Hedonic ratings and perceived healthiness in experimental functional food choices

ARTICLE IN PRESS

Appetite 47 (2006) 302–314 www.elsevier.com/locate/appet

Research report

Hedonic ratings and perceived healthiness in experimental functional food choices Nina Urala, Liisa La¨hteenma¨ki VTT Technical Research Centre of Finland, P.O. Box 1000, FIN-02044 VTT, Finland Received 20 October 2005; received in revised form 25 March 2006; accepted 21 April 2006

Abstract The associations of liking and perceived healthiness ratings between repeated food choices were studied in two experiments. Participants’ snack bar (n ¼ 41, Experiment I) and beverage (n ¼ 60, Experiment II) choices among six product alternatives were monitored for 4 and 3 weeks, respectively. In Experiment I, participants were allowed to familiarise themselves with snack bar alternatives (‘‘familiar assortment’’) prior to making choices. In Experiment II, the participants started making their beverage choices without familiarising themselves (‘‘unfamiliar assortment’’). In both experiments, the participants were divided into three groups according to their choice behaviour for each alternative: non-interested (0 choices), experimenters (1 choice) and potential frequent users (2 or more choices). In Experiment I, the overall difference between non-interested and potential frequent users of a product was 1.3 points in expected liking and 2.6 points in actual liking on a 7-point scale (ANOVA, po0:001). In Experiment II, the overall differences in blind hedonic ratings between non-interested participants and potential frequent users of a product were within a range of 0.9 points ðpo0:001Þ. The difference was wider for expected liking ratings, 1.3 points ðpo0:001Þ. Neither the perceived healthiness of the samples nor the background attitudes could be consistently associated with the choices (Pearson’s correlation coefficient). r 2006 Elsevier Ltd. All rights reserved. Keywords: Choice; Consumer; Liking; Healthiness; Ratings; Snack bar; Beverage

Introduction When purchasing food, we make a considerable number of minor choices every day. We process the alternatives through our individual attitudes, perceptions, use context and available resources. The hedonic properties derived from tasting a food are certainly one of the most essential factors in repetitive food choices, and hedonic ratings (Peryam & Pilgrim, 1957) have been widely used in predicting the future success of novel food products, but there is also evidence that the attitudes contribute to real food choices (Roininen & Tuorila, 1999). So, can the hedonic measurements correspond to the actual behaviour (de Graaf et al., 2005; Ko¨ster, 2003; Ko¨ster et al., 2002; Rosas-Nexticapa et al. 2005)? How can we choose food Corresponding author. VTT Technical Research Centre Of Finland, Consumer Studies, P.O. Box 1000, FIN-02044 VTT, Finland. Tel.: +358 20 722 5216; fax: +358 20 722 7071. E-mail address: nina.urala@vtt.fi (N. Urala).

0195-6663/$ - see front matter r 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.appet.2006.04.007

products based on hedonic liking when purchasing a new, ‘‘untasted’’ product? This dilemma has lately instigated discussions on the usefulness of the hedonic ratings in predicting long-term choices. The criticism has mainly focused on the hedonic tests that are performed in non-realistic laboratory settings (Meiselman, 1992) and on using methods that are not relevant to consumers (Lucas & Bellisle, 1987; Zandstra et al., 1999). Single liking measurements do not necessarily take into account the inconsistency of individual ratings (Levy & Ko¨ster, 1999; La¨hteenma¨ki & Tuorila, 1995), the environmental aspects (de Graaf et al., 2005), the context and situation (King et al., in press) and consumption aspects after the product is purchased (Grunert, 2003; Kozlowska et al., 2003). Most importantly, the role of such product attributes, which are not derived from direct experience of a food (Nelson, 1970), needs more investigation. In addition to direct experience of the food (i.e., tasting), the expectations have found to have an impact on

ARTICLE IN PRESS N. Urala, L. La¨hteenma¨ki / Appetite 47 (2006) 302–314

pleasantness and acceptance of foods (Cardello & Sawyer, 1992; Deliza & MacFie, 1996; Hurling & Shepherd, 2003; Jaeger & MacFie, 2001; Tuorila et al., 1998). Expectations thus have an impact not only on choosing the food but also on consumer liking and success on the market. Cardello and Sawyer (1992) showed that positive expectations increased the liking ratings compared to the ratings based on neutral expectations. In turn, negative expectations decreased the pleasantness ratings compared to the ratings based on neutral expectations (assimilation model). The actual perception can also be opposite to the expectations (contrast effect), but it seems rarer than the assimilation model (Cardello, 2003; Cardello & Sawyer, 1992; Deliza & MacFie, 1996; Hurling & Shepherd, 2003). Consumers’ expectations are typically guided by marketing communication. One interesting type of foods is functional foods, because their marketing is strongly based on health-related information attached to a single product (Diplock et al., 1999), but from consumers’ point of view the healthiness seems to be only one choice factor among price, pleasantness, convenience and familiarity (Urala & La¨hteenma¨ki, 2003). Earlier studies have shown that health-related information influences not only the healthiness perception but also the liking ratings of foods (Kihlberg et al. 2005; Mialon et al., 2002; Westcombe & Wardle, 1997), the effect of health-related information seems to be product dependent (Ka¨hko¨nen et al., 1997) and the effect may depend on the degree of liking (Kihlberg et al. 2005). Yet, Stein et al. (2003) found that healthrelated information did not affect the hedonic ratings of beverages. The effect on choice-behaviour was, however, clear: participants who were given information on the health effects of an unfamiliar beverage chose more bottles as a reward after the experiment compared to the group who did not receive the information. Similarly to other food products, sensory properties and hedonic pleasure are important in functional products, but we still know little about their relative importance in relation to health claims. In Belgium, neither most of women nor men have been willing to use functional foods that were stated to have a worse taste than their conventional counterparts (Verbeke, 2005, 2006). Studies including actual tasting of functional foods indicate that excellent hedonic properties in functional foods are necessary and, for instance, a bitter taste in functional foods does not support the perceived healthiness of such foods (Tuorila & Cardello, 2002). Huotilainen et al. (2006) showed that the liking ratings of new functional drinks were the strongest predictors of the preferred use frequencies and Bower et al. (2003) reported that Scottish participants’ purchase intent of spreads labelled with a proven health benefit was strongly related to the degree of liking, in addition to the health label. The expectations, actual liking, healthiness perception and consumers’ attitudes form a complex entity. In the current study, we try to understand better how the perceived healthiness, expected and actual liking ratings

