Objective: The aim of this study was to evaluate two interventions (one reward-based and one exposure-based) for increasing children's acceptance of an unfamiliar vegetable compared with a no-treatment control. It was predicted that the exposure condition would increase liking for, and consumption of, the vegetable relative to either the reward or control group.
Design: Using a randomized controlled design, participants were assigned to one of two intervention groups (exposure or reward) or to a no-treatment control condition, for a 2 week period. Liking for, and consumption of, red pepper was assessed before and after the treatment period.
Setting: The study was conducted in three primary schools in London.
Subjects: Parental consent was obtained for 49 out of a possible 72 children.
Interventions: Interventions comprised eight daily sessions during which participants in the exposure group were offered a taste of sweet red pepper and told that they could eat as much as they liked. Participants in the reward group were shown a sheet of cartoon stickers and told that they could choose one of them on condition that they ate at least one piece of the pepper.
Results: The exposure-based intervention significantly increased both liking (P=0.006) and consumption (P=0.03) compared with the control group. The outcome of the reward intervention was intermediate and did not differ significantly from the exposure or control conditions.
Conclusions: Repeated exposure to the taste of unfamiliar foods is a promising strategy for promoting liking of previously rejected foods in children.
Sponsorship: This study was financed by Cancer Research UK.
There is now a consensus that the nutrients obtained from a diet high in fruit and vegetables can contribute to the prevention of cancer and cardiovascular disease and, in addition, may displace other foods from the diet, thereby indirectly reducing fat consumption (Hu et al, 2000; Liu et al, 2000; World Cancer Research Fund, 1995). However, fruit and vegetable consumption falls substantially below recommended levels in many developed Western countries (Gregory et al, 1990; Li et al, 2000; Steinmetz & Potter, 1996; Thompson et al, 1999).
Recent campaigns to increase fruit and vegetable consumption have often targeted children. This is partly because intervention earlier in the lifecycle might be expected to yield maximum health benefits (Perry et al, 1998). It is also because eating habits established in childhood are likely to track into adulthood (Kelder, et al, 1994; Resnicow et al, 1998; Singer et al, 1995). Furthermore, there is widespread concern amongst parents about their children's diets, and especially consumption of vegetables, and therefore considerable motivation to develop new strategies to improve them (Gibson et al, 1998). Interventions directed at modifying children's intake of fruit and vegetables have had limited success. For example, the ‘5-a-day’ programme in the USA comprised a series of interventions tailored to different sections of the community (Basch et al, 1994; Perry et al, 1998; Sorensen et al, 1999). Results so far appear limited in the light of the enormous investment involved (eg Havas et al, 1998), and there is still an urgent need for new strategies to increase consumption of fruit and vegetables.
The focus of ‘5-a-day’ and several other campaigns (eg Domel et al, 1993) has been on increasing knowledge and changing attitudes, at which they have been reasonably successful. Achieving significant changes in behaviour has proved more challenging, however, and it is possible that significant barriers to behaviour change lie elsewhere. An alternative perspective focuses on hedonic factors, eg affective reactions to taste, texture or appearance. Attempts to explain why vegetable consumption is so low, in adults as well as children, have often highlighted taste preferences as a significant barrier (Drewnowski et al, 1999; Wardle, 1995). Children's liking for vegetables has been shown to be the most significant predictor of their intake (Gibson et al, 1998). If liking—or rather disliking—is a major deterrent, then interventions which are designed to modify food preferences could provide some leverage in changing children's acceptance of vegetables.
