Food Quality and Preference 37 (2014) 19–26

Contents lists available at ScienceDirect

Food Quality and Preference

journal homepage: www.elsevier .com/locate / foodqual

Evaluative conditioning of food technologies in China: Moderating effectof social trust

http://dx.doi.org/10.1016/j.foodqual.2014.04.0160950-3293/� 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author. Tel.: +45 8716 5189.E-mail addresses: n.loebnitz@asb.dk (N. Loebnitz), klg@asb.dk (K.G. Grunert).

1 Tel.: +45 8716 5007.

Natascha Loebnitz ⇑, Klaus G. Grunert 1

Business and Social Sciences, Department of Business Administration, MAPP Centre for Research on Customer Relations in the Food Sector, Aarhus University, Bartholins Allé 10,8000 Aarhus, Denmark

a r t i c l e i n f o a b s t r a c t

Article history:Received 26 September 2013Received in revised form 18 April 2014Accepted 18 April 2014Available online 26 April 2014

Keywords:Evaluative conditioningFood technologiesSocial trustChina

This study provides an initial examination of the evaluative conditioning (EC) of consumers’ attitudestoward food technologies in China, including how EC can affect consumer acceptance of new technologywhen participants possess different levels of social trust. In a study using the EC paradigm and a combi-nation of between-subjects control groups and within-subjects control conditions, participants consid-ered three food technologies (conventional, enzyme, and genetic), paired with affectively positive,neutral, and negative images. Subsequent evaluative measurements revealed that EC can explain attitudeformation toward food technologies in China when consumers see affective images, but the strength ofthe effects varies at different levels of social trust. Participants with a high level of trust in the institutionsthat promote and regulate the technologies can be conditioned both positively and negatively, indepen-dent of food technology. Participants with a low level of trust can be conditioned too, but only when thetechnology is paired with negative unconditioned stimuli. If social trust is low, positive conditioning offood technologies is not demonstrated in this study.

� 2014 Elsevier Ltd. All rights reserved.

Introduction et al., 2011), only one study adopts an EC framework to investigate

Consumer attitudes toward new food technologies vary acrosscountries: Public opinions in Europe are generally ambivalent orcritical towards high-tech foods, whereas in China, consumers seemmore positive toward them (Zhang, Huang, Qiu, & Huang, 2010).These attitudes in turn influence purchasing decisions, leading man-agers and food researchers to seek out explanations of how consum-ers’ attitudes toward different food technologies form. Althoughresearchers have documented widely divergent attitudes todifferent food technologies (e.g., organic production versus geneticmodification), our knowledge about how consumers develop theseattitudes remains limited.

In this context, we turn to the concept of evaluative conditioning(EC), which refers to a change in the valence of a conditioned stim-ulus (CS) due to its pairing with another, unconditioned stimulus(US) (De Houwer, 2007). Many studies demonstrate EC effects in afood context and show that EC can explain the acquisition of foodpreferences (Kerkhof, Vansteenwegen, Baeyens, & Hermans, 2009;Verhulst, Hermans, Baeyens, Spruyt, & Eelen, 2006). Although somescholars allude to EC as a post hoc explanation of consumer accep-tance of food technologies (Olsen, Grunert, & Sonne, 2010; Olsen

attitude formation (Loebnitz & Grunert, 2013). At least two issuesdemand further exploration. First, previous studies primarily con-sider the EC paradigm in relation to developed, mostly Westernnations (i.e., United States, European countries). Yet China is oneof the largest producers of genetically modified crops in the world(Curtis, McCluskey, & Wahl, 2004), and Chinese attitudes towardnew food technologies appear to diverge from those of Europeanconsumers, so it is imperative to understand how Chinese consum-ers form attitudes toward new food technologies. In an extension ofrecent findings that indicate EC can explain attitudes toward foodtechnologies (Loebnitz & Grunert, 2013), we ask, Does evaluativeconditioning apply in the context of attitudes to food technologiesin China? Second, past research indicates that Chinese consumershave more positive attitudes regarding food technologies (i.e.,genetic modification; Zhang et al., 2010) and higher trust in institu-tions promoting and regulating these technologies than consumersin other countries (Curtis et al., 2004). Social trust is an importantdeterminant of technology’s acceptance (Siegrist, 2000), so is socialtrust perhaps driving Chinese attitudes toward food technologies?In answering these questions, we seek to accomplish two objec-tives: (1) investigate if attitude formation toward food technologiesin China can be explained by EC and (2) discover if social trust mod-erates the evaluative conditioning effect on consumers’ attitudetoward food technologies. Accordingly, we collected data from con-sumers in China.

20 N. Loebnitz, K.G. Grunert / Food Quality and Preference 37 (2014) 19–26

Literature review

Consumer acceptance of food technology

Two main lines of thought attempt to explain how attitudes totechnologies form and become integrated into overall productjudgments. First, in a top-down approach, socio-political attitudespredict consumers’ attitudes toward different food technologies.Thus for example, views on nature and the environment (Frewer,Howard, Hedderley, & Shepherd, 1997), science and technology(Hamstra & Smink, 1996), social trust (Siegrist, 2000), and indus-trial food production (Beckmann, Brokmose, & Lind, 2001) guideoverall attitude formation and serve as a higher-order anchor forevaluations of a food technology (Grunert, Bredahl, & Scholderer,2003). Consumers who have highly positive attitudes toward nat-ure and the environment then should have less favorable attitudestoward high-tech food technology, especially if they seem to inter-fere with nature, such as genetic modification. Social trust is one ofthe strongest predictive factors for the formation of attitudestoward food technologies (Qiu, Huang, Pray, & Rozelle, 2012),though Søndergaard, Grunert, and Scholderer (2005) find consider-able differences in the influence of social trust on attitude forma-tion across European countries, so it may be culturally specific(Poppe & Kjærnes, 2003). In China, consumers appear more trust-ing of institutions that promote and regulate food technologies(Curtis et al., 2004); in Europe, social trust is heterogeneous, withScandinavian consumers conveying more trust than British con-sumers (Frewer, 1999). Generalizing from these international dif-ferences in social trust, we anticipate that consumers’ attitudeformation is culturally dependent (Gaskell, Allum, & Stares, 2003;Poppe & Kjærnes, 2003). Independent of the country of investiga-tion though, prior results suggest a positive relationship betweensocial trust and consumers’ acceptance of food technology (Lin,Somwaru, Tuan, Huang, & Bai, 2006).

