of those pairs areimportant-unimportant, interested-uninterested, andexciting-unexciting. The first step was to judge the pro-posed items for content validity—how well the chosenitems represent the defined concept. Content validityof

deletion of poor word pairs, and (2)

the 168 word pairs was tested in two phases: (1) initialmore appropriate word judging of thesumer behavior) were given this study's definition ofThree expert judges (senior Ph.D. candidates in con-involvement and instructed to rate the 168 word pairsthree times: first, replacing the word "object" in thedefinition with "product;" second, replacing the word"object" with "advertisement;" and third, replacing theword "object" with "purchase decision." Each wordpair was rated on the following scale: (1) clearly rep-resentative of

of involvement, and (3) not representative of involve-involvement, (2) somewhat representativement. Word pairs that were not rated as representativeof involvement for any advertisement, purchase deci-sion, or product were dropped. Examples of deletedword pairs are adequate-inadequate, controversial-noncontroversial, and es of attitudes used in the psychology and mar-Word pairs that were dropped included traditionalketing literature. Word pairs such as good-bad, pleas-ant-unpleasant, nice-awful,

Loken 1984; Mitchell and Olson 1981) were judged toand like-dislike (e.g.,be unrepresentative of involvement. The judges decidedthat other word pairs, such as valuable-worthless andappealing-unappealing would remain, as they seemedto measure involvement. Items at the low end of thebipolar scale that represent the low end of involvement343were generally not negative—as they would be if mea-suring attitudes—but rather were "who cares" descrip-tors, e.g., unimportant, unexciting, doesn't matter, orof no using the same procedure. Only 23 items wereFive new judges then rated the remaining 43 wordconsistently rated as representing the involvement con-struct (80 percent agreement over products, purchasedecisions, and advertisements for each word pair). Thismeant that at least 12 of the possible 15 judgments foreach word pair (five judges over three objects) had tobe rated as representative of the involvement ent across judges and within each area for the23 word pairs was as follows: advertisements, 84 per-cent; products, 87 percent; and purchase decisions, items with which to start data collection (French andTwenty-three was assumed to be too low a numberMichael 1966; Nunnally 1978). Thus, seven additionalitems were added to the item pool to raise the initialnumber to 30 (five of these seven were eventuallydropped). For example, trivial-grand (45 percentagreement) was changed to trivial-fundamental, andinspiring-discouraging (55 percent agreement) waschanged to inspiring-uninspiring and returned to thelist. Therefore, a thirty-item scale emerged from thecontent validity phase that trained and knowledgeablejudges agreed measured involvement over three do-mains: products, advertisements, and purchase deci-sions. However, this study focused on, and further val-idation procedures were carried out on, involvementwith AL SCALE RELIABILITYscale over different product categories to measure theThe next task was to administer the 30 items as ainternal consistency or inter-item correlation. Twoproduct classes—watches and athletic shoes—were se-jects. One hundred and fifty-two undergraduate psy-lected because they were thought to be used by the sub-chology students completed the scale during class imately half of the subjects filled out the scalepertaining to athletic shoes and the other half filled outthe scale pertaining to watches. The results show thatfor both product categories, 26 bipolar items had anitem-to-total score correlation of 0.50 or more, and aCronbach alpha level of ations were dropped; interestingly, most of theseSix adjective pairs with relatively low item-to-totaladjective pairs had been returned earlier to the itempool. Factor analyses, using varimax rotation withsquared multiple correlations in the diagnonals for fac-tor extraction, were carried out over both products tocheck if the items selected for deletion loaded onto oneparticular dimension or were amorphous across both watches and athletic shoes, one factor ex-

344plained the major variation in the data, accounting for70.3 percent and 69.3 percent of

