Brain Research Bulletin 67 (2005) 422–427

A behavioral window on the mind of the market: An applicationof the response time paradigm

Fred W. Masta,∗, Gerald Zaltmanb

a Department of Psychology, Cognitive Neuroscience, University of Zurich, Treichlerstrasse 10, CH-8032 Zurich, Switzerlandb Harvard Business School, Boston, USA

Available online 5 July 2005

Abstract

This article focuses on the role of implicit knowledge consumers have about particular brands, products or names. The major findings ofseveral studies, conducted at the Mind of the Market Laboratory at Harvard Business School, are presented with specific emphasis on studiesin which response time measurements were the core method. The results revealed that implicit measures provide a rich understanding of themeaning conveyed by a product or brand. Moreover, there is also considerable evidence that implicit measures may be better than traditionalexplicit measures as predictors of consumer behavior. We discuss the implications for the study of consumer behavior and the importance ofc ientists.©

K

1

twtnapkuctsopftqv

pro-ed tomers.ins ofy will

kindshen

ians

elyover,loca-sive, theiras-

delyr linkve

ratenot

0d

ombining several methods including neuroimaging, which has received recent attention by marketers, economists and social sc2005 Elsevier Inc. All rights reserved.

eywords: Association; Consumer behavior; Implicit cognition; Marketing; Neuroeconomics; Priming

. Introduction

Researchers in consumer behavior address several impor-ant and interesting questions. For example, how do we knowhether a new product will appeal to consumers and whether

hey will in fact purchase it? How do we know which brandame, logo or package design best fits the intended image ofbrand? How do we know if and how a particular brand is

erceived differently from a competing brand? How do wenow whether a brand name is appropriate for a given prod-ct? Answering these questions seems easy. Just go and askonsumers directly what their purchase intentions are, whathey like and do not like about a package design and howtrongly they feel about these attributes or whether they liker even remember a proposed name. Consumers will indeedrovide answers that they believe reflects their thinking. In

act, this way of answering such questions is common prac-ice in marketing and generally involves the use of standarduestionnaires administered in writing or by telephone inter-iew and by focus groups where 8–10 consumers under the

guidance of a group moderator will discuss a topic andvide their opinions. Face-to-face interviews are also usexplore conscious thoughts and feelings among consuThese same methods are used frequently in other domainquiry. Job candidates are asked about the strengths thebring to a company and are asked to complete variousof personality tests, patients describe their symptoms wasked by the physician and journalists interview politicand ask them about their plans for the future.

Verbal reports, whether written or spoken, are widaccepted as an important source of information. Morethey are easy to administer, can be implemented in anytion and in a variety of ways and they are often inexpenand quick. However, as widely used as verbal reports areutility in actually predicting behavior has become increingly unclear. For example, verbal self-reports are wiused in attitude research and a strong attitude–behaviohad been suggested[1]. However, researchers who hainvestigated this link empirically have failed to demonstit [23]. Often, when actual behavior is observed, it doescorrespond to the earlier expressed attitude.

∗ Corresponding author. Tel.: +41 1 634 15 60; fax: +41 1 634 15 89.E-mail address: f.mast@psychologie.unizh.ch (F.W. Mast).

There is mounting evidence that verbal reports, while use-ful in some instances, are much less reliable than previously

361-9230/$ – see front matter © 2005 Elsevier Inc. All rights reserved.oi:10.1016/j.brainresbull.2005.06.004

F.W. Mast, G. Zaltman / Brain Research Bulletin 67 (2005) 422–427 423

thought[16]. One reason they are unreliable as predictorsof consumer behavior is that they do not uncover or addressthe actual consumer cognitions which drive behavior. Simplybecause people can express apparent preferences when askeddoes not mean these are accurate reflections of their thoughtsand feelings or that they are the most relevant thoughts andfeelings for the topic being investigated[10,22]. Numerousstudies in psychology have revealed that self-report measurescorrelate poorly with more implicit measures of the same con-struct[7,8,18,20].