Second, the bottom-up approach emphasizes how informationabout the potential risks and benefits of a food technology caninfluence consumer acceptance (Grunert et al., 2003; Scholderer& Frewer, 2003). Various studies investigate the effects of givingconsumers information (e.g., benefits, risks, consequences; Olsenet al., 2010) by measuring consumer acceptance in terms of prefer-ences (Lähteenmäki et al., 2002), intentions to buy (Grunert,Bech-Larsen, Lähteenmäki, Ueland, & Aström, 2004), attitudetoward the technology (Eiser, Miles, & Frewer, 2002), and percep-tions (Siegrist, 2000) as dependent variables. Thus, food research-ers seeking to modify attitudes toward new technologies usingthe bottom-up approach focus on providing educational informa-tion (Teisl, Fein, & Levy, 2009), which emphasizes the benefitsand deemphasizes their risks.

Both approaches have merits, yet they may fail to account forthe emotional aspects of the process. Existing studies only dealwith cognitive processes that lead to acceptance or rejection of atechnology but ignore any corresponding affective processes thatare not grounded in cognitive processing of information. This gapmay explain two conundrums in prior research: (1) Although theprovision of information is generally effective in increasing knowl-edge, it cannot change existing attitudes and only reinforces nega-tive attitudes (Frewer, Scholderer, & Bredahl, 2003) and (2) attitudeformation still occurs, even in cases with limited information(Olsen et al., 2010).

Evaluative conditioning

Evaluative conditioning refers to the extent to which pairing anaffectively meaningful with a neutral stimulus changes the valence

of the neutral stimulus (Walther, Weil, & Langer, 2011). The cooc-currence of positive/negative images (US) with a neutral stimulus(CS) may result in a spillover of the positivity/negativity from theunconditioned to the conditioned stimulus (for a review, seeHermans, Baeyens, & Eelen, 2003). Hofmann, De Houwer, Perugini,Baeyens, and Crombez (2010) also describe two characteristics ofEC that distinguish it from classical conditioning: (1) EC takes placewithout contingency awareness (Olson & Fazio, 2001), though ECeffects are stronger for participants with higher compares withlower contingency awareness, and (2) EC effects appear resistantto extinction, such that CS-alone presentations will not interferewith them (Baeyens, Diaz, & Ruiz, 2005), though they eventuallymay decrease in extinction studies (Hofmann et al., 2010).

Some studies fail to confirm EC effects (e.g., Field & Davey,1999), but researchers across different disciplines have demon-strated that preferences can form and attitudes change throughthis conditioning (De Houwer, Thomas, & Baeyens, 2001). Further-more, it is reasonable to assume that real-world factors such asadvertising (Gibson, 2008; Stuart, Shimp, & Engle, 1987) and brandplacement (Schemer, Matthes, Wirth, & Textor, 2008) result in ECeffects when a previously neutral/novel stimulus is contingentlypaired with another (positive or negative) stimulus, resulting inconsumer liking or disliking. Empirical research also indicates thatparticular technologies (e.g., genetic modification of food) oftenappear in negative contexts, while others appear in positive con-texts. As a result of negative media coverage in Europe for example(Gaskell, Bauer, Durant, & Allum, 1999), genetic technology is oftenassociated with high risk and uncertainty (Bredahl, 2001), whereasconventional technology appears congruent with high consumerbenefits (e.g., health, safety; Grunert et al., 2001).

Significant research has applied EC to the acquisition of foodpreferences, suggesting that food likes/dislikes can be explainedaccording to an EC paradigm (Kerkhof et al., 2009; Verhulst et al.,2006). Beyond studies that use a standard picture–picture para-digm (visual stimuli) to examine food preference formation,researchers confirm EC effects using sensory liking (flavor;Verhulst et al., 2006), expected consequences (Verhulst et al.,2006), odors (Hermans, Baeyens, Lamote, Spruyt, & Eelen, 2005),and gustatory stimuli (Zellner, Rozin, Aron, & Kulish, 1983). Afew researchers even have shown that it is possible to change con-sumer attitudes toward new food technologies by exposing themto a superior sensory product experience, which induces positivesensory-based affect (Grunert et al., 2004; Scholderer, Grunert, &Søndergaard, 2006). For example, Grunert et al. (2004) demon-strate that attitudes to new food technologies grow more positiveafter trial of products that contain the technology, though only ifthose products produce a positive sensory experience. Similarly,Scholderer et al. (2006) find that participants evaluate a technologymore positively after even a single trial with a superior sensoryproduct experience. Olsen et al. (2011) confirm the effect of a sen-sory product experience on consumer attitudes to food technolo-gies and further note that the weight consumers grant todifferent production methods in their product evaluations dependon their product experience. Participants who taste the productprior to their choice place less emphasis on the production methodthan participants with no sensory product experience. In summary,the cooccurrence of a positive (superior) sensory experience with anovel stimulus (technology) can result in a spillover of the positiv-ity of taste to the stimulus. These findings allude to EC, but moststudies use it solely as a post hoc explanation of unpredictedresults, making it impossible to conclude that EC is responsiblefor the attitude change. Olsen et al. (2011) suggest the need tounderstand EC as an explanatory mechanism of consumer prefer-ence formation for food technologies.

N. Loebnitz, K.G. Grunert / Food Quality and Preference 37 (2014) 19–26 21

Moderators of evaluative conditioning

Considering the ample evidence that people differ in the extentto which they accept food technologies, because of the differencesin their socio-political attitudes, practitioners are well advised toinvestigate which consumers are more likely to be influenced byconditioning procedures. Empirical evidence suggests that atti-tudes toward food technologies are mainly driven by generalsocio-political attitudes and social trust in particular. We thereforepropose that social trust is a moderator of the EC effect on con-sumer attitudes toward food technologies.