Rozin's (1976) notion of ‘food neophobia’ provides a useful starting point in understanding food acceptance. Humans, in common with most omnivores, manifest an avoidance of new foods which may be adaptive in terms of providing protection from potentially poisonous substances. With experience (ie eating small amounts of the food and suffering no ill effects), the initial neophobic response can be transformed into acceptance via ‘learned safety’ (Kalat & Rozin, 1973). Rozin and Schiller (1980) also note that some tastes are initially unpalatable, eg bitter or spicy tastes, but even these can become liked with increasing experience. Exposure alone (often called ‘mere exposure’ in the literature) could therefore provide a means to modify acceptance of initially rejected or disliked foods. Experimental studies with infants, children and adults have provided strong support for the efficacy of exposure, demonstrating that tasting a food or drink more often increases liking for it (Birch & Marlin, 1982; de Silva, 1988; Pliner, 1982; Sullivan & Birch, 1994). Merely looking at the food has no such effect (Birch et al, 1987), but watching others consume the food may provide a form of modelling or ‘exposure by proxy’, which could also reduce rejection (Birch, 1980; Hobden & Pliner, 1995). Some further support for the efficacy of modelling comes from a series of small-scale studies evaluating a multi-component programme combining video-based peer modelling together with rewards for eating fruit and vegetables and daily letters of encouragement (the ‘Food Dudes’ study; Horne et al, 1995, 1998). The full programme produced positive effects on intake of fruits and vegetables, although the peer modelling video alone failed to achieve significant changes.
Promise of a reward (eat up your peas and you can have some ice cream) is a time-honoured parental tactic for promoting consumption where the child is resistant. However, it has been argued in the psychological literature that treating food consumption in this way, as the instrumental component of a contingency, may actually decrease liking for that food. Lepper and Greene's (1978)Overjustification Theory suggests that offering a reward for an action devalues it for the child. On this analysis, the child interprets his or her own behaviour with concepts such as ‘If I am rewarded for eating peas, I must not like them very much’. In support of this idea, a number of studies have reported decreased liking for foods when children are rewarded for eating them (Birch et al, 1982; Birch et al, 1984; Newman & Taylor, 1992). The detrimental effects of reward have come to be widely accepted in the child-feeding literature (Birch, 1999). However a recent review and meta-analysis of studies of reward suggested that the case may have been greatly over-stated (Eisenberger & Cameron, 1996), and that, under some circumstances, rewards can enhance willingness to perform a behaviour. For example, offering verbal rather then tangible rewards, small rather than large rewards, and rewards for quality rather than quantity of behaviour, seem to be more successful. Furthermore, the multi-component programme described above (Horne et al, 1995, 1998) included a significant reward component comprising a range of ‘Food Dudes’ prizes (stickers, baseball caps, lunchboxes etc), which were awarded to children who consumed specified quantities of target foods. There were some positive effects of reward alone, but not as great as when modelling and rewards were combined (Lowe et al, 1998).
The discrepancies in findings between studies may arise from differences in outcome measures, the context in which rewards are given, or indeed in the types of rewards offered, and these have never been investigated systematically. It is also important to consider the fact that any reward-based intervention must, insofar as the child consumes any of the food, involve exposure. Therefore, some degree of positive change might be expected as a consequence of this exposure, even if it is reduced relative to ‘exposure alone’ as a consequence of the overjustification effect. These considerations suggest that it would be valuable to compare the two procedures directly in the same population and under the same conditions.
The aim of the present study was to evaluate the efficacy of two interventions, one based on exposure alone, and one incorporating reward, in addressing the real-world problem of increasing children's liking for vegetables. Using a randomized controlled design, 5–7-y-old children were assigned to one of the two intervention groups (exposure or reward), or to a no treatment control condition, for a 2-week period. We predicted that the daily exposure condition would increase liking for, and consumption of the vegetable relative to either the reward or control group. On the basis of the existing laboratory studies, rewarding consumption was not expected to increase liking, but consumption was expected to rise from baseline enough to meet the reward criterion.
Ethical approval for the study was granted by the Bethlem and Maudsley Hospital Research Ethics Committee.
Participants were recruited from three primary schools in London. From a potential sample of 72 children, 49 (68.1%) parental consent forms were received. The children were reasonably balanced in terms of gender (23 boys and 26 girls) and ranged in age from 5 y 8 months to 7 y 10 months with a mean of 6 y 7 months. All participants attended both pre-and post-treatment assessments.