Still, in 2002, an estimated US$ 6.8 billion was spent onconventional marketing research tools in the United Statesalone. This is an enormous amount of money consideringthat there is little scientific evidence to support the widespreaduse of focus groups and growing concerns about the accuracyof other verbal report techniques. There are several reasonswhy conventional methods and in particular, verbal reportsobtained from consumers may be inaccurate. Some reasons,often noted in the psychological literature, are that partici-pants in a research project try to please the interviewer, havehypotheses about the goals of the research they are participat-ing in which, in turn, shapes their responses and try to conveya certain image of themselves. Another reason is that peoplemay not have complete or ready access to their own thoughtsand feelings and, even when they do, they may not be ableto articulate them well. Often, too, people are respondingt nk isi whatp picb mays ink-i ofa d tod oingt ectedfi ll them 80%o ringt

eful-n us-t ucts[ pur-c ases.S thev pur-c

allf wec e roleo ioust ageu era-t havea pro-c clea

that verbal protocols miss much information, which is abso-lutely essential for understanding how the mind works. Buthow can we tap into these implicit processes if languagedoes not do it? What is the key to the hidden motives? Dowe now need to bring in Freudian psychoanalysts to betterunderstand consumers’ thoughts? Freud’s merit is his con-ceptually pioneering work on introducing the unconsciousand the idea that many of our motives are out of reach ofconscious control can be captured by projective measures.The idea behind the use of projective measures is to presentan ambiguous stimulus or situation to a participant and to lethim or her interpret the stimulus. It is assumed that the hiddenmotives or intentions of the individual are projected onto thestimulus. Although projective measures are appealing, theyare time-consuming to administer and most of the tests arenon-standardized.

2. Response latency

Cognitive psychology has started to use response timesto study how the mind works. This approach is called men-tal chronometry and is strictly experimental. Based on howlong it takes the participants to perform a particular task,we can draw important inferences about the nature of theunderlying mechanisms. Response times are measured inm n, ina vationo so-n tative,a erfuls

digmo rma-t rk ed toi lusb eneda ere-f latedi Thep ane Fore ter atd sedt sawt ord“ edw een“ im-i notk hatk wen d int essS

o what the researcher thinks that study participants thimportant. These assumptions may not correspond toarticipants actually do think about with regard to the toeing investigated. So, the results of market researchimply reflect researcher thinking more than consumer thng [2,24]. In fact, there is evidence that as much as 80%ll market research is confirmatory, that is, it is intendeemonstrate a particular point of view held by those d

he research and as a result does not allow for unexpndings to surface. It is no surprise, then, that despite aoney spent on new product research that as much asf all new products are judged a commercial failure du

heir first year in the market.Compelling evidence suggesting caution about the us

ess of explicit information for marketing research is illrated by a study on purchase intentions for new prod15]. Indeed, there was no correlation between statedhase intentions for new products and their actual purchimilarly, purchase intentions provided verbally based oniewing of advertisements does not predict well actualhase behavior.

But if questionnaires, verbal reports and interviewsail to predict behavior, what other methods are thereould use instead? Perhaps, we are overestimating thf language to convey hidden knowledge or unconsc

houghts and feelings. Why should literal human langused in surveys be the tool to describe how mental op

ions work? Indeed, early experimental psychologistsssumed that introspection is a tool to study mentalessing. The insights remained moderate and it became

r

illiseconds and therefore have a high temporal resolutiony case better than most techniques that image the actif the human brain at work (e.g. functional magnetic reance imaging). Response time measures are quantind therefore, enable us to analyze the data with powtatistical tools such as analyses of variance.

A particular type of response time measure is the paraf priming. Consumers do not take in and process info

ion in a neutral and unbiased fashion[24]. Instead, prionowledge serves as an organizing framework that is usnterpret new information. The prime being the input stimurings conceptually related items into a state of heightccessibility. The response latency to certain stimuli th

ore reveals whether and to what degree conceptually retems are preactivated by a particular prime stimulus.rime stimuli are varied systematically thus allowing forxperimental manipulation of the context information.xample, it has been shown that participants are fasetecting the letters “doctor” as a real word (as oppo

o a random configuration of letters) when they justhe word “nurse” compared to when they just saw the wbread” [14]. “Nurse” and “doctor” are conceptually relathereas “bread” and “doctor” are not. The relation betw

nurse” and “doctor” is evident but the potential of the prng paradigm in marketing is to reveal relations we donow might exist. The question is what gets primed by wind of input. To better illustrate the use of priming,ow describe in more detail an experiment we conducte

he Mind of the Market Laboratory at the Harvard Businchool.

424 F.W. Mast, G. Zaltman / Brain Research Bulletin 67 (2005) 422–427

3. Using priming to study brand associations

The goal of this study was to investigate consumers’ asso-ciations related to “coke” and compare these associationswith those related to “water” (a local brand familiar to allconsumers tested was used in the study). The study waspurely exploratory in nature and not financed by a private cor-poration. Forty consumers (20 females and 20 males, agedbetween 19 and 33 years) volunteered to take part in thisstudy. They were either students from Harvard University orprofessionals from Boston area.