Hofmann et al. (2010) identify several other moderators thatunderlie the effects of EC, but those moderators represent featuresof the EC procedure (e.g., the modality of US or CS), not potentialmoderating influences of individual attitudinal perspectives thatexist prior to the EC task. A few studies also investigate thestrength of the EC mechanism; for example, for EC procedures tosucceed, the experimental stimuli should avoid eliciting prejudices(Bierley, McSweeney, & Vannieuwkerk, 1985; Blair & Shimp, 1992).By using standard pictures for the CS (snack food) but unpleasantpictures for the US (obesity as a consequence of eating unhealthyfoods), Hollands, Prestwich, and Marteau (2011) show that theEC effects on implicit attitudes are greater among participants withpositive prior implicit attitudes toward snacks. They find no ECeffects for participants with negative attitudes toward CS prior tothe EC task, suggesting ceiling effects as potential explanation.Because they only employ aversive images for the US, their studyis inconclusive regarding the moderating effects of prior attitudeswhen positive images represent the US. In another study,Schemer et al. (2008) use a different conditioning paradigm,employing a fictitious brand as the CS and rap artists with eitherpositive or negative reputations as the US. In addition to demon-strating successful EC, this study investigated music preferenceas moderator. Schemer et al. propose that a rap music preferenceresults in EC effects independent of US valence, whereas dislikeof rap music reinforces negative conditioning and attenuates posi-tive conditioning. Their results show that participants’ preferencefor rap music strengthens the EC effect, but a lack of preferencereduced the influence of EC, regardless of US valence. Participantswith a preference for a particular music genre thus could be condi-tioned in either direction, but those without preferences could notbe conditioned at all. Schemer et al. do not address the effect ofindividual attitudinal perspectives prior to the EC task though;instead, they use music preference as a moderating effect, whichpertains not to CS or US but rather to the experimental setting.

We propose that social trust is critical for determining whetherconsumers’ attitudes toward food technologies can be positively ornegatively conditioned. Although EC effects are stronger when par-ticipants have no knowledge of the used stimuli (Cacioppo,Marshall-Goodell, Tassinary, & Petty, 1992), food technologies con-stitute a special case. Previous research has shown that even whenconsumers have very limited knowledge, they still have attitudesabout food technologies. Siegrist (2000) reveals that trust has apositive impact on their perceived benefits and a negative influ-ence on perceived risk. If consumers have only negative associa-tions with a particular food technology (perceived risks), ECcould simply activate existing negative associations (Dasgupta &Greenwald, 2001). If consumers lack trust in the institutions pro-moting and regulating these technologies for example, the condi-tioning procedure is unlikely to work as expected. It also seemsunlikely that a brief pairing of food technologies with positiveimages would be sufficient to affect skeptics or opponents of foodtechnologies. Rather, we expect that for consumers with low trust,EC functions only in the negative direction, reinforcing opinions bypairing the food technology with negatively valenced images.

Consumers with little social trust instead are less likely to be influ-enced by EC when the food technology appears paired with a posi-tive image. However, if consumers express trust in the institutionspromoting and regulating these technologies, conditioning proce-dures might work as expected (Hollands et al., 2011; Schemeret al., 2008), such that consumers undergo a conditioningprocedure when the CS is paired with either positive or negativeimages.

Method

This study replicates, in China, the EC framework employed byLoebnitz and Grunert (2013), which successfully conditioned con-sumers’ attitudes toward food technologies in Denmark. In addi-tion, it extends the EC replication by including social trust as apotential theoretical contribution. The study features a between-subjects control group and standard, within-subjects control con-ditions, such that the CS–US pairings were subject to a 3 � 3 Latinsquare design, enabling us to control simultaneously for the effectof valence and technologies. With the applied nature of our study,we sought to extend prior EC studies in three ways: (1) We usedthree existing food technologies that elicit variations in consumers’evaluative measurements and represent realistic options inreal-life settings; (2) we employed simultaneous CS–US presenta-tions (Pleyers, Corneille, Luminet, & Yzerbyt, 2007), becauseconsumers often see positive/negative media coverage ofparticular technologies; and (3) we investigated social trust as amoderating influence.

Participants

Participants (n = 936) were recruited from a consumer panel inChina. Three participants were excluded because their total time tocomplete the study fell below the minimum pre-tested time(n = 25), which left 933 participants for hypothesis testing.

Their average age was 39.8 years, 57% were men (43% female),they considered their financial situation modest (31.8%) to reason-able (41.3%), and they were relatively well educated, such that65.3% had a university or college degree. Furthermore, 82.3% weremarried, 13.2% lived alone, and 69.3% resided in big cities or towns,whereas only 19.1% lived in small cities.

Stimulus material

The three CS were different technologies used in food produc-tion: conventional, enzyme, and genetic technology. These threetechnologies represent likely variations in terms of expected con-sumer unease, from low (conventional) to high (genetic) andincluding a midpoint (enzyme). We used these generic terms forthe technologies to avoid evoking different levels of concern duesolely to terminology. Thus for example, we used genetic technol-ogy instead of genetic modification. At the outset of the experi-ment, all respondents received a brief text describing eachtechnology, as follows:

a. Conventional technology has been around for many years.This technique involves using chemical substances to extractand change food ingredients. Through conventional technol-ogy it is possible to achieve traditional flavor under anextensive energy production.

b. Enzyme technology is a recent development in modern bio-technology. This technique involves using enzymes toextract and change food ingredients. Through enzyme engi-neering it is possible to achieve healthier food under a sus-tainable production.

22 N. Loebnitz, K.G. Grunert / Food Quality and Preference 37 (2014) 19–26

c. Genetic technology has been around for many years in bio-technology. This technique involves changing genetic struc-ture of food ingredients. Through genetic technology it ispossible to achieve food with improved properties for theconsumer.

The nine US consisted of three positive (balloons 8162, a sum-mer lake 5760, and mountains 8190), three negative (pile of skulls9440, shark with open mouth 1931, and cigarette butts 9830), andthree neutral non-food related pictures (reading newspaper on abench 2102, napping on the subway 2397, neutral face expression2570) taken from the International Affective Picture System (IAPS)picture database. Pictures from this database appear in many ECstudies (e.g., Sweldens, Van Osselaer, & Janiszewski, 2012). Pretestdata, obtained from a different sample (n = 17), provided Likertevaluations of various US (1 = very negative, 7 = very positive),which indicated the pictures with the highest (M = 6.21; SD = .26)and lowest (M = 1.56; SD = .43) scores. We used them for the posi-tive and negative versions, respectively, then selected a picturethat scored 3–4 for the neutral image (M = 3.59; SD = .71).