303

and consumers’ attitudes are related to the repetitive food choices. The research questions of our study were as follows. (1) How are the perceived healthiness, expected and actual liking ratings related to the choices of food alternatives that have been tasted several times before the choice period (‘‘familiar assortment’’) compared to the more realistic situation where the available alternatives have not been tasted before (‘‘unfamiliar assortment’’)? We assume that the actual liking correlates strongly with the choices made freely from the familiar assortment and that the expectations correlate with the choices made from the unfamiliar assortment. (2) How do background attitudes influence the repetitive choices of so-called functional products (labelled with health-related claims), organic products (have positive health image among consumers) and their conventional alternatives? We assume to find associations between background attitudes and choices. We expect that positive attitudes towards healthy eating and functional foods would positively correlate with the choices of samples labelled with a health-related claim and that interest towards natural products would positively correlate with the choices of samples labelled as organic. Methods Participants Two panels of Finnish volunteers were recruited separately for two choice experiments. Participants were told that the aim of the study was to examine the liking for different types of snack bars (Experiment I) or beverages (Experiment II). Participants were not working in food or health-related areas. For both experiments, all participants were recruited from work places using the convenience sampling method and only users of snack bars (Experiment I) or beverages (Experiment II) were recruited in both experiments. There were more women than men and the participants were relatively well educated (Table 1). The majority reported that they bought at least half of their household food supplies and, in general, they reported their own health as being good (Table 1). All participants in both experiments were very motivated: no one withdrew from the study. Participants were debriefed and given a small gift after they completed the study. For Experiment I, 41 participants were recruited from VTT (Technical Research Centre of Finland). Motivation to use products with health claims was inquired by asking participants’ need to pay attention to health issues linked to the products. In all, 66% of the participants rated at least 5 on the 7-point scale (1, no need at all; 7, very strong need) their need to pay attention to the well-being of their gut, 64% to their physical condition, 42% to fibre intake and 44% to physical recovery after exercise. Respondents reported being familiar with snack bars, in general, as

ARTICLE IN PRESS 304

N. Urala, L. La¨hteenma¨ki / Appetite 47 (2006) 302–314

Table 1 Descriptions of the participants in Experiments I and II Experiment I

Experiment II

Mean age (years) Age range (years) Total Women Men

37 24–59 41 33 8

80% 20%

35 24–60 60 42 18

70% 30%

University degree Reported very good state of own health Bought at least 50% of household’s food supplies

18 29 37

44% 70% 90%

19 45 58

32% 75% 97%

Mean

SD

Mean

SD

GHI NPI

4.3 3.5

1.3 1.1

4.5 ND

1.2 ND

FF REW FF NEC FF CON

3.9 3.3 4.6

1.0 0.7 1.0

3.8 3.5 4.3

1.0 0.7 1.0

GHI, General Health Interest; NPI, Natural Product Interest; (Roininen et al. 1999), FF REW, perceived reward from using functional foods; FF NEC, necessity for functional foods; FF CON; confidence in functional foods (Urala & La¨hteenma¨ki, in press); ND, not measured.

expected according to the recruiting criteria. When asked about their experience of different types of snack bars before the study, an ordinary chocolate bar and white chocolate had been tasted by all of the respondents. The recovery bar and organic snack bar were, in turn, the least familiar products: 24% and 22% of the participants reported that they had tasted these products, respectively. For Experiment II, 60 participants were recruited from two large office buildings in Espoo, Finland. Motivation to use products with health claims was inquired like in Experiment I, but with health issues targeted to the beverage alternatives. Most of the participants reported the need to pay attention (at least 5 on the 7-point scale) to the well-being of their gut (60% of the respondents), to fibre intake (57%) and to balance blood glucose levels (35%). When asked about participants’ experience of different types of juices before the study, pure fruit juices had been tasted by all the respondents. Juice with added fibre had been tried by 63% and juice with a blood glucose balancing effect had been tasted by 57% of the respondents.

The respondents’ attitudes towards functional foods were measured using a short version of a functional food questionnaire (Urala & La¨hteenma¨ki, 2004; Urala & La¨hteenma¨ki, in press). Functional food was defined in the background questionnaire as a food product that, in addition to its usual nutritional effect, has a special effect that maintains or promotes health or decreases the risk of disease. Attitude statements of rewarding feeling from using functional foods (FF REW), necessity for functional foods (FF NEC), confidence in functional foods (FF CON) and safety of functional foods (FF SAF) were included into the questionnaire (Urala & La¨hteenma¨ki, in press). In both experiments, the respondents were relatively neutral in their attitudes towards functional foods (Table 1). The reliability (Cronbach’s alpha) of FF SAF was poor in both studies and this attitude measurement was not included in further analysis. All attitude statements were evaluated on the 7-point Likert-scale: 1, completely disagree and 7, completely agree.

Background attitudes

Six product alternatives were used in both selections. Commercial snack bars were used as the samples in Experiment 1. The original packages were removed and samples were wrapped in transparent plastic. Beverage samples were, in turn, bottled in semi-transparent plastic bottles (100 ml). In both studies, four samples were labelled with different types of health-related information and two samples had no health-related information. This approach allowed us to seek the relationship between background attitudes and the choices of products with different types of health-related claims.

In both experiments, participants’ attitudes towards healthy eating were measured by the General Health Interest (GHI) scale developed by Roininen et al. (1999). In Experiment I, the respondents’ interest in natural products was measured by the Natural Product Interest scale (Roininen et al., 1999) as organic alternatives were included in the selection. The participants were reasonably neutral in their health interest. Natural Product Interest in Experiment I was also neutral (Table 1).

Products and their labels

ARTICLE IN PRESS N. Urala, L. La¨hteenma¨ki / Appetite 47 (2006) 302–314

305

Table 2 Descriptions and total choices of the snack bar samples of Experiment I Code

Choices (N)

Choices (%)

FF-cereal FF-choc ORG-cereal ORG-choc CONV-cereal CONV-choc

101 34 12 55 200 131

19 6 2 10 38 25

Total

533

100

Claim Oat fibre improves gut well-being. This product contains plenty of oat fibre. Aids in physical recovering after exercise. This snack bar contains plenty of protein. Organic Organic — —

FF, labelled with health-related claim; ORG, labelled as organic; CONV, no health-related claim; choc, covered with chocolate; cereal, cereal bar. Codes are only used for indicating the samples in this paper and they were not shown to the participants.