respectively (eigenvalues 13.3 and 13.2). Watches hadthe common variance,two more factors, accounting for 11.6 percent and 5.6percent of the common variance (eigenvalues 2.2 andcounting for 11.7 percent, 5.9 percent, and 5.7 percent1.1), and athletic shoes had three more factors, ac-ofthe common variance (eigenvalues 2.2, 1.2, and l.l).items selected for deletion did not load together on anyThe results of the factor analyses showed that theunique factor across either product category. Since thefirst factor accounts for approximately 70 percent ofthe variance, and none of the remaining items had aloading of zero or less on that first dimension, the scaledevelopment continued on the assumption of a simplelinear combination of the individual items (Comreyficient, and that it is the scale taken as a whole that1973). The assumption is that no individual item is suf-tends to measure the involvement construct (Nunnally1978).TEST-RETEST RELIABILITYexamined over two new subject samples and four newTest-retest reliability of the remaining 24 items wasproduct categories. Sixty-eight psychology students ini-tially rated calculators and mouthwash. Forty-five MBAstudents rated breakfast cereals and red wine. The orderjects in each group rated one product category first, andof the products was counterbalanced—half of the sub-the other half rated the other product category scales were administered during class time and tookabout five minutes to same product categories to the same subjects. Thir-Three weeks later the scales were administered overteen psychology students and 19 MBA subjects werelost to attrition; thus, 55 psychology students and 26MBA students were used to measure test-retest reli-ability. The average Pearson correlation between Timeto-item correlations ranged from 0.31 to 0.93. Four ad-I and Time 2 on the 24 items was 0.90. Individual item-ditional items with average test-retest correlations below0.60 were deleted. The resulting twenty-item involve-ment score test-retest correlations for each product wereas follows: calculators,breakfast cereals, r = 0.88; and red wine, r = 0.88; mouthwash, r = 0.89;product categories were also tested for internal scale r = 0.93. Thesereliability. The Cronbach alpha ranged from 0.95 to0.97 over the four reliability and stability phases of scale develop-Therefore, a twenty-item scale emerged from the in-ment for products. Twenty items allowed an adequatesampling of the possible items that represent involve-ment with products and yet was long enough to ensurea high level of reliability.' On a practical level, the scaleis for subsets ofthe scale items, the case may be that a smaller number'Although the current analyses do not suggest what the reliabilityTHE JOURNAL OF CONSUMER RESEARCHfits neatly on one page and only takes a few momentsto complete. The scale was then counterbalanced sothat ten random items were reverse scored. Since eachbipolar item was rated on a seven-point scale, the totalpossible score ranged from a low of 20 to a high of scale was named the Personal Involvement Inven-tory (PII) and is listed in Appendix CONTENT VALIDITYfrom the open-ended responds of 45 MBA studentsA second measure of content validity was obtainedover three product categories: 35mm cameras, red wine,and breakfast cereals. After completing the scales foreach product, subjects answered the following open-ended question:Now we would like you to state, in your own words,you rated each product category as you did. whySubjects were then divided into three groups—high,medium, or low—for each product class according totheir scale scores.^ Examples ofthe open-ended respon-ses appear in the behavior) blind to the scale scores evaluated theTwo expert judges (senior Ph.D. candidates in con-total set of open-ended responses. For each productcategory, the judges sorted the comments into threegroups indicative of low involvement, medium in-volvement, and high involvement with the product cat-egory, based on how well the responses represented in-volvement, as defined s was 80 percent agreement for 35mm cameras,Interjudge reliability on the classification ofthe re-agreement for breakfast cereals. Classifications on which84 percent agreement for red wines, and 80 percentthe two expert judges did not agree were then given toof items would be almost as reliable as the 20 items. The problem ofreducing the scale to fewer items lies in deciding which items to selectas subsets, since individual items differed in their reliability acrossproduct categories. A subset of items that may approach the reliabilityofthe 20 items for one product may not approach the same reliabilityfor another product. This variation is evident in that the test-retesttotal score correlation ranged from 0.88 to 0.93 over products, andtest-retest for the 20 individual items ranged from 0.44 to 0.93 overvarious products. The twenty-item measure should outperform anysubset of the scale; besides, decreasing the number of items wouldnot really make the scale any easier to administer, but may serve todecrease the domain of items judged as being representative of in-volvement and also lower the reliability ofthe scale. Researchers whomay use this scale are warned not to haphazardly reduce Ihe numberof based on an overall distribution developed over !3 product cat-^The classification of subjects into low, medium, and high scoresegories (Table 3) and several hundred subjects. All scores were tab-ulated on the PII scale range presented in the Figure. Subjects whosePII scores fell into the bottom 25 percent of the overall distributionwere classified as having low involvement with the product. Subjectswhose PH scores fell into the middle 50 percent ofthe distributionwere classified as having medium involvement, and subjects whosePIl scores were in the top 25 percent ofthe distribution were classifiedas having high involvement with the product. For development ofthis classification scheme see Appendix B.