Prior to the study, we defined a set of concepts, whichwe thought would differentially distinguish between “coke”and “water”. We expected the concepts “enjoy”, “mystery”,“fresh taste” and “energy” to show stronger associations with“coke” and likewise, we expected the concepts “satisfy”,“nature”, “clean taste” and “vitality” to show stronger asso-ciations with water. Each of these concepts was representedby 20 different words defined prior to the study. These 20words were preselected from a list of 40 words provided toan independent group of consumers similar to those involvedin the response latency part of the study. They were asked torate to what extent each of these 40 words represented theconcept (e.g. how much does the word “delight” representthe concept of “enjoy”) on a scale from one to five. For eachconcept, we selected those 20 words, which best representedt

task,w andn ordso Con-s r thec ord.I 50%o o thew mso sholdo sses.T . ap r).A s ofp

whenp edw erei hadd re”,w hism tery”a ar-i ly,o dif-f datas ence.T wasf , the

“mystery” words were not differentially primed. The reversepattern was revealed for the “nature” words. Only the maleconsumers show the priming effect for “coke” on the “nature”words.

4. Behavioral relevance of priming studies

Clearly, the lab is not the store where many factors maybe operating. However, the artificial setting of the labora-tory provides a special advantage: we can isolate variables,which may be crucially involved in a particular marketingproblem. Whether the evidence revealed in the lab holdsup in a real world context is of course very important inmarketing and other settings. That is, are the effects wefind in a response time task behaviorally relevant? With ref-erence to the example above, will female consumers buy“coke” more often when it is presented in a “mystery” con-text? What about the “nature–coke” relation we revealed forthe men? Do male consumers prefer “coke” when they areoutside, for example, in a ballpark attending a game? Like-wise, it could make sense to present “coke” in magazines forsport and outdoor recreation, which are read more frequentlyby men.

Until now, only few studies have addressed the relation-ship between implict measures and actual purchase behavior.F -t sageb lim-i thisi psy-c vior.F edw esep uentc ertpi col-l ce atT lsh ivatet ore,p alsoh minge s cer-t ntsu ad ofp thei pur-c andm resentc

est.I reaterp aredt even

he concept.The words were then presented in a lexical decision

hich required consumers to distinguish between wordson-words. Consumers sit in front of a computer and wr non-words appear in the center of a computer screen.umers had to decide as quickly as possible whetheonfiguration of letters represented a word or a non-wn 50% of the cases, the consumers saw a word and inf the cases, the consumers saw a non-word. Prior tords or non-words, the prime stimuli appeared for 150n the computer screen. This duration is above the thref detection but too short for elaborated cognitive procehe primes were either pictures of “coke” or “water” (e.gicture of a “coke” bottle or of the brand of bottled watell words and non-words were combined with both typerimes.

For each word, we subtracted the response timesrimed with “coke” from the response times when primith “water”. Had there been no difference, the primes w

nefficient. However, this was not the case. The primesifferential effects on the concepts “mystery” and “natuhich were more strongly associated with “coke”. Teans that the participants responded faster to “mysnd “nature” when they were primed with “coke” in comp

son to when they were primed with “water”. Surprisingther concepts, such as “enjoy” and “fresh” showed no

erence between the two primes. We then analyzed theeparately by gender. The results showed a clear differhe strong connection between “mystery” and “coke”

ound only for female consumers. For the men only group

or example, Maison et al.[11] report that implicit associaions with juice and soda brands predict actual product uetter than explict self-reports. Even though there is only

ted evidence in the domain of consumer behavior sinces a new approach, there is a wealth of knowledge inhology showing that priming affects subsequent behaor instance, in a study in which participants were primith words belonging to the concept of “aggression”, tharticipants interrupted others more often in a subseqonversation[3]. Similar research in this vein finds that covrimes for helpfulness make people help more[13] while

ntelligence primes (i.e. that activate a stereotype of aege professor) temporarily improve people’s performanrivial Pursuit[6]. Thus, the implicit associations individuaold and the degree to which a given context might act

hem play an important role in guiding behavior. Therefriming not only helps uncover relevant thoughts butelps predict actual behavior. Can we extend these priffects to the domain of purchase behavior? The answer i