Table 3 Descriptions and total choices of the beverage samples of Experiment II Code

Choices (n)

Choices (%)

FF-well5

127

14

FF-well4

108

12

FF-gi5

64

7

FF-gi4

136

15

CONV-3 CONV-2

288 177

32 20

Total

900

100

Claim

Price (VTT Coins)

Fibres promote gut well-being. This product contains fibre. SHAKE BEFORE USE Fibres promote gut well-being. This product contains fibre. SHAKE BEFORE USE Slow carbohydrates help to balance the blood glucose levels. This product contains slow carbohydrates (low glycaemic index). Slow carbohydrates help to balance the blood glucose levels. This product contains slow carbohydrates (low glycaemic index). — —

5 4 5

4

3 2

FF-well, labelled with effect on gut well-being; GI-gi, labelled with blood glucose level balancing effect. Numbers 2–5 refer to price in VTT Coins.

In Experiment I, there were two types of health-related claims stated on the snack bar packages: two samples were labelled as an organic product by adding the word ‘‘organic’’ (‘‘luomu’’ in Finnish) on the package (ORG in this paper) and two as a functional product by adding a targeted health-related claim on the package (FF in this paper). Two had no health-related information and they were considered as conventional alternatives (CONV in this paper) (Table 2). The organic products represented foods that have a positive health image among consumers without scientific proof (Zanoli & Naspetti, 2002) and the functional products represented healthiness that is connected to a single product (Diplock et al., 1999) (Table 2). In Experiment II, we concentrated on functional foods and the conventional products were used for filling the choices. Two types of health-related claims were provided on the beverage bottles: two beverages were labelled as functional by a fibre-related health claim referring to the positive effect on gut well-being (FF-well in this paper) and two beverages also as functional with a carbohydraterelated claim referring to the positive effects on blood glucose levels, i.e., low glycaemic index (FF-gi in this

paper). Again, two products had no health-related information representing conventional alternatives (CONV in this paper) (Table 3). Product-related evaluations In both experiments, four product-related evaluations were carried out at different stages of the experiments: pleasantness, familiarity, perceived healthiness and willingness to use. These were measured on a 7-point scale: 1, not at all (pleasant, familiar, healthy or willing) and 7, extremely (pleasant, familiar, healthy or willing). Samples were evaluated in a randomised order. Data analysis For examining the relationships between the ratings and the choices, both the ratings and the choices were converted into so-called relative shares (or ratings). This was done for each rating and participant separately by summing, for instance, the hedonic ratings for all six alternatives and dividing the single hedonic rating given for each alternative with this sum. For instance, if a participant

ARTICLE IN PRESS N. Urala, L. La¨hteenma¨ki / Appetite 47 (2006) 302–314

Procedure Experiment I The participants were allowed to familiarise themselves with the snack bars in a tasting session arranged in a meeting room at VTT and also in their homes during a weekend. Overall, the alternatives were evaluated at four stages of the experiment. The procedure is presented in Fig. 1. First, the participants took part in a tasting session where the expected and actual hedonic liking ratings for the alternatives were given. Based on the expectations, the participants were asked to choose one snack bar as a snack. The choice was marked on the questionnaire. After the expectation measurements, the participants were allowed to taste all the samples and evaluate the samples again. Directly after evaluations, participants were asked to make another choice and it was marked on the questionnaire. When leaving the tasting session, the participants were given one of each snack bar and a new questionnaire for home test. Participants were asked to taste the samples and make the evaluations at home during the weekend. Home tasting was carried out because it allowed the participants taste all six snack bars in their normal eating environment according to their own schedule.

EXPERIMENT I, SNACK BARS

EXPERIMENT II, BEVERAGES 1) BLIND EVALUATIONS

2) EXPECTATIONS

1) EXPECTATIONS + CHOICE

2) TASTING + CHOICE

3) HOME TASTING

NO FAMILIARISATION PERIOD

had given hedonic ratings of 4, 5, 2, 1, 2 and 5 for the six alternatives, respectively, the overall hedonic sum would have been 19. Thus, the relative hedonic rating for the first sample would have been 4/19 ð¼ 0:21Þ. For the relative shares describing the choices, the individual’s choice frequencies of each sample were divided by the sum of all choices made during the choice period (i.e., 10 for the snack bars and 15 for the beverages). Relative shares of ratings were used to adjust ratings to the range of scale used and number of choices made by individual participants. With this approach, we transformed absolute liking ratings into relative and thus comparable values across participants for the given selection of samples. These relative ratings were used throughout the data analysis for both experiments. Pearson’s correlation coefficient was used for investigating the linear associations between (a) liking ratings and the choice frequencies and (b) the background attitudes and the choice frequencies (Lawless & Heymann, 1998). In addition, to demonstrate how the liking and healthiness ratings differ between different consumer groups and how they are related to the choices, we divided the participants into three groups according to the choice frequencies (0, 1 and 2 or more choices) of each alternative. Groups were chosen to represent different choice behaviour per product: non-interested participants (did not choose the product at all), experimenters (chose the product only once) and potential frequent users (chose the product at least twice). The means of the hedonic and healthiness ratings (dependent) between these groups (independent) were tested with ANOVA.

FAMILIARISATION PERIOD

306

3) CHOICE PERIOD 4) CHOICE PERIOD

5) TASTING

4) BLIND EVALUATIONS

Fig. 1. Procedures and measurements of Experiments I and II. In Experiment I, the participants were allowed to familiarise themselves with the alternatives at three stages before the choice period (familiar assortment). In Experiment II, the participants were allowed to only look at the appearance of the samples before making their choices and no familiarisation period was included (unfamiliar assortment).