MEASURING INVOLVEMENTEXHIBITOPEN-ENDED RESPONSES ON CONTENT VALIDITY3Smm Cameras1. High involvement for cameras (scwe greater than 110).a. Subject 1. Cameras are important, but not essential. TTieyprovide s creative and historical outlet for me.b. Subject 12. Cimeras interest me and are ^i impcMiant bcrfibyto me.2. Uow involvement for cameras (scwes less than 70).a. Subject 17. Because I never use 35mm cameras and am notextremely interested in them.b. Subject 37. It's a nice product to have but not a high priority.I have several but as I recall, none of the purchases was an"involved" Wine1. High involvemait for red wine (score greater than 110).a. Subject 22. Red wine adds a tot to the approfsiate meals.b. Subject 6-1 have always wanted to know more about wmesand fflijoy it when people I know teach me about them.2. Low involvement for red wine (score less than 70).a. Subject 20.1 m not interested in vt/ines nor do I particularlyappreciate the mystique that surrounds wines, in general.b. Subject 36. OK for socials and getting ast cereals1. High involvement for breakfast cereals (score greater than 110).a. Subject 27.1 eat cereal, healthy efficient 'wake up America."Cereal is good for you.b. Subject 8. Because they are diet foods.2. Low involvement for breakfast cereals (score less than 70).a. Subject 3. ! think breakfast cereals are a sham. I only eatgr^ienuts. it infuriates me to see breakfast cerealsadvertised to be eaten with toast, juice, etc. What's the use,aw exercise? I refuse to buy cereal for my child.b. Subject 31.1 eat cereal for convenience; it is easy and fast. Ihave no interest in them nor am I fascinated with them.a third judge to classify. The categories of responses, asgrouped by the scale scores, were compared to the cat-egories of responses as grouped by the expert results are presented in Table 1. These data indicatea significant relationship between the scale scores andthe open-ended responses from the subjects, thus addingan additional modicum of support to the validity of ION-RELATED VALIDITYparing the scores from the developed instrument withCriterion-related validity is demonstrated by com-one or more external variables that provide a directmeasure of the characteristic in question (French andMichael 1966). The external variable selected as a cri-terion was the simple ordering or classification of prod-ucts into low or high involvement -one products classified in other studies as345TABLE 1RELATIONSHIP BETWEEN THE SCALE SCORES ANDTHE OPEN-ENDED RESPONSESCollapsed forJudges' ratingsChi-squareScalesccwesLowMediumHigh(Total)LowMediumHigh35 mm Cameras'Low710(8)11137Medium4127(23)0410High0410(14)(Total)(11)(17)(17)(45)Red wine"Low1210(13)1210Medium898(25)81014High016(7)(Total) (20) (11) (14) (45)Breakfast cereals^Low1930(22)19Medium99109)911303High022(4)(Total)(28)(14)(3)(45)• x:' - 10,4, Of = 2. p < 0.01."'x'= 17.0,c//-2,p<0.001,'x^= = 2,p<0.01NOTE: As more tban 20 percent ot the exp«:tecl ceS frequencies dropped below 5, eitherthe low or rwgh row was coKapsed mio the medium row to compute the either high or low involvement (Bowen andChaffee 1974; Hupfer and Gardner 1971; Lastovickaand Gardner 1978a; Traylor 1981) were presented to agroup of 68 undergraduate psychology students. As inHupfer and Gardner (197!), subjects rated each producton an eight-point scale: extremely unimportant in mylife (1) to extremely important in my life (8).surement: bubble bath (meanFrom these 21 products, four were selected for mea-(X =Bubble bath was previously selected as a low involve- 5.25); jeans (X = 6.6); and automobiles (X) = 2.35); facial tissue (X = 7.9).ment product, and jeans a high involvement productby Clarke and Belk (1978). Facial tissue was previouslyidentified as a low involvement product, and automo-biles as a high involvement product by Lastovicka andGardner (1978a).istered for each of the four product categories to a freshThe twenty-item involvement scale (PII) was admin-sample of 47 undergraduate psychology students duringclass time. The PII mean scores and standard deviationsfor each product were as follows^ bubble bath5-38 (Males= Sl,s = X = 55, Females X = X = 69^= 122, s= 26; jeans X = 99, 19. s = 21; and automobiles X 74); facial tissue Xoverall significant difference among the product meansA repeated measures analyses of variance showed an

346, 138) = 39.9,/J<0.001). Furthermore, each meanwas found to be significantly different from each of theothers {p < 0.01). These results are in agreement withprevious studies that have stated that facial tissue andbubble bath have lower involvement levels than jeansand UCT VALIDITYStudies of construct validity check the theory under-lying the test (French and Michael 1966). Three stepsare involved in construct validity. First, from the in-volvement literature, propositions are made about thebehavior of people with high and low scores. Second,data is gathered to test if the scale discriminates on be-havior, and third, an inference is made as to whetherthe theory is adequate to explain the data tical Propositions of InvolvementVarious propositions about differences in low andhigh involvement behavior were selected after reviewingtheoretical papers by several authors (e.g., Belk 1981;Bowen and Chaffee 1974; Lastovicka 1979; Lastovickaand Gardner 1978b; Mitchell 1979; Robertson f976;Tyebjee 1979). Generally, there seems to be someagreement on what constitutes the differences betweenhaving high or low involvement in a product class. Un-der the low involvement condition, researchers propose:1. A relative lack of active information seeking aboutbrands2. Little comparison among product attributes3. Perception of similarity among different brands4. No special preference for a particular brandBased on these various theoretical propositions, thefollowing specific statements were developed and thenadministered to subjects with the PII over variousproduct categories:1. I would be interested in reading information abouthow the product is made.2. I would be interested in reading thearticle about this product category. Consumer Reports3. I have compared product characteristics amongbrands of this product.4. I think there are a great deal of differences amongbrands of this product.5. 1 have a most preferred brand of this statements were rated on a seven-point scale fromstrongly disagree (I) to strongly agree (7).to involvement were administered over three productsMethod. The PII and the specific questions relatedto 28 clerical and 29 administrative staff members (7THE JOURNAL OF CONSUMER RESEARCHmales and 50 females) at a major university. Subjectswere personally contacted at work by the researcher andasked if they would participate. Those who agreed weregiven a questionnaire and asked to complete it at theiroffice desk. The researcher then returned in about anhour to collect the questionnaires from the additional subjects agreed to fill out the ques-tionnaire but never completed it. The median age rangeof the subjects was 35-44 years and the median edu-cation level was some college. The products selected forevaluation were instant coffee, laundry detergent, andcolor television.^ These were chosen to represent a rangeof products thought to be used by the ts were classified into three groups to comparethe responses between subjects who had low involve-ment with the product category and those who had highinvolvement with the product category. This classifi-cation scheme is the same as that found in the secondcontent validity section and further explained in Ap-pendix B. The particular question of interest was: "Didsubjects having low PII scores for the product categoryrespond to the five statements differently than did thosehaving high PII scores for the product category?"Planned comparisons, by simple t-tests, were carriedout between the low and high PII scores for each state-ment and product category. Before comparing the dif-ferences between the low and high group in their re-sponses to the statements within each product category,a one-way MANOV A was computed over the five state-ments for each product category to determine if theoverall pattern of responses across the five propositionswas addition, the Pearson correlations between thescale score (range = 120) and the responses to the state-ments (range = 7) were computed. The cell means, cor-relations, and results of t-tests between the low and highcells are presented and summarized in Table s. The Cronbach alpha, the mean, and thestandard deviation of the PII scores for the three productcategories were as follows:Cronbach alphaMean {s)Instant coffeeColor television.97Laundry detergent.9966(40).97103 (23)97 (30)These scores provide two unexpected results: first, a rel-atively high PII score for laundry detergent (103), andsecond, a relatively low PII score for color television(97). These results should be interpreted in the contextof the sample population used in this study: this rela-tively homogeneous group of middle-aged females mayhave viewed laundry detergent as more involving thansituations for red wine were also included for other purposes.'PII scales for the product category of red wine and two purchase