ainly yes. It is entirely possible to design field experimesing the methods described above. For example, insterime stimuli on the computer, it is possible to present

nformation at the entrance to the store or as point ofhase stimuli where a product is located within a storeeasure the actual purchases when the information is p

ompared to when it is absent.In this context, the “mere exposure effect” is of inter

t describes the phenomenon whereby people have a greference for objects they have previously seen comp

o new objects being viewed. This preference holds

F.W. Mast, G. Zaltman / Brain Research Bulletin 67 (2005) 422–427 425

when the objects were only polygons presented for a dura-tion which was too brief to consciously encode the stimuli[4]. The participants were later shown those polygons inter-spersed with new ones and they preferred the ones they wereexposed to before. Using the “mere exposure” technique, ithas been shown that fictitious brands can become “famousovernight”[9]. In this study, the consumers recognized thosebrands as famous, which they saw on the day before. Again,they had no conscious recollection of having been exposedto those names. Moreover, the “mere exposure” effect canalso account for the results reported by Shapiro et al.[17].The consumers read a magazine article delivered on a com-puter screen. The article was flanked by target ads designed toreceive minimal reader attention. Later, the consumers wereengaged in a buying scenario and they named eight productsconsidered for later purchase. No reference was made to theads they viewed before. In addition, the consumers performeda recognition task, in which they had to identify the ads theysaw from a list of four alternatives. Their conscious recog-nition did not differ from the control group, in which noneof the consumers actually saw the ads. Interestingly, the buy-ing scenario revealed that those consumers who had viewedthe ads in the periphery of their visual field were more likelyto include the corresponding products in their considerationset.

Importantly, response time techniques are not at all staticb xt inw iblet ea-s muli(w riort thep con-n ouldc text,a ar-v mings tudyw rticu-l milart xten-s mersa whop andc ars,w hichr fac-t

ea-sl ands atinga test,p k.T lly

found that the time required to name the color ink of thecolored words takes longer when the information is incon-gruent (e.g. the word “red” written in blue ink) than whenthe ink and word are congruent (e.g. the word “red” writtenin red ink). Therefore, the task to name the color of the inkinterferes with the intention to automatically read the colorwords.

How can we apply the Stroop paradigm in the context ofmarketing research? For example, it could be used to selecta name for a new product or service. Imagine you have aproduct (represented by different pictures) and alternativenames. The idea is to figure out whether there is an existingaffinity between the product and one of the alternative names.Consumers would see pictures of the product on the screenfollowed immediately by one of those potential names printedin color. The consumer has to read aloud the color in whichthe ink is printed. An existing connection between any ofthe names and the product would enhance response timeswhen naming the color. The implicit connection between theproduct and its potential name would make it harder to ignorethe meaning of the word and thus it will take the consumerslonger to solve the actual task, which is to name the color ofletters[5] (unpublished data).

We suggest that when comparing the results from theresponse time latency measures to explicit judgments basedon questionnaires or rating scales, the former will be morea tly, wet mok-i kersa “try-t ledw ol-l ther weens dis-c s yeta diesa

5

manya it isn romt comei haved neticr boutb muli[ haveb hilec Fore es,C stingi onal

ut rather flexible and easily adapted to the specific contehich the study will be conducted. For example, it is poss

o measure the affective connotation of products by muring positive–negative judgments to neutral target stimeaningless strings of letters, such asY*u7Z). Again, weould present as primes pictures of different brands p

o the target stimuli. The faster the consumers will hitositive response button, the more positive an affectiveotation is associated with the product. The results complement product positioning research. In this connother study form the Mind of the Market Laboratory at Hard Business School is of interest. We conducted a pritudy for a major auto manufacturer. The aim of the sas to reveal the attributes consumers associate with pa

ar automotive designs. The experimental design was sio the study with beverages described above with the eion of the specific comparisons with the attributes consussociate with the competitor’s designs. The consumersarticipated in this study were all experienced driversar owners. The prime stimuli were different pictures of chich preceded words like daring, fast and supple, w

evealed significant differences between different manuurers.

Priming is not the only possible effect we can mure with response time latency. In 1935, Stroop[21] pub-ished his landmark article on attention and interferenceince then, numerous studies have been done replicnd extending his initial results. In the classic Strooparticipants view words written in different colors of inheir task is to identify the color of the ink. It is usua

ccurate predictors of the most successful name. Recenested potential names for a new resource center on sng. This priming study revealed that consumers (smond non-smokers) responded differently to names like

ostop” or “make smoking history”. The latencies reveaith priming differed from the subjective ratings we c

ected with questionnaires, and most interestingly, onlyesponse time data revealed significant differences betmokers and non-smokers. The explicit survey failed toriminate between smokers and non-smokers. This inother demonstration that the results from priming sture behaviorally relevant.