With this procedure, we ‘‘forced’’ the participants to familiarise themselves with the alternatives three times prior to choosing. In addition, it was possible to demonstrate how the ratings given at different stages of the familiarisation process (expected, tasted and home tasting) are related to the snack bar choices during the choice period. During the next 4 weeks, the respondents freely chose one snack bar from a counter approximately three times per week (choices 1–10). The counters were placed in the cafeterias of participants’ workplaces during coffee breaks and the participants were asked to consume the product during their break. The choices were made on alternate days to avoid forcing the participants to select snacks with a relatively high energy content (500–1100 kJ/bar) every day. Some exceptions in the participants’ schedules were made due to work or other excuses. The length of the choice period was based on the assumption that the participants could choose each sample at least once during the period, but there would be several opportunities to vary the choices. After the last choice, the snack bars were reevaluated in a tasting session. Experiment II The purpose of Experiment II was to demonstrate how the ratings and the background attitudes are related to the choices when it is not possible to taste the products or be familiar with the alternatives before purchasing. Thus, the

ARTICLE IN PRESS N. Urala, L. La¨hteenma¨ki / Appetite 47 (2006) 302–314

participants started making their choices without tasting the samples beforehand. The procedure is presented in Fig. 1. To imitate the choice situation in real life, the beverages were also given a price in ‘‘VTT Coins’’ and the participants were given 17 (on the first week) or 18 VTT Coins per week for purchasing the beverages. The samples with a health-related claim were priced higher than the samples without any specific claim (Table 3). The purpose for giving the coins was to lead the participants away from choosing only the most preferred alternatives, to support the realistic impression of purchasing the alternatives from a food store and to guide the participants to ponder their choices. To measure the liking, familiarity, perceived healthiness and willingness to use ratings of the samples, a blind test with 12 samples including six beverages belonging to the experiment was carried out 5 days before the choice period. As in Experiment I, the beverage choices were made from a counter in the participants’ workplace cafeteria. The choices were made on consecutive days during 3 weeks as the energy content of the beverages was reasonable (50–190 kJ/portion) for daily use. The choices were made at lunchtime and the respondents were asked to use the beverage during that day as a snack. The length of the choice period was based on the assumption that the participants could choose each sample at least twice during the period, but there would be several opportunities to vary the choices. To ensure that the participants read the labels before the first choice, they were asked to first look at the appearance of the bottles, read the label carefully and give the expectation ratings. They were not allowed to taste the alternatives. Results

307

For all the samples, Pearson’s correlation coefficients between the expected healthiness ratings and the ratings given in the first tasting, at home and at the end of the choice period, were r ¼ 0:74, 0.74 and 0.75 ðpo0:01Þ, respectively. Correlation coefficients among single samples varied between r ¼ 0:29 and r ¼ 0:80 ðpo0:01Þ for the expected and tasted healthiness, respectively, and between r ¼ 0:31 ðpo0:01Þ and r ¼ 0:76 ðpo0:01Þ for the expected healthiness and healthiness evaluated at home, respectively. The highest correlations were found in the perceived healthiness of the organic samples, indicating perhaps a more stable healthiness image of the organic than of other alternatives. The expected liking for the snack bars was rated at the middle of the scale, with the exception of ORG-cereal, which was rated slightly lower (Fig. 2c). After tasting, the liking ratings changed and differences between the liking ratings for the bars emerged. The liking for FF-choc decreased clearly and the liking for CONV-cereal increased after tasting. Samples of CONV-cereal, CONV-choc, FFcereal and ORG-choc were rated as the most pleasant and the FF-choc and ORG-cereal were the least liked products. In general, the means of liking ratings remained consistent after the first tasting, but similarly to the healthiness ratings there were differences between the measurements. Pearson’s correlation coefficients for all samples were r ¼ 0:47 ðpo0:01Þ between the expected and first tasted liking, r ¼ 0:82 ðpo0:01Þ between the first tasting and home evaluation, and r ¼ 0:78 ðpo0:01Þ between the first tasting and the last at the end of the study. When the samples were studied separately, the correlation coefficients between the first tasted liking and the expected liking varied between r ¼ 0:33 ðpo0:05Þ and r ¼ 0:67 ðpo0:01Þ. The reported willingness to use ratings correlated (Pearson’s) significantly with the reported liking ratings ðr ¼ 0:6720:92Þ throughout the experiment.

Experiment I Perceived familiarity, healthiness and liking As expected, the familiarity scores of the snack bars were less than 4 on the 7-point scale, indicating that the participants were not familiar with them. The familiarity increased during the choice period (Fig. 2a). However, at the end of the study, the samples were still regarded as neutral in their familiarity. The perceived healthiness of the snack bars remained stable during the study and followed very well the evaluations that were given according to the appearance and the labels of the samples (without tasting). Functional alternatives were not rated as being healthier than organic or conventional alternatives (Fig. 2b). Meanwhile, the products coated with seeds and cereal (FF-cereal, ORGcereal and CONV-cereal) were rated as being more healthy than the products coated with chocolate (FF-choc, ORGchoc and CONV-choc). When the products were tasted, the differences between the healthiness evaluations changed slightly; but after the first tasting, the means remained stable.

Snack bar choices In total, 533 snack bar choices were made during Experiment I (Table 2). According to the relative shares, the expected liking has guided the choice that was made directly after evaluating the expectations (no tasting) (Fig. 3). After tasting, three snack bars dominated the choices (CONV-cereal, CONV-choc and FF-cereal) and the shares of their choices were greater (85% of the first choices) than those of the hedonic scores (62% of the given liking ratings) (Fig. 3). Also, the participants avoided choosing the least preferred sample (ORG-cereal) as their first choice, while almost half of the participants chose the most preferred alternative (CONV-cereal). After home use, the participants did not choose the least preferred sample at all, but, in general, the overall situation did not change. During the choice period, the overall choices remained quite consistent, but the participants varied in their individual choices. The three most preferred snack bars still clearly dominated the choices. At the same time, the participants avoided choosing the least pleasant samples (ORG-cereal and FF-choc).

ARTICLE IN PRESS N. Urala, L. La¨hteenma¨ki / Appetite 47 (2006) 302–314

308

FAMILIARITY

6

7 FF-cereal FF-choc ORG-cereal ORG-choc CONV-cereal CONV-choc

PERCEIVED HEALTHINESS

7

5 4 3 2 1

(a)

6 5 4 3 2 1

Expected

Tasted

At home

At the end

Expected

(b)

Tasted

At home

At the end

7 6

LIKING

5 4 3 2 1 Expected

(c)

Tasted

At home

At the end

Fig. 2. Perceived (a) familiarity, (b) healthiness and (c) liking for six snack bars without tasting (expected), after tasting (tasted), at home and at the end of Experiment I (1, not at all pleasant, familiar, healthy or willing; 7, extremely pleasant, familiar, healthy or willing). FF, labelled with targeted healthrelated information; ORG, labelled as organic; CONV, no health-related information; choc, covered with chocolate; cereal, cereal bar.