MEASURING INVOLVEMENTTABLE 2HELATIONSHIP BETWEEN CONSTRUCT VALIDITY STATEMENTS AND LOW. MEDIUM. OR HIGH PII SCORES:MEANS, STANDARD DEVIATIONS, AND CORRELATIONSInstant coffeeConstruct v^iditystatement'1. ! would be interested inreading mfc>rmationabout how the product ismade.2. 1 would be intwested inreading Itie ConsumerReports article about thisproduct.3. 1 have compared productcharacteristics amongt»'£inds.4. 1 think there are a greatdeal of differencesamong brands.5. I have a most-preferredbrand of this (32)3.28(2.0)Medium(12)4.42(2.3)High(12)4.25(2.3)r".30=Laundry detergentLow(4)1.25'(.5)Medium(28)4.04(1.7)High(25)4.48(2.4)Low(9)3.56(2.1)Color televisionMedium(26)4.00(1.9)High(12)4.23(2.1)347r.37=r.143.00'(1.8)4.75(2.3)4.92(2.3).47"2.75*(2.9)4.46(2.0)5.00(2.1).33=4.56(1.9)4.65(1.9)5.36(2.1).27=2.59*(1.8)3.94'3.42(2.1)4.67(1.1)4.83(1.8)5.25(2.0)6.33(-8)6.17(1.7).52=1.75(1.5)2,25"(1-0)2.50"(3.0)4.36(1.8)4.00(1.7)4.68(1.6)4.80(2.4)5.20(2.1)5.44(1.9).42=3.11'(1.9)4.11'(1.2)2.56'(1-4)3.85(1.9)4.85(1.5)4.77(17)4.59(2.3)5.73(1.7)5.55(1.9).23".63=.42=.33=0-6)2.88'(1.9).68=.42=.50=• TTie construct valkSty statements are measi^ed on a seven-point scale: (t) strongly disagree to (7) strongly agree' r - Pearson CoireJation betwem PI! score and response to constrvici vaBdity question.'p <0.01,"p <0.05• Low scores signiflcantty (Sflerent than htgh scores p < 0-01.' Low scores significantly cfiffereot than high scores p < ; Nwnbera in parentfwses in Tatjie heading ore numbers of subjects in each group. NiHT*ef^ in parentheses in T^jle body are stand^fl television because it may be their responsibilityto do the family laundry. If this is true, they wouldvalue the product's benefits and they would be iikelyto be interested in the quality of the product becausethey need the product to perform their household sions, however, may not fall under their respon-sibility for maintenance or interest them as much. Elec-tronics, solid state, or color tuning may not be relevantto them. And if a television does not affect them per-sonally, housewives might have relatively low involve-ment with this results of the MANOVA for all five statementswere significant for the three products (instant coffeeK10, 98) = 6.56, p < ; laundry detergent F( 10, 100)= 2.34, p < 0.05; and color television F(1O, 100) = 2.00,p < 0.05). This indicated that there were significant dif-ferences due to the PII scores on the responses to all fivebehavioral statements pertaining to involvement. Theseoverall significant results allow the interpretation of eachproposition for product information. High involvementconsumers should be more interested in acquiring in-formation about the product than low involvementconsumers. For example, Engel and Blackwell (1982)defined involvement as the activation of extendedproblem solving behavior, and Bettman (1979) citedlevel of involvement as a mediating variable in infor-mation search. Thus, high scale scorers should indicatemore interest in product information than low ts were given two statements pertaining to in-formation search over the three product categories. Thefirst statement was "I would be interested in readinginformation about how the product is made (instantcoffee and laundry detergent) or works (color televi-sion)." The second statement was "I would be interestedin reading the Consumer Reports article about ."The results of the analyses of the information searchquestions are generally in agreement with the theory ofinvolvement. High scorers tended to be more interestedin information pertaining to the product than low scor-ers. All correlation coefficients were significantly dif-ferent from zero, with the exception of the product cat-egory of televisions. Perhaps the change in the wordingof the question led to the weak results for that productcategory. Perhaps interest in how televisions work canindicate interest in technology or interest in quality ofperformance other than interest in the product per ative evaluation. One of the characteristics ofhigh involvement is the evaluation of competing alter-natives. Since the highly involved consumer searchesfor relevant information, the available alternatives arethought to be consciously compared before a selection