Participants were randomly assigned to one of three experimental groups: exposure (group E, n=15, 7 boys, 8 girls), reward (group R, n=16, 8 boys, 8 girls) and control (group C, n=18, 8 boys, 10 girls). In order to control for possible school effects, each group contained children from all three schools. Groups E and R were scheduled for eight sessions of either ‘mere exposure’ or reward—although of the 31 children assigned to the two intervention groups, nine children failed to attend all treatment sessions. In the reward group, one child missed one session, and four missed five; in the exposure group three children missed one session, and one missed six. Sweet red pepper (capsicum) was selected as the ‘target’ vegetable for this study since exploratory taste tests with this age group had suggested that it was both comparatively novel and relatively disliked. Of the sample in the present study, 67% had tried it before. Liking and intake of red pepper were assessed before and after the eight treatment sessions. In order to examine the process of change in the two intervention groups, liking and intake were also assessed in each treatment session.
Consumption was measured by counting the number of small pieces of raw red pepper (average weight 1.8 g) eaten by each child. Liking was assessed after tasting using a five-point ‘faces’ scale designed to represent the following responses ‘I like it a lot’, ‘I like it a little bit’, ‘It is okay’, ‘I dislike it a little bit’, ‘I dislike it a lot’ (scored 2 to −2). The scale comprises cartoon faces, varying in expression from a broad smile to deep frown and designed to appear neutral with respect to their sex. The reliability of such a measure of preference in young children has been demonstrated by Birch et al (1980) and, more recently, by Guthrie et al (2000). Participants' understanding of the scale was established through discussion of hypothetical situations prior to using it to rate the taste of the pepper.
The study was carried out in schools over a 2-week period (10 school days). At the same time every day (between 10 am and 12 noon), children were seen individually in a quiet room close to their classroom. All children completed pre- and post-treatment assessments, which took place on day 1 and day 10, respectively. Children were presented with fresh, sweet red pepper cut into small pieces, and invited to eat as much as they liked, having previously watched the experimenter eating a piece. After they had eaten one piece, children were asked to rate their liking for the taste of the pepper using the measure detailed above. Consumption was measured by counting the number of pieces eaten.
Between the pre- and post-treatment assessments, participants randomized to the exposure and reward groups completed eight treatment sessions from day 2 to day 9 of the study period. The daily procedure for the exposure group was the same as for the pre- and post-treatment assessments. They were offered a taste of red pepper and told: ‘Please have a taste of this red pepper. You can eat as much as you like’. Liking was rated using the ‘faces’ scale and number of pieces eaten recorded.
Participants in the reward group were shown a sheet of stickers featuring a variety of well-known cartoon characters and told: ‘Please have a taste of this red pepper. If you eat at least one piece, you may choose one of these stickers. You can eat as much as you like’. They rated their liking in the same way as the exposure group and were then invited to choose their sticker, having established that the child did not want to eat any more pieces. Participants in the control condition were not seen between the pre- and post-treatment tests and received no further intervention.
Statistical analysis was carried out using SPSS (Version 9.0). Analyses were performed on data from all participants (n=49) since evidence concerning the degree of exposure required to alter preferences is equivocal at present and, furthermore, results were qualitatively replicated when participants who failed to attend every treatment session were excluded. Correlations and analysis of variance were used to establish whether age or gender represented confounding variables. In order to test for differential effects of the three conditions on liking and consumption, repeated measures analysis of variance was used to compare changes in liking and consumption from before to after the intervention period across the three groups. Paired t-tests were used to examine the extent to which participants within each group changed their liking and consumption. Further repeated measures analyses were used to examine the profile of changes in liking and consumption over the treatment sessions (day 2 to day 8) in groups E and R, and to compare the groups in terms of the patterning of changes in consumption and liking.
No significant effects of age or sex were observed on liking, consumption nor any changes over time, so data from all participants were combined for all the analyses.
Effects on liking
Pre-treatment, mean liking for red pepper was lower in the exposure group than in either reward or control groups, although this difference was not significant (F(2, 46)=1.56, P=0.22).