Evaluative measure

To obtain evaluative measurements of the three technologies,we used three items and a seven-point Likert scale: ‘‘Using conven-tional/enzyme/genetic technology in food production is. . .’’ (1 =extremely bad, 7 = extremely good); ‘‘Using conventional/enzyme/genetic technology in food production is. . .’’ (1 = extre-mely foolish, 7 = extremely wise); and ‘‘I am. . .’’ (1 = stronglyagainst using conventional/enzyme/genetic technology in foodproduction, 7 = strongly for using conventional/enzyme/genetictechnology in food production). We gathered responses to thesemeasures twice, once before (EM1) and once after (EM2) the con-ditioning task. The evaluative measurement items for each CS thenwere averaged (Cronbach’s alphas, EM1: conventional = .881;enzyme = .890; genetic = .945; EM2: conventional = .920;enzyme = .946; genetic = .967).

Conditioning procedure

Participants were exposed repeatedly to simultaneous presen-tations of the positive, negative, or neutral images selected fromthe IAPS and different food technologies (conventional, enzyme,genetic), on a computer screen. To avoid demand characteristics,we provided a modified version of Pleyers’ et al. (2007) instruc-tions to explain the task. Participants were instructed as follows:

The study comprises two phases. In the first phase, you will seea newspaper headline involving various food technologies with apicture appearing underneath the headline but not the actualnewspaper article text: just a headline and a picture. They willbe presented in random order by the computer program (you donot need to memorize them). In the second phase, you will simplybe asked to spontaneously answer a set of questions. Please pressthe space bar to start the experiment.

Following Baeyens et al. (2005), we used two possible presenta-tion schedules (Loebnitz & Grunert, 2013), which varied with theparticipants’ assigned condition.

Paired conditionWith a 3 � 3 Latin square design, we simultaneously controlled

for the effect of valence and technologies. The three treatment con-ditions ensured that each CS paired with all three US across partic-ipants, namely,

� G1: conventional (N), enzyme (�), and genetic (+).� G2: conventional (�), enzyme (+), and genetic (N).

� G3: conventional (+), enzyme (N), and genetic (�).

The experimental design thus involved nine combinations oftreatment levels and blocking variables. For each participant, a ran-dom CS picture always paired with the same US picture. The CS–USassignments were presented seven times, mixed with four CS-onlytrials, for a total of 33 presentation trials in random order. Each CS–US trial consisted of one CS, positioned as a headline atop a US.Each of the seven CS–US pairs displayed on the computer screenfor 1 s, directly followed by a dark screen for 1500 ms.

Unpaired conditionParticipants assigned to the block/sub-block (BSB) condition

(Field, 1996) were subjected to the same CS and US as participantsin the paired condition, but to avoid CS–US associations, we pre-sented them in isolated blocks (i.e., CS block and US block). Inthe CS (US) block, each CS (US) paired seven (five) times with itself,using the same parameters as in the CS–US trials in the paired con-dition. Thus, the sequence was as follows: simultaneous presenta-tion of stimulus for 1 s, followed by a black screen for 1500 ms,then another stimulus presentation for 1 s, and so on, until thestimulus had been presented 14 times in the CS block and 10 timesin the US block. Half the participants saw the CS block followed bythe US block, while the other half saw the US block before the CSblock (Baeyens et. al., 2005; Field, 1996).

Contingency awareness

Immediately after the EC task, we collected a contingencyawareness measurement, using similar questions to the ones posedby Baeyens, Eelen, Crombez, and Van den Bergh (1992), about eachCS. The questions assessed participants’ knowledge of the CS–UScontingency relationship. Specifically, participants rated each CSon three items: (1) ‘‘Do you think that it was followed by a partic-ular image?’’(yes; no; do not know); (2) ‘‘Indicate whether it wasconsistently presented with positive, negative, or neutral imagesduring the first phase of the experiment’’ (possible answersincluded ‘‘do not know’’); and (3) ‘‘Please choose out the followinglist of images which image it was paired up with.’’ A response thatwas accurate for all three questions and very confident received ascore of 2, a response that was only accurate on the picture selec-tion question received a score of 1, and a response that was com-pletely inaccurate received a 0 score. We then totaled the scoresfor each participant, for a potential range of 0–6, such that highernumbers indicated higher contingency awareness.

Socio-political attitudes

Participants indicated, on a seven-point scale (1 = no trust,7 = high trust) how much trust they had in various institutions(Siegrist, 2000). At the end of the questionnaire, we collecteddemographic information.

Results

Conditioning effects

We began the analysis by running repeated measures, multipleanalyses of variance (MANOVA) to test for changes in the evalua-tive measurements (EM1, EM2) and how they were affected bythe paired condition (G1, G2, G3, BSB). The analysis indicated a sig-nificant interaction between the moment of evaluative measure-ment (EM1, EM2) and the experimental group (F(9, 2787) = 29.1,p < .001). Thus, for different paired conditions, the change in eval-uative measurement from before to after the EC task varied. We

Panel a: Conventional technology

Panel b: Enzyme technology

4

4.2

4.4

4.6

4.8

5

5.2

5.4

5.6

EM 1 EM 2

Mea

n C

S - C

onve

ntio

nal T

echn

olog

y

Evaluative Measurement

G 1G 2G 3BSB

4

4.2

4.4

4.6

4.8

5

5.2

5.4

EM 1 EM 2

Mea

n C

S - E

nzym

e Te

chno

logy

Evaluative Measurement

G 1G 2G 3BSB

N. Loebnitz, K.G. Grunert / Food Quality and Preference 37 (2014) 19–26 23

used contrasts to break down this interaction, which revealed asignificant effect of the pairing group in the before-and-after com-parisons for conventional technology (F(3, 929) = 35.42, p < .001),enzyme technology (F(3, 929) = 26.98, p < .001), and genetic tech-nology (F(3, 929) = 24.32, p < .001). For conventional technology,simple effects revealed significant differences for paired conditionG2 (negative valence) (F(1, 929) = 94.47, p < .001; MEM1 = 5.03,SD = .10; MEM2 = 4.33, SD = .10) and G3 (positive valence) (F(1,929) = 14.54, p < .001; MEM1 = 5.12, SD = .10; MEM2 = 5.39,SD = .10) but not for G1 (neutral) (F(1, 929) = .09, p = .764; ns;MEM1 = 5.03, SD = .10; MEM2 = 5.05, SD = .10).