348is made. To tap this dimension, subjects were asked theextent to which they agreed with the statement "I havecompared product characteristics among brands oficantly greater agreement with the statement than low." For all products, the high scorers had tested was that high involvement scoTers wouldPerception of brand differences. The next proposi-perceive greater differences among brands in the prod-uct class than low involvement scorers. This propositionstems from writings of Robertson (1976), who suggeststhat high involvement implies that beliefs about productattributes are strongly held, whereas low involvementindividuals do not hold strong beliefs about productattributes. Thus, the strength ofthe belief system to theattributes emphasizes the perception of differencesamong brands on the attributes where beliefs arestrongly held. Subjects were asked to respond to thestatement "I think there are a great deal of differencesamong brands of

ceived greater differences." High scorers always per-low scorers in the product class. {p < 0.01) among brands thanproduct class were hypothesized to have a most pre-Brand preferences. People highly involved in aferred brand in the product category. The preferenceof a particular brand stems from the perception of dif-ferences among brands. Since high involvement impliesperceiving greater differences about product attributes,then the consumer should have a greater preferencebased on that product differentiation. Again, over allthree products, high scorers showed a significantly (/J< ) greater agreement with the statement "I havea most preferred brand of " than low used the correlation of two paper and pencil testsIn conclusion, the various measures of construct va-on the same subjects as evidence that the proposed scaledoes tap the construct of involvement, as applied toproduct categories. Although no one result is an excel-lent test of the scale, each finding adds to the weight ofevidence that the scale is an acceptable measure of in-volvement, as applied to product ANALYSES OF THE PIIitem scale was carried out for each product categoryAn investigation ofthe dimensionality ofthe twenty-used in the scale development. The items were factoranalyzed using varimax rotation with squared multiplecorrelations in the diagonal for factor extraction. Thegeneral pattern of results showed one main factor and(usually) one minor or residual factor for every productcategory. The major factor accounted for a range ofcommon variance from 65 percent for jeans to 100 per-cet for instant coffee. Over ail products, all items loadedpositively on the first factor, which indicates that theasumption of a simple linear combination ofthe scaleitems was not JOURNAL OF CONSUMER RESEARCHSENSITIVITY TO SITUATIONALDIFFERENCESand the construct validity sections have demonstratedThe second content validity, the criterion validity,that the level of involvement with product categoriesgory, there seems to be individuals who have low in-varies greatly over individuals. For any product cate-volvement with the product and individuals who havehigh involvement with the product. Additionally, theaverage level of involvement varies across the differentproducts. For example, students rated bubble bath 69on the PII and rated automobiles 122 on the PII. Thisdemonstrates that different products are perceived dif-ferently by the same people. The scale is also proposedto be sensitive to different situations, a third factor thatcauses involvement, given the same people and the sameproducts.(1981) demonstrated that some purchase situations canPrevious studies by Clarke and Belk (1978) and Belkbe more involving than others. They found that thepurchase of some previously uninvolving products forgifts can raise the level of involvement in the purchasedecision. To investigate the possibility of rating pur-chase situations on the scale, the PII was administeredover two purchase situations for wine to 41 membersof the clerical and administrative staff used in the pre-vious construct validity study."* Each subject rated twopurchase situations: (i) the purchase of a bottle of winefor everyday consumption, and (2) the purchase of abottle of wine for a special dinner party. The scale itemswere internally reliable for these purchase ch alphas were 0.98 and 0.97 respectively, andthe item-to-total correlations were generally above anced across subjects. The mean scale score forFor this data collection, these situations were coun-the everyday consumption was 78special dinner party was 106sures t-test was significant at ;(40) - 5.42, (s = 24). A related mea- (,s = 34), and for thetherefore, the two purchase situations were rated dif- p < 0.001;ferently on the PIL The analysis for differences betweenthe two purchase situations was also carried out as abetween-subjects design. Twenjy-two subjects first ratedthe everyday consumptionfirst rated the special dinner party {X = 76) and 19 subjectstween-subjects t-test for the first rating was also signif- (X = 98). The be-icant at /(39) = 2.34,the PII is sensitive to different situations, if people and p < 0.05. These results show thatproduct remain constant. The PII may hold promise asa measure of involvement with purchase decisions, per-haps even applicable as a manipulation check for ex-periments that deal with manipulation ofthe situationas manipulation of involvement level (e.g.. Park andCacioppo, and Goldman 1981).Young 1983; Petty and Cacioppo 1979, 1981; Petty,*The other 16 subjects did not receive these scales as part of theirquestionnaire.