CONV-cereal 100%

CONV-choc

FF-cereal

ORG-choc

FF-choc

ORG-cereal

80% 60% 40% 20%

en d he tt -A in g Li k

-A th om C e ho ic e -F irs C ho t ic es 1 -1 0

Li ki ng

-T as te d

te

-T as te d

C ho ic e

Li ki ng

ce ho i C

Li k

in g

-

-E

Ex

pe c

xp ec

te

d

d

0%

Fig. 3. Relative shares of hedonic ratings and of the snack bar choices measured at three stages before the choice period and at the end of Experiment I. FF, labelled with targeted health-related information; ORG, labelled as organic; CONV, no health-related information; choc, covered with chocolate; cereal, cereal bar.

ARTICLE IN PRESS N. Urala, L. La¨hteenma¨ki / Appetite 47 (2006) 302–314

309

Table 4 Pearson’s correlation coefficients between the hedonic rating given in different stages of the study, participants’ background attitudes and the relative share of snack bar choices (Experiment I) Relative hedonic ratings Expected Relative share of snack bar choices FF-cereal FF-choc ORG-cereal ORG-choc CONV-cereal CONV-choc

0.31*
0.38* 0.07 0.14 0.27 0.43**

Background attitudes

Tasted Beginning

At home

At the end

GHI

NPI

FF REW

FF NEC

FF CON

0.40**
0.06 0.58** 0.40 0.39* 0.44**

0.39*
0.04 0.46** 0.45** 0.40* 0.68**

0.54** 0.08 0.52** 0.61** 0.53** 0.74**

0.31* 0.02
0.02
0.04 0.02
0.12

0.10 0.05
0.10
0.09 0.05
0.11

0.21
0.32* 0.24 0.00
0.32* 0.05


0.35* 0.24
0.34* 0.00 0.24 0.16

0.13
0.27 0.27
0.10
0.27 0.01

FF, labelled with targeted health-related claim; ORG, labelled as organic; CONV, no health-related claim; choc, covered with chocolate; cereal, cereal bar. Codes are only used for indicating the samples in this paper and they were not shown to the participants. GHI, General Health Interest; NPI, Natural Product Interest; FF REW, perceived reward from using functional foods; FF NEC, necessity for functional foods; FF CON, confidence in functional foods. *po0.05; **po0.01; ***po0.001.

In general, the association between the background attitudes and the snack bar choices was weak (Table 4). Pearson’s correlation coefficients, indicating the association between the choices and the GHI, varied between r ¼
0:12 (not significant) and r ¼ 0:31 ðpo0:05Þ. Statistically significant correlation coefficients were not found between Natural Product Interest and the choices of organic snack bars (r ¼
0:10 and r ¼
0:09). Correlations between the perceived reward from using functional foods (FF REW) and the choices of snack bars with health-related claims were, in turn, r ¼
0:32 ðpo0:05Þ and r ¼ 0:21 (not significant), indicating both negative and positive attitude effects. The correlation between confidence in functional foods (FF CON) and functional choices varied between r ¼
0:35 ðpo0:05Þ and r ¼ 0:24 (Table 4). The correlations between the necessity for functional foods (FF NEC) and functional choices were not statistically significant. The clearest differences between the non-interested, experimenters and frequent users were found among the hedonic ratings given after the first tasting of the familiarisation period (ANOVA). In all snack bars, there was a difference between the hedonic ratings of the noninterested and the experimenters. The difference was statistically significant in four snack bars out of six (Tukey’s post hoc, po0:05). When all the samples were considered, the overall difference in hedonic ratings between the non-interested participants and frequent users was even as high as 2.6 points on the 7-point scale (ANOVA, po0:001) (Fig. 4). No differences were found in the healthiness ratings, with the exception of the CONVchoc sample (ANOVA, po0:01). Experiment II Perceived familiarity, healthiness, liking and willingness to use The beverage samples were perceived as being more familiar when they were tasted in the blind test than when

evaluated based on the appearance of the bottles (Fig. 5a). Samples, even those without a health-related claim, were perceived to be healthier when evaluated by looking at the bottles than in the blind test (Fig. 5b). In either situation, the functional alternatives were not perceived as being healthier than the conventional alternatives. There were no significant differences in the mean liking scores between blind testing and expectations. However, the correlation between tasted and expected liking ratings for all the samples was r ¼ 0:44 ðpo0:01Þ and varied between r ¼ 0:18 (not significant) and r ¼ 0:47 ðpo0:01Þ (Fig. 5c). The hedonic ratings given in the first blind test correlated with the blind evaluations at the end of the study (r ¼ 0:73; po0:01 for the samples that belonged to the choice experiment and r ¼ 0:66 po0:01 for the six extra samples). Correlation coefficients for blind tasted liking and willingness to use were relatively high, varying between r ¼ 0:71 ðpo0:01Þ and r ¼ 0:87 ðpo0:01Þ. However, participants were more willing to use the bottled beverages (that were not tasted) than the same samples in the blind tasting (Fig. 5d). Beverage choices In all, 900 beverage choices were made (Table 3). In the first choice, the share of the functional beverages with a gut well-being claim dominated the choices (FF-well5 and FFwell4). Their selection accounted for over 48% of the choices (Fig. 6). The share of their hedonic ratings was approximately 36%. During the week, the choices were balanced (due to the monetary constraint). The hedonic ratings given in the blind test were poorly related to the choices made during the experiment (Table 5). The correlation between the choices and the expected liking was higher, showing that the expectations were closely associated with the choices. The attitudes towards functional foods or the GHI did not have any statistically significant association with the beverage choices (Table 5).

ARTICLE IN PRESS N. Urala, L. La¨hteenma¨ki / Appetite 47 (2006) 302–314

310

Non-interested p=0.010

non sig.

Experimenters p=0.006

Frequent users

p=0.004

non sig.