For enzyme technology, simple effects revealed significant dif-ferences for paired condition G1 (negative valence) (F(1,929) = 90.73, p < .001; MEM1 = 5.12, SD = .09; MEM2 = 4.42,SD = .09) and G2 (positive valence) (F(1, 929) = 4.57, p = .033;MEM1 = 5.16, SD = .09; MEM2 = 5.32, SD = .09) but not for G3 (neu-tral) (F(1, 929) = 1.01, p = .315; ns; MEM1 = 5.17, SD = .90;MEM2 = 5.10, SD = .09).

For genetic technology, simple effects revealed significant dif-ferences for paired condition G1 (positive valence) (F(1,929) = 14.73, p < .001; MEM1 = 3.62, SD = .12; MEM2 = 3.96,SD = .12) and G3 (negative valence) (F(1, 929) = 62.35, p = .033;MEM1 = 3.69, SD = .12; MEM2 = 2.98, SD = .12) but not for G2 (neu-tral) (F(1, 929) = .112, p = .737; ns; MEM1 = 3.45, SD = .13;MEM2 = 3.48, SD = .13). Fig. 1 contains graphs of the mean evalua-tive measurements of conventional technology (Panel a), enzymetechnology (Panel b), and genetic technology (Panel c) for eachpaired condition before and after the EC.

In the BSB group, the repeated measure MANOVA for the threeCS (EM1, EM2) revealed, as expected, no significant changes in EMfor conventional technology (F(1, 235) = .716, ns, MEM1 = 4.99,SD = .10; MEM2 = 5.02, SD = .10), enzyme technology (F(1,235) = .612, ns, MEM1 = 5.08, SD = .09; MEM2 = 5.11, SD = .09), orgenetic technology (F(1, 235) = 1.13, ns, MEM1 = 3.48, SD = .12;MEM2 = 3.44, SD = .12). Furthermore, in the control BSB group, wefound no effect of pair type order (CS block–US block versus USblock–CS block, p < .05).

Panel c: Genetic technology Notes: G1 = positive; G2 = neutral; G3 = negative.

2.5

2.7

2.9

3.1

3.3

3.5

3.7

3.9

4.1

EM 1 EM 2

Mea

n C

S - G

enet

ic T

echn

olog

y

Evaluative Measurement

G 1G 2G 3BSB

Fig. 1. Evaluative measurements of different food technologies, before and after EC,across different paired conditions.

Social trust as a moderator of EC effects

To examine the main effect of social trust on evaluative mea-surements of different food technologies, we included social trustin the preceding analysis as a covariate. It emerged as a significantpredictor of participants’ evaluations of food technologies (conven-tional technology, F(1, 929) = 46.244, p < .001; enzyme technology,F(1, 929) = 142.55, p < .001; genetic technology, F(1, 929) =274.903, p < .001). To determine whether the EC effect was greaterfor those with relatively high versus little to no trust, we catego-rized participants as high and low in social trust (median split)and included it as a between-subjects factor. Our data analysis thusfeatured a 4 (paired condition: G1, G2, G3, BSB) � 2 (social trust:high and low social trust) � 2 (evaluative measurement: EM1,EM2) analysis of variance (ANOVA), with repeated measures onthe last variable. The paired condition � social trust � evaluativemeasurement interaction (F(9, 2775) = 4.35, p < .001) indicatedthat evaluative measurements for the three technologies dependon both the level of social trust and the pairing group to whichthe participants belonged. To tease this effect apart, we conductedtwo separate 4 (paired condition: G1, G2, G3, BSB) � 2 (evaluativemeasurement: EM1, EM2) ANOVAs for each technology (conven-tional, enzyme, genetic), according to whether participants werecategorized as low or high trust.

In each ANOVA we uncovered a paired condition � evaluativemeasurement interaction, whether participants were categorizedas low trust (F(9, 1317) = 8.13, p < .001) or high trust (F(9,

1467) = 23.77, p < .001). That is, independent of levels of trust,the results demonstrate successful EC effects.

According to univariate tests, participants scoring low on socialtrust exhibited a significant paired condition � evaluative mea-surement interaction for conventional (F(3, 436) = 7.47, p < .001),enzyme (F(3, 436) = 5.84, p = .001), and genetic (F(3, 436) = 11.02,p < .001) technology.

24 N. Loebnitz, K.G. Grunert / Food Quality and Preference 37 (2014) 19–26

The simple effects analysis of this interaction revealed a signif-icant difference only when the conventional technology was pairedwith negative images (G2), F(1, 436) = 22.124, p < .001;MEM1 = 4.66; SD = .15; MEM2 = 4.21; SD = .14). No significant differ-ences arose for conventional technology paired with neutralimages (G1), F(1, 436) = .178, p = .673, ns; MEM1 = 4.84; SD = .16;MEM2 = 4.79; SD = .15) or positive images (G3), F(1, 436) = 1.415,p = .235; MEM1 = 4.95; SD = .16; MEM2 = 5.08; SD = .16).

Furthermore, we found significant differences only whenenzyme technology was paired with negative images (G1), F(1,436) = 25.724, p < .001; MEM1 = 4.90; SD = .13; MEM2 = 4.39;SD = .14, not positive (G2), F(1, 436) = .168, p = .682, ns;MEM1 = 4.77; SD = .13; MEM2 = 4.73, SD = .13, or neutral (G3), F(1,436) = .291, p = .59; MEM1 = 4.57; SD = .14; MEM2 = 4.52, SD = .15.