. From response times to the brain

Even though the response time techniques havedvantages when compared to conventional tools,ot always possible to draw conclusive evidence f

he response time data. Recently, marketers have benterested in the advances in the brain sciences andiscovered that technologies such as functional magesonance imaging can provide potentially useful data arain activation patterns in response to marketing sti

24]. Researchers in both marketing and neuroscienceegun using neuroimaging to study brain activations wonsumers are involved in a marketing related task.xample, an fMRI study involving two popular beveragoca Cola and Pepsi Cola, provides a number of intere

nsights that might not be as readily obtained using traditi

426 F.W. Mast, G. Zaltman / Brain Research Bulletin 67 (2005) 422–427

methods[12]. One such finding was that brand knowledgefor Coca Cola has a stronger impact on behavioral preferencethan did brand knowledge for Pepsi Cola.

Unlike, for example, questionnaire data, neuroimagingdata are much less susceptible to demand characteristics orsocial desirability. We are not able to voluntarily control acti-vation within a given brain area. However, as increasinglyavailable these techniques have become, the more we need tobe able to design tasks that are meaningful to consumers. Theimportance of this point is often underestimated. The rapidrise of cognitive neuroscience was possible only because awealth of knowledge was already available from decades ofresearch in the neurosciences and cognitive psychology. Inorder to make neuroimaging a useful method for marketingresearch, one needs to be aware that the starting conditionsfor its use in marketing are totally different than those ofcognitive neuroscience. Going from focus groups and opin-ion polls to the use of neuroimaging tools is a much biggerleap than was necessary going from cognitive psychology tocognitive neuroscience.

The use of neuroimaging techniques in marketing con-texts is very much in its infancy. Although very promising,a great deal of experimental work is required to determinewhat set of problems are best addressed using this approachand how neuroimaging techniques might best supplementexisting methods. Neuroimaging does not generate hypothe-s ssaryt n ther gingm tantiakC xten-s estsa re isa orea em-o urseo ore-o entsa wardp areao ing.O ausei thata out-cu uringe einge

eralr standh s thati etterp ar-k ioralv rela-

tion between conscious and unconscious information and itsimpact on consumer behavior. While explicit and implicitknowledge may sometimes correspond, the exciting oppor-tunity for marketers is that often there is a discrepancy, i.e.what a consumer reports explicitly, their espoused attitudesand beliefs, may have no bearing on their actual behavior. Theuse of implicit measures will provide more reliable predic-tions and may also help us learn how to improve the designand use of explicit measures.

References

[1] I. Ajzen, M. Fishbein, The prediction of behavioral intentions in achoice situation, J. Exp. Soc. Psychol. 5 (1969) 400–416.

[2] V. Barabba, G. Zaltman, Hearing the Voice of the Market, HarvardBusiness School Press, Boston, Mass, 1991.

[3] J. Bargh, M. Chen, L. Burrows, Automaticity of social behavior:direct effects of trait constructs and stereotype activation on action,J. Pers. Soc. Psychol. 71 (1996) 230–244.

[4] R.F. Bornstein, D.R. Leone, D.J. Galley, The generalizability of sub-liminal mere exposure effects: influence of stimuli perceived withoutawareness of social behavior, J. Pers. Soc. Psychol. 53 (1987) 70–79.

[5] K.A. Braun-Latour, N.M. Puccinelli, F.W. Mast, Negative affectstates: do they impair or enhance consumer cognition, submittedfor publication.

[6] A. Dijksterhuis, R. Spears, T. Postmes, D.A. Stapel, W. Koomen,doing. Psy-

es,

ndi-est,

sure:Mar-

[ A,

[ testll. 2

[ .R.rally

[ ction

[ s ofns, J.