MEASURING INVOLVEMENTSUMMARYmeasure the construct of involvement. Hence, a se-The purpose of the study was to develop a scale tomantic differential scale was developed to capture theconcept of involvement for products. This Personal In-volvement Inventory was developed over four data setsof 268 undergraduate psychology students; two data setswith 49 MBA students; and two data sets with 57 clericaland administrative staflf members. The scale was dem-onstrated to have content validity by expert judges attwo phases of the scale development: first, for the se-lection of items, and second, through classification ofopen-ended responses from subjects. The reliability orstability of the scale over time was checked over twosubject populations for an average test-retest correlationof 0.90. The criterion-related validity of the scale waschecked by demonstrating agreement with the order ofvarious products as found in previous studies. The con-struct validity—the test ofthe scale to theoretical prop-ositions—was then carried out. The scale was admin-istered to clerical and administrative staff and coveredthree different product categories and several statementsof behavior proposed to be representative of involve-ment. Over all three product categories there was a pos-jects' responses to the statements of theoretical prop-itive relationship between the scale scores and the sub-ositions pertaining to tionsvergent and discriminant validity. The tests of conver-Missing from this scale development are tests of con-gent validity with another measure of involvement werenot carried out because at the time of this scale devel-opment no other general involvement measure in theliterature had been tested for reliability and validity.^Tests of discriminant validity to the concept of expertiseor knowledge structure were carried out and are re-ported in Zaichkowsky (1985). PII scores were foundto be unrelated to expertise but related to product r tests of discriminant and convergent validityneed to be carried out with respect to other particular, the relationship of PII scores to attitudesshould be further examined, since several items on thescale appear to be similar to a measure of products, the initial aim was to select items so thatAlthough the usefulness ofthe PII was demonstratedthe same scale might also be applied to advertisementsor purchase decisions. Some data were collected overdifferent purchase decisions and showed that the PIIwas internally reliable for different purchase situationsfor the same product. Additionally, the purchase situ-ations differed in their involvement scores, as would beexpected; the special dinner party scored higher on thefor possible validation to an independently developed Likert scale.*There is currently some research translating this scale to French349PII than the everyday purchase situation. Further, somepreliminary research indicates that the PII is an inter-nally reliable measure when applied to r, more research needs to be carried out to verifythe stability and construct validity of the PII to adver-tisements and purchase er behavior. It offers the potential of a validThe PII should have several benefits to the study ofinstrument to replace the ad hoc and untested ap-proaches that have previously been used in the involvement is proposed to be a variable in thedecision process, the PII offers researchers a quicklyadministered tool, generalizable across product cate-gories, that can be used as a covariate to other researchquestions. The ultimate test of the scale is whether ornot the instrument can be used in empirical studies totest various aspects of involvement. I am conductingsuch research, and as others use the instrument andgeneralizable norms develop, its true validity will bedetermined,APPENDIX APersonal Involvement Inventoryto measure a person's involvement with products. To changeThe following Personal Involvement Inventory is designedthe instructions to measure involvement with advertisementsor purchase decisions, the words in the parentheses shouldbe changed accordingly. To measure involvement with ad-vertisements, the words "various products they regularly pur-chase or have purchased in the past" would he changed to"the advertisements you have just seen (read)." To measureinvolvement with purchase decisions, the words "variouspurchase decisions people make" would he top ofthe scale page. Examples for three applications ofThe name ofthe object to he judged should he inserted atthe different contexts of the ohject are: (1) if the product wasthe object, then "red wine" would be judged; (2) if an ad wasthe ohject, then "the ad for Gallo wine" would be judged;and (3) if the purchase decision was the object, then the "pur-judged. The reader is reminded that the construct validity ofchase ofa bottle of wine for a special dinner party" would bethe scale has only been supported for ction PageInstructionsment or interest in (various products they regularly purchaseThe purpose of this study is to measure a person's invoive-or have purchased in the past). To take this measure, we needyou to judge various (products) against a series of descriptivescales according to how YOU perceive the product you willhe shown. Here is how you are to use these scales:If you feel that the (product) that appears at the top of thepage isplace your check mark as follows: very closely related to one end ofthe scale, you should

350Unimportant jc_: : : : :

Unimportant : ; :

or: Important; : : _x_ ImportantIf you feel that the (product) isthe other end of the scale (but not extremely), you should quite closely related o one orplace your check mark as follows:Appealing : JL: : : '• • UnappealingAppealing : : :

or: : xj. UnappealingIf you feel that the (product) seemsnot really neutral) to one end of the scale, you should piace only slightly related (butyour check mark as follows:UninterestedInterestedUninterestedorInterestedImportant1. Be sure that you check every scale for every (product); do2. Never put more than one check mark on a single omit each item a separate and independent judgment. Workat fairly high speed through this questionnaire. Do not worryor puzzle over individual items. It is your first impressions,the immediate feelings about the items, that we want. On theother hand, please do not be careless, because we want yourtrue Questions?Scale Page(insert name of object to be judged)importantunimportant*of no concernof concern to meirrelevantrelevantmeans a lot to memeans nothing to me*uselessusefulvaluableworthless*trivialfundamentalbeneficialnot beneficial*matters to medoesn't matter*uninterestedinterestedsignificantinsignificant*vitalsuperfluous*boringinterestingunexcitingexcitingappealingunappealing"mundanefascinatingessentialnonessential*undesirabledesirablewantedunwanted*not neededneeded' Indicates item is reverse on the left are scored (1) low invoivemenl lo (7) high involvement on the ng the 20 items gives a score from a low of 20 to a high of JOURNAL OF OONSUMER RESEARCHFIGUREOVERALL DISTRIBUTION OF PII SCORES'FREQUENCYeo96139MedtanLow invotvementHigh innolvBment90 111 ~Sample meanfs=321 SCORES"O indicalos end-pofit users 26 at low ot 20; 12 at hi^ ot : Skevnwss = -IX BDistribution of Scores Over AU Productsvolvement were on the scale, an overall distributionTo decide where cut off points for low and high in-was tabulated and is presented in the Figure. This over-jects over 13 product categories shown in Table distribution is based on data collected from 751 sub-However, some ofthe subjects