For genetic technology, similar results emerged: We observedsignificant differences only when it paired with negative images(G3), F(1, 436) = 42.69, p < .001; MEM1 = 2.89; SD = .17; MEM2 =2.07; SD = .16, not positive (G1), F(1, 436) = .302, p = .583, ns;MEM1 = 3.00; SD = .16; MEM2 = 3.07, SD = .15, or neutral (G2), F(1,436) = .002, p = .962; MEM1 = 2.83; SD = .15; MEM2 = 2.83, SD = .15.

Among participants categorized as high trust, we found signifi-cant paired condition � evaluative measurement interactions forconventional (F(3, 489) = 33.37, p < .001), enzyme (F(3, 489) =24.1, p < .001), and genetic (F(3, 489) = 14.92, p < .001) technology.

The simple effects analysis revealed a significant difference whenconventional technology paired with positive (G3), F(1, 489) =16.12, p < .001; MEM1 = 5.25; SD = .13; MEM2 = 5.64, SD = .13, andnegative images (G2), F(1, 489) = 84.86, p < .001; MEM1 = 5.46;SD = .14; MEM2 = 4.48, SD = .14, not neutral images (G1), F(1,489) = .615, p = .433, ns; MEM1 = 5.21; SD = .13; MEM2 = 5.28,SD = .13.

Similar results arose for enzyme technology: We observed sig-nificant differences when enzyme technology was paired withpositive images (G2), F(1, 489) = 11.11, p = .001; MEM1 = 5.60;SD = .12; MEM2 = 5.98, SD = .12, or with negative images (G1), F(1,489) = 67.52, p < .001; MEM1 = 5.31; SD = SD = .11; MEM2 = 4.44,SD = .11. No significant differences were observed when enzymetechnology was paired with neutral images (G3), F(1, 489) = .73,p = .393, ns; MEM1 = 5.63; SD = .11; MEM2 = 5.54, SD = .11.

Finally, for genetic technology, we noted significant differenceswhen it was paired with negative images (G3), F(1, 489) = 24.33,p < .001; MEM1 = 4.32; SD = .16; MEM2 = 3.69, SD = .16, or positiveimages (G1), F(1, 489) = 20.28, p < .001; MEM1 = 4.15; SD = .17;MEM2 = 4.74, SD = .17, but not neutral images (G2), F(1, 489) = .173,p = .677, ns; MEM1 = 4.16; SD = .18; MEM2 = 4.21, SD = .18.

These results suggest that participants with a high level of trustin institutions that promote and regulate these technologies can beconditioned both positively and negatively, independent of thetype of food technology. Whereas participants with a low level oftrust in these institutions can be conditioned too, this effect onlyholds with increasing negative valence for the technology. Whensocial trust is low, positive conditioning of food technologies isnot demonstrated in this study.

Awareness data

As expected, we observed that in the unpaired condition (BSBgroup), participants either did not know or indicated that CS werenot followed by a particular picture. In addition, the majority ofthese participants responded that they did recognize the valenceof the picture paired with the CS, could not select a picture fromthe list, and were very uncertain with regard to their decision.

On the basis of the contingency awareness questions, we calcu-lated a total contingency awareness score for each participant. Aparticipant who correctly noticed that CS was followed by a partic-ular image, noted the correct valence, and identified the correct

paired images received a score of 2; one who stated that CS wasfollowed by particular image and noted the correct valence butcould not identify the particular picture earned a score of 1; anda participant who answered all questions incorrectly received ascore of 0. Across the three CS, the total score therefore rangedfrom 0 to 6. To examine the role of contingency awareness(Baeyens et al., 2005; Pleyers et al., 2007), we categorized partici-pants into one of the three levels of awareness on the basis of theirscores: 0–2 = level 1 or no awareness, 3–4, = level 2 or moderatelyaware, and 5–6 = level 3 or highly aware.

Discarding the unpaired condition (BSB group), 195 participantsshowed no awareness (28%), 296 were moderately aware (42.5%),and 206 (29.6%) were highly aware. In separate analyses for thesethree groups, we investigated the role of contingency awareness inconditioning effects, such that we ran three 3 (paired condition:G1, G2, G3) � 2 (evaluative measurement: EM1, EM2) MANOVAs,with repeated measures on the last variable. The results revealedan evaluative measurement � paired group interaction for all threegroups: unaware participants at level 1 (F(6, 382) = 11.35, p < .001,g2 = .15), moderately aware participants in level 2 (F(6,584) = 14.43, p < .001, g2 = .12), and highly aware participants, orlevel 3 (F(6, 404) = 9.93, p < .001, g2 = .12). That is, EC effects occurfor all three groups, even among participants categorized as una-ware (level 1). Furthermore, higher awareness appears to attenuateEC effects slightly.

Discussion

This study has focused on the formation of Chinese consumers’attitudes toward different food technologies. In particular, we con-sider whether evaluative conditioning can explain attitude forma-tion in response to affective images. We also have investigated themoderating effect of social trust in the context of consumer atti-tudes toward food technologies. Consistent with studies conductedin the West, when the conditioned stimulus (CS), food technology,pairs with a positive (negative) unconditioned stimuli, it is evalu-ated more positively (negatively) by Chinese consumers. Theseresults strengthen the case for EC as responsible for attitude forma-tion. This is not to suggest that EC effects are equally strong for allparticipants though. Rather, we reveal that EC effects on consum-ers’ evaluations of food technologies differ at various levels ofsocial trust; that is, the EC effect is subject to moderation. If a con-sumer exhibits high social trust in institutions that promote andregulate these technologies, EC significantly modifies his or herattitudes toward food technologies, regardless of the US valence.A different story emerges for a consumer with low social trust inthese institutions; in this case, EC significantly alters attitudestoward food technology only if the respective food technology ispaired with a negative US. This study thus contributes to EC liter-ature by identifying another moderator of EC.

Although speculative, it appears that the attitudes of partici-pants with low social trust toward food technologies were rein-forced by a negative pairing. Further research is necessary toprovide more direct evidence of this negative reinforcement. Thenotion that social trust, which is a main determinant of technologyacceptance, also affects EC strength suggests that food marketersmust examine intervention strategies to improve consumers’ trustin the institutions that promote and regulate food technologies.