Overall mean

p<0.001

p<0.001

7

6

b b

ACTUAL LIKING

b

b

c

b

b

ab

5

b 4

a

a

a 3

a

a a

2

1 n=12 8 21 FFcereal

24 9 8 FF-choc

35 3

3

ORGcereal

17 12 12

3 5 33

11 9 21

102 46 98

ORGchoc

CONVcereal

CONVchoc

All

Fig. 4. Hedonic ratings given by non-interested (0 choices), experimenters (1 choice) and potential frequent users (2 or more choices) of a particular snack bar (1, not at all pleasant, familiar, healthy or willing; 7, extremely pleasant, familiar, healthy or willing). FF, labelled with targeted health-related information; ORG, labelled as organic; CONV, no health-related information; choc, covered with chocolate, cereal, cereal bar. Different letters indicate the difference between the means according to Tukey’s post hoc test, significance level po0:05.

As in Experiment I, the participants were divided into three groups according to their potential choice behaviour. When the blind liking ratings were compared between these choice groups (ANOVA), the differences were reasonably small (Fig. 7a). When we compared the expected liking ratings, the differences were greater. Non-interested participants had rated that particular beverage lower than the participants who at least tried it. In total, the expected liking rating for the non-chosen alternatives was 1.3 points lower than the rating for the beverage chosen two or more times on the 7-point scale ðpo0:001Þ (Fig. 7b). The expected healthiness ratings differed significantly between the choice groups (Fig. 7c). In total, the difference between the choice groups was 1.3 on the 7-point scale ðpo0:001Þ. Discussion In several earlier studies (Arvola et al., 1999; La¨hteenma¨ki & van Trijp, 1995; Tuorila & Cardello, 2002), direct hedonic perception, i.e., tasting and liking of a food, strongly guided the food choices. Experiment I concurred with these results by showing that the actual liking ratings associated with the snack bar choices. As could be expected, the hedonic scores of non-interested participants were lower than those of experimenters. We were able to demonstrate how a wide difference in the actual hedonic scores between these options could predict the success (or failure) on the market, too. Kramer et al. (2001) reported less than 0.5 point differences on a 9-point scale

(1, dislike extremely; 9, like extremely) when comparing soldiers’ overall liking between food items chosen once and twice or more often. Unfortunately, no hedonic ratings were reported by Rosas-Nexticapa et al. (2005), where participants’ ðn ¼ 101Þ liking ratings of yoghurts were compared with the buying frequencies during 1 year. Our results need deeper investigation in future as this was one of the first attempts to understand the relationship between liking ratings and repetitive choices. In contrast, when the participants made their choices in a more realistic situation, without any direct experience of the hedonic properties of the alternatives and under a strict monetary constraint, the differences in actual hedonic ratings between non-interested, experimenters and potential frequent users were smaller. However, the differences in the expected liking ratings were wider than in blind liking ratings and statistically significant. This is in line with earlier studies that have clearly shown that the expectations have a central role in the novel product acceptance (Arvola et al., 1999; Lange et al., 1999; Tuorila et al., 1998). In our study, participants did not have a clear picture of the beverage selection before they started making choices and the expectations were the only cues for what to choose (in addition to the monetary constraint). In Experiment I, the first choices that were made without tasting were in line with the expected liking, but several most liked samples started dominating the choices after tasting. This supports the finding by Lange et al. (1999): participants ðn ¼ 123Þ wanted to be consistent with their

ARTICLE IN PRESS N. Urala, L. La¨hteenma¨ki / Appetite 47 (2006) 302–314

311

7

7 Blind

PERCEIVED HEALTHINESS

Expected

FAMILIARITY

6 5 4 3 2 1 FF-gi5

FF-gi4

3 2

7

7

6

6 WILLINGNESS TO USE

LIKING

4

(b)

CONV-3 CONV-2

5 4 3

FF-well5 FF-well4

FF-gi5

FF-gi4

CONV-3 CONV-2

FF-well5 FF-well4

FF-gi5

FF-gi4

CONV-3 CONV-2

5 4 3 2

2

1

1 (c)

5

1 FF-well5 FF-well4

(a)

6

FF-well5 FF-well4

FF-gi5

FF-gi4

(d)

CONV-3 CONV-2

Fig. 5. Perceived (a) familiarity, (b) healthiness, (c) liking and (d) willingness to use six beverages in a blind test (Blind) and evaluated without tasting by the appearance and information (Expected) (1, not at all pleasant, familiar, healthy or willing; 7, extremely pleasant, familiar, healthy or willing). FF-well, labelled with an effect on gut well-being; GI-gi, labelled with a blood glucose level balancing effect. The numbers 2–5 refer to the price in VTT Coins.

FF-well5

FF-gi5

FF-well4

FF-gi4

Conventional

100% 80% 60% 40% 20%

at th e

1

en d

-1 5

t irs

ta st ed

C ho ic es

e C ho ic

-B lin d Li ki ng

Li k

in g

-E

xp

ec t

Li ki ng

ed

(n o

-B lin d

tt

as

-F

te

ta st ed

d)

0%

Fig. 6. Relative shares of hedonic ratings and of the beverage choices at different stages of Experiment II. FF-well, labelled with an effect on gut wellbeing; GI-gi, labelled with a blood glucose level balancing effect. The numbers 5 and 4 refer to the price in VTT Coins.

expectation scores, but they transferred their choices to the second and the third preferred alternatives when the price of the most preferred alternative increased. Also,

La¨hteenma¨ki & van Trijp (1995) found that the participants made their sandwich choices from among the three most preferred alternatives. These results indicate that consumers

ARTICLE IN PRESS N. Urala, L. La¨hteenma¨ki / Appetite 47 (2006) 302–314

312

Table 5 Pearson’s correlation coefficients between the relative liking ratings, participants’ background attitudes and the relative share of beverage choices (Experiment II) Relative liking ratings Expected

Tasted

Relative share of beverage choices FF-well5 FF-well4 FF-gi5 FF-gi4

Background attitudes

0.30* 0.34** 0.29* 0.44**

Beginning

At the end

GHI

FF REW

FF NEC

FF CON

0.21
0.05 0.21 0.26**

0.45** 0.24 0.37** 0.45**


0.07 0.04 0.01
0.18

0.05 0.09
0.09
0.02


0.09 0.13 0.11 0.01

0.06
0.06
0.12 0.10

FF-well, labelled with effect on gut well-being; GI-gi, labelled with blood glucose level balancing effect. Numbers 5 and 4 refer to price in ‘‘VTT Coins’’. Codes are only used for indicating the samples in this paper and they were not shown to the participants. GHI, General Health Interest; NPI, Natural Product Interest; FF REW, perceived reward from using functional foods; FF NEC, necessity for functional foods; FF CON, confidence in functional foods. *po0.05; **po0.01; ***po0.001.