than one product, and thus are counted more than out the PII for moreThe overall PII mean for these products is 89.55,whereas the true theoretical mean is 80. This deviationfrom the theoretical mean is most likely due to theproduct-dependent nature of the distribution. No at-tempt was made to consciously select products that werejects. It seems that the scale was developed, perhaps,thought a priori to be more or less involving to the sub-over products that were somehow more involving. Ad-dition of other products, such as nails or canned peas,that might not be involving to the subjects might pushthe mean toward the theoretical mean of points 20 and 140 (these values are computed intoDeleted from the pictured distribution are the twothe mean scores). Twenty-six points were deleted at thelow end of 20, and 12 points were deleted from the highpoint of 140. These scores indicate that the rater onlyused the endpoints of one and seven to rate the productin fy scorers into either low, medium, or high in-The distribution derived from the data was used tovolvement when comparison among groups of individ-uals was of interest. Low scorers were defined as thosescoring in the first quartile ofthe distribution; they hadscores ranging from 20 to 69. Medium scorers were de-

MEASURING INVOLVEMENTTABLE 3PRODUCTS USED FOR SCALE DISTRIBUTIONMean PIIProductscore(S)Sample (N)Instant coffee66(40)57"Bubble bath69(38)45"Breaktast cere^69(29)43"^Mouthwash74(28)68"Red wine82(31)45 =Red wine84(31)57"Facial tissues87(26)47'*Headactie remedy91(25)68"35mm camera96(26)45=Color TV97(30)57"Jeans99(21)47"Laundry detergent103(23)57"Calculator112(16)68"Automobile122(19)47"* Clencal and admnisb^^ve staff.^ UrxJeryaduate psychology stutlems.° MBA : Theoretical mean = 80. Aclual mean based on ^xwe products = 90 (s = 32).fined as those scoring in the middle 50 percent of thedistribution; they had scores ranging from 70 to scorers were defined as those scoring in the topquartile of

velopment of a two-group classification when neces-111 to 140. The distribution was also used for the de-the distribution; they had scores ranging fromsary—i.e., those relatively high scores and those rela-tively low scores. The scale mean (89.55) was used asthe break point for this grouping. By using this overalldistribution as a guide to classification, some compar-ison is possible between low and high scoring subjectson the PII.[Received March 1984. Revised June 1985.]REFERENCESBdk, Russell W. (1981), "Effects of Gift-Giving Involvementon Gift Selection Strategies," inResearch,Association for Consumer Research. 408-411. Vol. 9, ed. Andrew Mitchell, Ann Arbor, MI: Advances in ConsumerBettman, James R. (1979), An Information Processing TheoryBloch, Peter H. and Marsha L. Richins (1983), "A TheoreticalOf Consumer Choice. Reading, MA: for the Study of Product Importance Perceptions,"Bowen, Lawrence and Steven H. Chaffee (1974), "ProductJournal of Marketing. 47 (Summer), ement and Pertinent Advertising Appeals,"nalism Quarterly. 51 (Winter), 613-621. Jour-Clarke, Keith and Russell W. Belk (1978), "The Effects ofProduct Involvement and Task Definition On Antici-pated Consumer Effort," insearch, Advances In Consumer Re-sociation for Consumer Research, 313-318. Vol. 5, ed. H. Keith Hunt, Ann Arbor, MI: As-35!Cohen, Joel B. (1983), "Involvement and You: 1000 GreatIdeas," inRichard Bagozzi and Alice Tybout, Ann Arbor, MI: As- Advances in Consumer Research, Vol. 10, ion for Consumer Research, in Post-Decision Product Evaluation,"and Marvin E. Goldberg (1970), "The DissonanceComrey, Andrew L. (1973),of Marketing Research, 1 Journal A First Course in Factor Analyses, (August), , James E., David Kollat, and Roger D. Blackwell (1978),New York: Academic er Roger D. Blackwell (1982), New York: Dryden Press. Consumer , John W. and William B. Michael (1966),New York: Dryden Educational and Psychological Tests and Manuals, StandardsGreenwald, Anthony G. and Clark Leavitt (1984), "AudienceWashington, D.C.: American Psychological ement in Advertising: Four Levels,"Consumer Research, 11 (June), 581-592. Journal ofHouston, Michael J. and Michael L. Rothschild (1978),"Conceptual and Methodological Perspectives in In-volvement," inlogues and Directions, Research Frontiers in Marketing: Dia-Marketing Association, 184-187. ed. S. Jain, Chicago: AmericanHoward, John A. and Jagdish N. Sheth (1969),Hupfer, Nancy and David Gardner (1971), "Differential In-of Buyer Behavior, New York: John Wiley. The Theoryvolvement with Products and Issues: An ExploratoryStudy," insearch,sociation for Consumer Research, 262-269. ed. David M. Gardner, College Park, MD; As- Proceedings: Association for Consumer Re-Kassarjian, Harold H. (1978), "Presidential Address," invances in Consumer Research. Ad-Ann Arbor, MI: Association for Consumer Research, Vol. 5, ed. H. Keith Hunt,xiii-xiv.(1981), "Low Involvement: A Second Look," in Ad-roe, Ann Arbor, MI: Association for Consumer Research,vances In Consumer Research, Vol. 8, ed. Kent B. n, Herbert E. (1962), "An Application of LearningTheory to TV Copy Testing,"26 (Winter), 626-634. Public Opinion ng Without Involvement,"(1965), "The Impact of Television Advertising:terly. 29 (Fall), 349-356. Public Opinion Quar-ment,"(1967), "The Measurement of Advertising Involve-Perception,"(1977), "Memory Without Recall, Exposure Without Public Opinion Quarterly, 30 (Winter), ), 7-12. Journal of Advertising Research, 17 (Au-Lastovicka, John L. (1979), "Questioning the Concept of In-volvement Defined Product Classes," inConsumer , MI: A^ociation for Consumer Research, 174- Vol. 6, ed. William L. Wilkie, Ann Advances ent," inand David M. Gardner (1978a), "Components of In-eds. John C. Maloney and Bernard Silverman, Chicago: Attitude Research Plays For High Stakes,American Marketing Association, High Involvement Cognitive Structures,"and David M. Gardner (1978b), "Low Involvementvances in Consumer Research, Vol. 5, ed. H. Keith Hunt, in Ad-