Theoretically, these results demonstrate that the CS–US contin-gency is not a causal factor for attitude formation toward foodtechnologies. An EC effect emerges even among participants whoremained unaware of the contingency. Although in line with previ-ous studies that demonstrate EC effects for participants who arenot aware of the CS–US contingency (Walther & Nagengast,2006), our results run contrary to Loebnitz and Grunert’s (2013)

N. Loebnitz, K.G. Grunert / Food Quality and Preference 37 (2014) 19–26 25

finding that EC effects are significantly stronger for participantsclassified as aware, which would support the prominent proposi-tional account as the mechanism underlying EC. Our results con-test this account, which depends on contingency awareness. AsHofmann et al. (2010) argue, EC studies apparently should be moresensitive when classifying participants as aware versus unaware,such as by using an intra-individual pairing method (Pleyerset al., 2007). To address these contrasting findings, additional stud-ies might manipulate contingency awareness experimentally,rather than relying on explicit measures.

To the best of our knowledge, this study is the first to demon-strate attitude formation toward food technologies, employing anEC procedure, in China. A dearth of EC studies exists outside Wes-tern and developed countries. Furthermore, our study identifiessocial trust as a moderating influence of attitudes, prior to the ECprocedure. These findings accordingly suggest a promising avenuefor research into the key mechanisms that underlie attitude forma-tion toward food technologies. They also provide insights for prac-titioners into which consumers are more subject to EC, which canlead to more effective interventions.

Acknowledgements

The research reported in this article is part of the project ‘‘LEAN-GREENFOOD’’, funded by the European Commission through con-tract no. 238084. We gratefully acknowledge Elisabeth NevinsCaswell for her comments on an earlier version of this article.

References

Baeyens, F., Diaz, E., & Ruiz, G. (2005). Resistance to extinction of human evaluativeconditioning using a between-subjects design. Cognition & Emotion, 19(2),245–268.

Baeyens, F., Eelen, P., Crombez, G., & Van den Bergh, O. (1992). Human evaluativeconditioning: Acquisition trials, presentation schedule, evaluative style andcontingency awareness. Behaviour Research and Therapy, 30(2), 133–142.

Beckmann, S. C., Brokmose, S., & Lind, R. L. (2001). Danske forbrugere og økologiskefødevarer. Copenhagen: Copenhagen Business School.

Bierley, C., McSweeney, F. K., & Vannieuwkerk, R. (1985). Classical conditioning ofpreferences for stimuli. Journal of Consumer Research, 316–323.

Blair, M. E., & Shimp, T. A. (1992). Consequences of an unpleasant experience withmusic: A second-order negative conditioning perspective. Journal of Advertising,35–43.

Bredahl, L. (2001). Determinants of consumer attitudes and purchase intentionswith regard to genetically modified food: Results of a cross-national survey.Journal of Consumer Policy, 24(1), 23–61.

Cacioppo, J. T., Marshall-Goodell, B. S., Tassinary, L. G., & Petty, R. E. (1992).Rudimentary determinants of attitudes: Classical conditioning is more effectivewhen prior knowledge about the attitude stimulus is low than high. Journal ofExperimental Social Psychology, 28(3), 207–233.

Curtis, K. R., McCluskey, J. J., & Wahl, T. I. (2004). Consumer acceptance ofgenetically modified food products in the developing world.

Dasgupta, N., & Greenwald, A. G. (2001). On the malleability of automatic attitudes:Combating automatic prejudice with images of admired and dislikedindividuals. Journal of Personality and Social Psychology, 81(5), 800–814.

De Houwer, H. (2007). A conceptual and theoretical analysis of evaluativeconditioning. Spanish Journal of Psychology, 10(2), 230–241.

De Houwer, J., Thomas, S., & Baeyens, F. (2001). Association learning of likes anddislikes: A review of 25 years of research on human evaluative conditioning.Psychological Bulletin, 127(6), 853–869.

Eiser, J. R., Miles, S., & Frewer, L. J. (2002). Trust, perceived risk, and attitudes towardfood technologies. Journal of Applied Social Psychology, 32(11), 2423–2433.

Field, A. P. (1996). An appropriate control condition for evaluative conditioning. Schoolof Cognitive & Computing Sciences, University of Sussex.

Field, A. P., & Davey, G. C. (1999). Reevaluating evaluative conditioning: Anonassociative explanation of conditioning effects in the visual evaluativeconditioning paradigm. Journal of Experimental Psychology: Animal BehaviorProcesses, 25(2), 211–224.

Frewer, L. (1999). Risk perception, social trust, and public participation in strategicdecision making: Implications for emerging technologies. Ambio, 569–574.

Frewer, L. J., Howard, C., Hedderley, D., & Shepherd, R. (1997). Consumer attitudestowards different food-processing technologies used in cheese production—Theinfluence of consumer benefit. Food Quality and Preference, 8(4), 271–280.

Frewer, L. J., Scholderer, J., & Bredahl, L. (2003). Communicating about the risks andbenefits of genetically modified foods: The mediating role of trust. Risk Analysis,23(6), 1117–1133.

Gaskell, G., Allum, N., & Stares, S. (2003). Europeans and biotechnology in 2002:Eurobarometer 58.0. Brussels: European Commission.

Gaskell, G., Bauer, M. W., Durant, J., & Allum, N. C. (1999). Worlds apart? Thereception of genetically modified foods in Europe and the US. Science,285(5426), 384–387.

Gibson, B. (2008). Can evaluative conditioning change attitudes toward maturebrands? New evidence from the Implicit Association Test. Journal of ConsumerResearch, 35(1), 178–188.

Grunert, K. G., Bech-Larsen, T., Lähteenmäki, L., Ueland, Ø., & Aström, A. (2004).Attitudes towards the use of GMOs in food production and their impact onbuying intention: The role of positive sensory experience. Agribusiness, 20(1),95–107.

Grunert, K. G., Bredahl, L., & Scholderer, J. (2003). Four questions on Europeanconsumers’ attitudes toward the use of genetic modification in food production.Innovative Food Science & Emerging Technologies, 4(4), 435–445.

Grunert, K. G., Lähteenmäki, L., Asger Nielsen, N., Poulsen, J. B., Ueland, O., & Aström,A. (2001). Consumer perceptions of food products involving geneticmodification—Results from a qualitative study in four Nordic countries. FoodQuality and Preference, 12(8), 527–542.