Non-interested

Experimenters

Frequent users 7

non sig.

non sig.

non sig.

p=0.003

Overall mean

p=0.036

p=0.001

p=0.003

p=0.001

p<0.001

p<0.001

7 6 ab b

5

b

b a

ab

4

EXPECTED LIKING

BLIND LIKING

6

a a

3

b a

b

b

ab

b

b

b

4

ab a

a a

3

a 2

2

1

1 (a)

5

n=12 15 33

11 18 31

22 21 17

10 17 33

55 71 114

FF-well5

FF-well4

FF-gi5

FF-gi4

All

non sig.

(b)

p=0.017

n=12 15 33

11 18 31

22 21 17

10 17 33

55 71 114

FF-well5

FF-well4

FF-gi5

FF-gi4

All

p=0.003

p=0.005

p<0.001

7

EXPECTED HEALTHINESS

6 b 5

b

b

ab

b

b b

b a

a

4

a 3

a

2

1 (c)

n=12 15 33

11 18 31

22 21 17

10 17 33

55 71 114

FF-well5

FF-well4

FF-gi5

FF-gi4

All

Fig. 7. (a) Blind hedonic ratings, (b) expected hedonic ratings and (c) excepted healthiness ratings given by non-interested (0 choices), experimenters (1 choice) and potential frequent users (2 or more choices) of a particular beverage (1, not at all pleasant, familiar, healthy or willing; 7, extremely pleasant, familiar, healthy or willing). FF-well, labelled with an effect on gut well-being; GI-gi, labelled with a blood glucose level balancing effect. The numbers 5 and 4 refer to the price in VTT Coins. Different letters indicate the difference between the means according to Tukey’s post hoc test, significance level po0:05.

ARTICLE IN PRESS N. Urala, L. La¨hteenma¨ki / Appetite 47 (2006) 302–314

not only choose the most liked options; instead they make compromises among several preferred alternatives between liking, price and other choice factors, such as healthiness, for instance. Understanding the relationship between these factors needs future studies in long-time choice and consuming studies. One possibility could be monitoring participants’ food receipts (Rosas-Nexticapa et al. 2005). In Experiment I, the health-related claims were provided at the beginning of the study and the healthiness perceptions did not change after this, indicating that the information ‘‘locked’’ the healthiness perception. The chocolate-covered bars were regarded as less healthy than the bars covered with seeds and cereal. The associations between sensory cues, information and healthiness perception provide an interesting frame for future studies. Recently, Mialon et al. (2002) reported that participants ðn ¼ 161Þ perceived the ‘‘wholemeal bread’’ and ‘‘multigrain muffin’’ as healthier than their ‘‘fibre enriched’’ and ‘‘white’’ counterparts even when tasted without labels. The GHI, the Natural Product Interest (Roininen et al., 1999) and the attitudes towards functional foods (Urala & La¨hteenma¨ki, 2004; Urala & La¨hteenma¨ki, in press) did not show any consistent correlations with the choices of products that included health-related information (except some single products). There are several explanations for the lack of association. Firstly, regardless of the welldistributed attitude means, the number of participants may still have been too small to achieve required statistical power. Secondly, it is possible that the products did not have enough variation in those characteristics towards which the attitudes were measured. For instance, the samples may have been perceived as being so similar in their naturalness that participants’ Natural Product Interest did not have any meaning in the choices. Unfortunately, any evaluations about perceived naturalness were not implemented. Thirdly, the personal relevance of the choices may have been so low that the attitudes simply had no meaning in the choices (Posavac et al., 2003). In addition, the choices may not have been made as spontaneously as they might have been made in real life. The number of alternatives was quite small when compared to the product alternatives in the real world, too. In these experiments, we demonstrated an approach for investigating the relationships between food choices made from a certain selection and liking, perceived healthiness and participants’ background attitudes. Although we were not able to verify any associations between choices and perceived healthiness or background attitudes, this approach provides a feasible framework to study food choices with an experimental set-up in the future. Further studies are needed especially to investigate the role of expectations and product experience in long-term choices. The effects of monetary constraint and background attitudes on choices should be studied with improved experimental designs by paying special attention to large enough consumer groups, a more natural choice context, wider selections of food

313

products and a more careful consideration of the monetary constraint. Acknowledgements We would like to acknowledge TEKES (National Technology Agency of Finland), which partly funded this research. This study is part of a joint project ‘‘Tools for consumer-oriented product development’’, and belongs to the Tailored Technologies for Future Foods programme at VTT Biotechnology. References Arvola, A., La¨hteenma¨ki, L., & Tuorila, H. (1999). Predicting the intent to purchase unfamiliar and familiar cheeses: The effects of attitudes, expected liking and food neophobia. Appetite, 32, 113–126. Bower, J. A., Saadat, M. A., & Whitten, C. (2003). Effect of liking, information and consumer characteristics on purchase intention and willingness to pay more for a fat spread with a proven health benefit. Food Quality and Preference, 14, 65–74. Cardello, A. V. (2003). Consumer concerns and expectations about novel food processing technologies: Effects on product liking. Appetite, 40, 217–233. Cardello, A. V., & Sawyer, F. M. (1992). Effects of disconfirmed consumer expectations on food acceptability. Journal of Sensory Studies, 7, 253–277. de Graaf, C., Cardello, A. V., Kramer, F. M., Lesher, L. L., Meiselman, H. L., & Scutz, H. G. (2005). A comparison between liking ratings obtained under laboratory and field conditions: The role of choice. Appetite, 44, 15–22. Deliza, R., & MacFie, H. J. H. (1996). The generation of sensory expectation by external cues and its effect on sensory perception and hedonic ratings: A review. Journal of Sensory Studies, 11, 103–128. Diplock, A. T., Agget, P. J., Ashwell, M., Bornet, F., Fern, E. B., & Roberfroid, M. B. (1999). Scientific concepts of functional foods in Europe: Consensus document. British Journal of Nutrition, 81, 1–27. Grunert, K. (2003). Purchase and consumption: The interdisciplinary nature of analysing food choice. Food Quality and Preference, 14, 39–40. Huotilainen, A., Seppa¨la¨, T., Pirttila¨-Backman, A.-M., & Tuorila, H. (2006). Derived attributes as mediators between categorization and acceptance of a new functional drink. Food Quality and Preference, 17, 328–336. Hurling, R., & Shepherd, R. (2003). Eating with your eyes: Effect of appearance on expectations of liking. Appetite, 41, 167–174. Jaeger, S., & MacFie, H. J. H. (2001). The effect of advertising format and means-end information on consumer expectations for apples. Food Quality and Preference, 12, 189–205. Ka¨hko¨nen, P., Tuorila, H., & Lawless, H. (1997). Lack of effect of taste and nutrition claims on sensory and hedonic responses to a fat-free yoghurt. Food Quality and Preference, 8, 125–130. Kihlberg, I., Johansson, L., Langsrud, Ø., & Risvik, E. (2005). Effects of information on liking of bread. Food Quality and Preference, 16, 25–35. King, S. C., Meiselman, H. L., Hottenstein, A. W., Work, T. M., & Cronk, V. The effects of contextual variables of food acceptability: A confirmatory study. Food Quality and Preference, in press. Kozlowska, K., Jeruszka, M., Matuszewska, I., Roszakowski, W., Barylko-Bikielna, N., & Brzozowska, A. (2003). Hedonic tests in different locations as predictors of apple juice consumption at home in elderly and young subjects. Food Quality and Preference, 14, 653–661.