352Ann Arbor, MI: Association For Consumer Research,, Barbara (1984), "Attitude Processing Strategies,"Journal of Experimental Social Psychology,Markin, Rom J. and Chem Narayana (1975), "Behavior272-296. 20 (July),Control: Are Consumers Beyond Freedom and Dignity?"in Advances in Consumer Research,B. Anderson, Ann Arbor, MI; Association for Consumer Vol. 3, ed. BeverleeResearch, ll, Andrew A. (1979), "Involvement: A PotentiallyImportant Mediator of Consumer Behavior," invances in Consumer Research, Ad-Wilkie, Ann Arbor, MI: Association for Consumer Re- Vol. 6, ed. William , s the Only Mediator of Advertising Effects on Brandand Jerry C. Olson (1981), "Are Product AttributeAttitude," Journal of Marketing Research, 18 (August),Nunnally, Jum C. (1978),318-332. Psychometric Theory, New York:Olshavsky, Richard W. and Donald H. Granbois (1979),McGraw-Hill."Consumer Decision Making—Fact or Fiction?"Osgood, Charles, George J. Suci, and Percy H. Tannenbaumof Consumer Research, 6 (September), 93-100. Journal(1957), The Measurement of Meaning,Park, C. Whan and S. Mark Young (1983), "Types and Levelsversity of Illinois Press. Urbana, IL: Uni-of Involvement and Brand Attitude Formation," invances in Consumer Research,gozzi and Alice Tybout, Ann Arbor, MI: Association for Vol. 10, eds. Richard Ba- Ad-Consumer Research, , Richard E. and John T. Cacioppo (1979), "Issue In-volvement Can Increase Or Decrease Message RelevantCognitive Responses,"Psychology. 37 (October), 1915-1926. Journal of Personality and SocialA Moderator of the Effects On Attitude of Advertisingand John T. Cacioppo (1981), "Issue Involvement AsTHE JOURNAL OF CONSUMER RESEARCHContent and Context," insearch, Advances in Consumer Re-Association for Consumer Research, 20-24. Vol. 8, ed. Kent B. Monroe, Ann Arbor, MI:"Personal Involvement as a Determinant of Argument-John T. Cacioppo, and Rachel Goldman (1981),Based Persuasion,"Psychology. 41 (November), 847-853. Journal of Personality and SocialRay, Michael L. (1973), "Marketing Communications andThe Hierarchy-of-Effects," inCommunication Research,CA: Sage Publications, 147-176. ed. P. Clarke, Beverly Hills, New Models for MassRobertson, Thomas S. (1976), "Low Commitment ConsumerBehavior,"19-24. Journal of Advertising Research, 16 (April),Rothschild, Michael L. (1984), "Perspectives in Involvement:Current Problems and Future Directions," inin Consumer Research, AdvancesArbor, MI: Association for Consumer Research, 216- Vol. 11, ed. Tom Kinnear, , Jagdish N. and M. Venkatesan (1968), "Risk ReductionProcess in Repetitive Consumer Behavior,"Traylor, Mark B. (1981), "Product Involvement and BrandMarketing Research. 5 (August), 307-310. Journal ofCommilment," Journal of e, Tyzoon T. (1979), "Response Time, Conflict, and51-56. 21 (December),Involvement in Brand Choice,"Wright, Peter (1973), "Cognitive Processes Mediating Accep-Research, 6 (December), 259-304. Journal of Consumertance of Advertising,"10 (February), 53-62. Journal of Marketing Research,Zaichkowsky, Judith Lynne (1985), "Familiarity: ProductPublic Opinion Quarterly,"Analyzing Media Effects on Advertising Response," 38 (Summer), 192™, Involvement or Expertise?" Insumer Research, Advances in Con-and Morris B. Holbrook, Provo, UT: Association for Vol. 12, eds. Elizabeth C. HirshmanConsumer Research, 296-299.