[ tionseting

[ orsiness

[ l ades. 24

[ licitPsy-

es and to formulate interesting hypotheses, it is neceo have a deep understanding of the issues involved iesearch context. For this reason, it is likely that neuroimaethods might best be used in areas where there is subs

nowledge already. The McClure et al. study[12] involvingoca Cola and Pepsi did just that since there is an eive literature on branding, blind and non-blind taste tnd the role of brand image and cultural meaning. Thelso promise in using neuroimaging to help to learn mbout how marketing information is consolidated in mry or how the activation patterns change over the cof multiple exposures to a particular advertisement. Mver, it might reveal more about the affective componssociated with a product or about the anticipated reromised by the purchase of a product. A very interestingf application involves the basic dynamics of choice makne neuroimaging study, potentially very important bec

t challenges a central assumption in economics, foundttitudes about payoffs and beliefs about the likelihood ofomes interact both behaviorally and neurally[19]. It can helps determine whether an advertisement is having an endmotional impact or whether particular metaphors are bffective[24,25].

In sum, we think that converging evidence using sevesearch methods in combination is key to better underow consumers think. Considerable evidence suggest

n many circumstances, an implicit measure may be a bredictor of actual consumer behavior. In the future, met research that combines explicit, implicit and behavariables will be necessary to further reveal the inter

l

A. van Knippenberg, D. Scheepers, Seeing one thing andanother: contrast effects in automatic behavior, J. Pers. Socchol. 75 (1998) 862–871.

[7] A.G. Greenwald, M.R. Banaji, Implicit social cognition: attitudself-esteem and stereotypes, Psychol. Rev. 102 (1995) 4–27.

[8] A.G. Greenwald, D.E. McGhee, J.L.K. Schwartz, Measuring ividual differences in implicit cognition: the implicit association tJ. Pers. Soc. Psychol. 74 (1998) 1464–1480.

[9] J.S. Holden, M. Vanhuele, Know the name, forget the expobrand familiarity versus memory of exposure context, Psychol.ket. 16 (1999) 479–496.

10] B. Libet, Mind Time, Harvard University Press, Cambridge, MUSA, 2004.

11] D. Maison, A.G. Greenwald, R. Bruin, The implicit associationas a measure of implicit consumer attitudes, Pol. Psychol. Bu(2001) 61–79.

12] S.M. McClure, J. Li, D. Tomlin, K.S. Cypert, L.M. Montague, PMontague, Neural correlates of behavioral preference for cultufamiliar drinks, Neuron 44 (2004) 379–387.

13] C.N. McCrae, L. Johnston, Help, I need somebody: automatic aand inaction, Soc. Cogn. 16 (1998) 400–417.

14] D.E. Meyer, R.W. Schaneveldt, Facilitation in recognizing pairwords: evidence of a dependence between retrieval operatioExp. Psychol. 90 (1971) 227–234.

15] V.G. Morwitz, J.H. Steckel, A. Gupta, When do purchase intenpredict sales, Report No. 97–112, Working Paper Series, MarkScience Institute, 1997.

16] N.M. Puccinelli, K. Braun, F. Mast, G. Zaltman, Implicit predictof consumer behavior, Case No. 502-043, 2001, Harvard BusSchool.

17] S. Shapiro, D. MacInnis, S. Heckler, The effects of incidentaexposure on the formation of consideration sets, J. Consum. R(1997) 94–104.

18] J.J. Skowronski, M.A. Lawrence, A comparative study of the impand explicit gender attitudes of children and college students,chol. Women Q. 25 (2001) 155–165.

F.W. Mast, G. Zaltman / Brain Research Bulletin 67 (2005) 422–427 427

[19] K. Smith, J. Dickhaut, K. McCabe, J.V. Pardo, Neuronal substratesfor choice under ambiguity, risk, gains, and losses, Manage. Sci. 48(2002) 711–718.

[20] W.D. Spangler, Validity of questionnaire and TAT measures ofneed for achievement: two meta-analyses, Psychol. Bull. 112 (1992)140–154.

[21] J.R. Stroop, Studies of interference in serial verbal reactions, J. Exp.Psychol. 18 (1935) 634–662.

[22] D. Wegner, The Illusion of Conscious Will, MIT Press, Cambridge,MA, USA, 2002.

[23] A.W. Wicker, Attitudes versus actions: the relationship of verbaland overt behavioral responses to attitude objects, J. Soc. Issues 25(1969) 41–78.

[24] G. Zaltman, How Consumers Think, Harvard Business School Press,Boston, Mass, 2003.

[25] G. Zaltman, Eliciting mental models through imagery, in: A.M. Gal-aburda, S.M. Kosslyn, Y. Christen (Eds.), The Language of the Brain,Harvard University Press, Cambridge, MA, USA, 2002, pp. 363–375.