ARTICLE IN PRESS 314

N. Urala, L. La¨hteenma¨ki / Appetite 47 (2006) 302–314

Ko¨ster, E. P. (2003). The psychology of food choice: Some often encountered fallacies. Food Quality and Preference, 14, 359–373. Ko¨ster, E. P., Couronne, T., Le´on, F., Le´vy, C., & Marcelino, A. S. (2002). Repeatability in hedonic sensory measurement: A conceptual exploration. Food Quality and Preference, 14, 165–176. Kramer, F. M., Lesher, L. L., & Meiselman, H. L. (2001). Monotony and choice: Repeated serving of the same item to soldiers under field condition. Appetite, 36, 239–240. Lange, C., Rousseau, F., & Issanchou, S. (1999). Expectation, liking and purchase behaviour under economical constraint. Food Quality and Preference, 10, 31–39. Lawless, H., & Heymann, H. (1998). Correlation, regression, and measures of association. In Sensory evaluation of foods—Principles and practices. USA: Chapman–Hall. La¨hteenma¨ki, L., & Tuorila, H. (1995). Consistency of liking and appropriateness ratings and their relation to consumption in a product test of ice cream. Appetite, 25, 189–198. La¨hteenma¨ki, L., & van Trijp, H. (1995). Hedonic responses, varietyseeking tendency and expressed variety in sandwich choices. Appetite, 24, 139–152. Levy, C. M., & Ko¨ster, E. P. (1999). The relevance of initial hedonic judgements in the prediction of subtle food choices. Food Quality and Preference, 10, 185–200. Lucas, F., & Bellisle, F. (1987). The measurement of food preferences in humans: Do taste-and-spit tests predict consumption? Physiology and Behavior, 39, 739–743. Meiselman, H. L. (1992). Methodology and theory in human eating research. Appetite, 19, 49–55. Mialon, V. S., Clark, M. R., Leppard, P. I., & Cox, D. N. (2002). The effect of dietary fibre information on consumer responses to breads and ‘‘English’’ muffins: A cross-cultural study. Food Quality and Preference, 13, 1–12. Nelson, P. (1970). Information and consumer behavior. Journal of Political Economy, 77, 311–329. Peryam, D. R., & Pilgrim, F. J. (1957). Hedonic scale method of measuring food preferences. Food Technology, 11, 9–14. Posavac, S. S., Herzenstein, M., & Sanbonmatsu, D. M. (2003). The role of decision importance and the salience of alternatives in determining the consistency between consumers’ attitudes and decisions. Marketing Letters, 14, 47–57.

Roininen, K., La¨hteenma¨ki, L., & Tuorila, H. (1999). Quantification of consumer attitudes to health and hedonic characteristics of foods. Appetite, 33, 71–88. Roininen, K., & Tuorila, H. (1999). Health and taste attitudes in the prediction of use frequency and choice between less healthy and more healthy snacks. Food Quality and Preference, 10, 357–365. Rosas-Nexticapa, M., Angulo, O., & O’Mahony, M. (2005). How well does the 9-point hedonic scale predict purchase frequency? Journal of Sensory Studies, 20, 313–331. Stein, L. J., Nagai, H., Nakagawa, M., & Beauchamp, G. K. (2003). Effects of repeated exposure and health-related information on hedonic evaluation and acceptance of a bitter beverage. Appetite, 40, 119–129. Tuorila, H., & Cardello, A. V. (2002). Consumer response to an off-flavor juice in the presence of specific health claims. Food Quality and Preference, 13, 561–569. Tuorila, H., Meiselman, H. L., Cardello, A. V., & Lesher, L. L. (1998). Effect of expectations and the definition of product category on the acceptance of unfamiliar foods. Food Quality and Preference, 9, 211–430. Urala, N., & La¨hteenma¨ki, L. (2003). Reasons behind consumers’ functional food choices. Nutrition & Food Science, 33, 148–158. Urala, N., & La¨hteenma¨ki, L. (2004). Attitudes behind consumers’ willingness to use functional foods. Food Quality and Preference, 15, 793–803. Urala, N., & La¨hteenma¨ki, L. Consumers’ changing attitudes towards functional foods. Food Quality and Preference, in press. Verbeke, W. (2005). Consumer acceptance of functional foods: Sociodemographic, cognitive and attitudinal determinants. Food Quality and Preference, 16, 45–57. Verbeke, W. (2006). Functional foods: Consumer willingness to compromise on taste for health? Food Quality and Preference, 17, 126–131. Westcombe, A., & Wardle, J. (1997). Influence of relative fat content information on responses to three foods. Appetite, 28, 49–62. Zandstra, E. H., de Graaf, C., van Trijp, H. C. M., & van Staveren, W. A. (1999). Laboratory hedonic ratings as predictors of consumption. Food Quality and Preference, 10, 411–418. Zanoli, R., & Naspetti, S. (2002). Consumer motivations in the purchase of organic food. A means-end approach. British Food Journal, 104, 643–653.