Hamstra, A. M., & Smink, C. (1996). Consumers and biotechnology in theNetherlands. British Food Journal, 98(4/5), 34–38.

Hermans, D., Baeyens, F., & Eelen, P. (2003). On the acquisition and activation ofevaluative information in memory: The study of evaluative learning andaffective priming combined. In J. Musch & K. C. Klauer (Eds.), The psychology ofevaluation: Affective processes in cognition and emotion (pp. 139–168). Mahwah,NJ: Erlbaum.

Hermans, D., Baeyens, F., Lamote, S., Spruyt, A., & Eelen, P. (2005). Affective primingas an indirect measure of food preferences acquired through odor conditioning.Experimental Psychology, 52(3), 180–186.

Hofmann, W., De Houwer, J., Perugini, M., Baeyens, F., & Crombez, G. (2010).Evaluative conditioning in humans: A meta-analysis. Psychological Bulletin,136(3), 390–421.

Hollands, G. J., Prestwich, A., & Marteau, T. M. (2011). Using aversive images toenhance healthy food choices and implicit attitudes: An experimental test ofevaluative conditioning. Health Psychology, 30(2), 195–203.

Kerkhof, I., Vansteenwegen, D., Baeyens, F., & Hermans, D. (2009). A picture-flavourparadigm for studying complex conditioning processes in food preferencelearning. Appetite, 53(3), 303–308.

Lähteenmäki, L., Grunert, K., Ueland, Ø., Aström, A., Arvola, A., & Bech-Larsen, T.(2002). Acceptability of genetically modified cheese presented as real productalternative. Food Quality and Preference, 13(7), 523–533.

Lin, W., Somwaru, A., Tuan, F., Huang, J., & Bai, J. (2006). Consumer attitudes towardbiotech foods in China. Journal of International Food & Agribusiness Marketing,18(1–2), 177–203.

Loebnitz, N., & Grunert, K.G. (2013). Evaluative conditioning of food technologies.In 4th regional EMAC CEE conference, St. Petersburg, Russia, September 26–27.

Olsen, N. V., Grunert, K. G., & Sonne, A.-M. (2010). Consumer acceptance of high-pressure processing and pulsed-electric field: A review. Trends in Food Science &Technology, 21(9), 464–472.

Olsen, N. V., Menichelli, E., Grunert, K. G., Sonne, A.-M., Szabo, E., Banati, D. N., et al.(2011). Choice probability for apple juice based on novel processing techniques:Investigating the choice relevance of mean-end-chains. Food Quality andPreference, 22(1), 48–59.

Olson, M. A., & Fazio, R. H. (2001). Implicit attitude formation through classicalconditioning. Psychological Science, 12(5), 413–417.

Pleyers, G., Corneille, O., Luminet, O., & Yzerbyt, V. (2007). Aware and (dis)liking:Item-based analyses reveal that valence acquisition via evaluative conditioningemerges only when there is contingency awareness. Journal of ExperimentalPsychology: Learning, Memory and Cognition, 33(1), 130–144.

Poppe, C., & Kjærnes, U. (2003). Trust in food in Europe. A comparative analysis. Oslo:National Institute for Consumer Research SIFO.

Qiu, H., Huang, J., Pray, C., & Rozelle, S. (2012). Consumers’ trust in government andtheir attitudes towards genetically modified food: Empirical evidence fromChina. Journal of Chinese Economic and Business Studies, 10(1), 67–87.

Schemer, C., Matthes, J. R., Wirth, W., & Textor, S. (2008). Does ‘‘passing theCourvoisier’’ always pay off? Positive and negative evaluative conditioningeffects of brand placements in music videos. Psychology & Marketing, 25(10),923–943.

Scholderer, J., Grunert, K. G., & Søndergaard, H. A. (2006). Learning to love what wedespise: Experiential re-evaluation of stigmatised technologies. In 35th EMACconference: Sustainable marketing leadership.

Scholderer, J., & Frewer, L. J. (2003). The biotechnology communication paradox:Experimental evidence and the need for a new strategy. Journal of ConsumerPolicy, 26(2), 125–157.

Siegrist, M. (2000). The influence of trust and perceptions of risks and benefits onthe acceptance of gene technology. Risk Analysis, 20(2), 195–204.

Søndergaard, H. A., Grunert, K. G., & Scholderer, J. (2005). Consumer attitudes toenzymes in food production. Trends in Food Science & Technology, 16(10),466–474.

Stuart, E. W., Shimp, T. A., & Engle, R. W. (1987). Classical conditioning of consumerattitudes: Four experiments in an advertising context. Journal of ConsumerResearch, 14(3), 334–349.

Sweldens, S., Van Osselaer, S. M., & Janiszewski, C. (2012). Evaluative conditioningprocedures and the resilience of conditioned brand attitudes. Journal ofConsumer Research, 37(3), 473–489.

26 N. Loebnitz, K.G. Grunert / Food Quality and Preference 37 (2014) 19–26

Teisl, M. F., Fein, S. B., & Levy, A. S. (2009). Information effects on consumer attitudestoward three food technologies: Organic production, biotechnology, andirradiation. Food Quality and Preference, 20(8), 586–596.

Verhulst, F., Hermans, D., Baeyens, F., Spruyt, A., & Eelen, P. (2006). Determinantsand predictive validity of direct and indirect measures of recently acquired foodattitudes. Appetite, 46(2), 137–143.

Walther, E., & Nagengast, B. (2006). Evaluative conditioning and the awarenessissue: Assessing contingency awareness with the four-picture recognition test.Journal of Experimental Psychology: Animal Behavior Processes, 32(4), 454.

Walther, E., Weil, R., & Langer, T. (2011). Why do we like the iPhone? The role ofevaluative conditioning in attitude formation. Social and Personality PsychologyCompass, 5(7), 473–486.

Zellner, D. A., Rozin, P., Aron, M., & Kulish, C. (1983). Conditioned enhancement ofhuman’s liking for flavor by pairing with sweetness. Learning and Motivation,14(3), 338–350.

Zhang, X., Huang, J., Qiu, H., & Huang, Z. (2010). A consumer segmentation studywith regards to genetically modified food in urban China. Food Policy, 35(5),456–462.