Changes in liking from pre- to post-treatment are shown in Figure 1, indicating greater liking after treatment, which was confirmed in the significant effect of time (pre vs post) in the repeated measures analysis of variance (F(1, 46)=24.75, P<0.001). As predicted, the change in liking between pre-test and post-test differed significantly between groups (group-by-time interaction: F(2, 46)=5.37, P<0.01). Post-hoc analyses showed that the exposure group changed significantly more than the control group (mean difference=1.79, P<0.01). The reward group had an intermediate level of change, which was not significantly different from either the exposure or the control groups. However, significant linear effects in the extent of change in liking were observed across the three groups (F(1, 46)=10.61, P<0.005), with the change in the reward group being intermediate between the others. Paired t-tests indicated that participants in both intervention groups showed increased liking for red pepper over the study period (group E: t=4.68; P<0.001; group R: t=2.51; P<0.05). Participants in Group C did not significantly change in their liking over the study period.
Effects on consumption
Changes in consumption from before to after the intervention are shown in Figure 2. There was a significant main effect for time (F(1, 46)=25.83, P<0.001) and a significant group-by-time interaction (F(2, 46)=3.51, P<0.05) for the number of pieces of red pepper that were consumed. Paralleling the liking results, post-hoc analyses showed that the difference between groups E and C was significant (mean difference=5.44, P<0.05), but comparisons involving group R did not reach significance. Again, significant linear effects in the extent of change in consumption were found across the three groups (F(1, 46)=7.01, P<0.05). Participants in all three groups significantly increased their consumption of red pepper over the study period (group E, t(14)=3.85, P<0.01; group R, t(15)=2.30, P<0.05; group C, t(17)=3.5, P<0.01).
Process analysis—liking and intake during the intervention sessions
Liking for red pepper at pre- and post-treatment sessions and at each treatment session (T1–T8) for children who attended all eight sessions in groups E and R are illustrated in Figure 3. This shows that liking increased from the pre-treatment test (which was, for many, their first experience of eating red pepper) to the first treatment session, especially in the reward group. The overall group by time effect over sessions 1–8, however, was not significant. Nevertheless, out of interest we examined the patterning of results to see if the profile of change over the treatment period looked at all different in the two groups. A repeated measures analysis of variance revealed that the liking ratings of group E showed a significant linear increase over the eight treatment sessions (F(1, 10)=5.23, P<0.05) while no such trend was observed in group R.
Intakes at pre- and post-treatment assessments and at each treatment session (T1–T8) for groups E and R are illustrated in Figure 4. The data showed a similar pattern to that for liking. In both groups, intake in the first treatment session was up on the pre-treatment test. On average they also consistently ate more than the minimum requested in the exposure protocol (taste one piece) or the reward protocol (if you taste a piece of pepper you can have a reward) in every session. Mean intakes over all treatment sessions were 6.73 (s.d.=5.30) for group E and 4.57 (s.d.=3.77) for group R, with a total intake over all sessions of 716 pieces in group E and 473 pieces in group R. Neither of these differences was statistically significant. However, group E showed a steady increase in consumption over the eight treatment sessions, reflected in a significant linear time trend in a repeated measures analysis of variance (F(1, 10)=9.95, P<0.05), which was not the case in group R.
The results of the present study support the prediction that daily exposure to the taste of a target vegetable can increase children's liking and consumption compared with no exposure. Eight sessions of exposure shifted children's liking from being on the margin between liking and disliking (−0.06 on a scale from −2 to +2) to maximum liking (2.00). Intake increased from just over one piece of pepper on the first test session to more than nine pieces on the last session. These results are supportive of the findings of previous research in this area (Birch & Marlin, 1982; Sullivan & Birch, 1994) and suggest that exposure is a promising technique for improving children's liking of vegetables, which merits further investigation in a real-world setting.
The outcome of the reward intervention was intermediate, lying between, and not differing significantly from, the exposure or the control conditions. This could be dismissed as a negative result, in the light of the lack of statistical significance in the difference from the control condition. However, it should be noted that the ordering of the three outcomes was significantly linear, indicating that the reward effect lay between the effects of control and exposure groups. This means that, contrary to other findings in this area (Birch et al, 1982, 1984; Newman & Taylor, 1992), the promise of a reward did not decrease liking. Furthermore, the amount of red pepper consumed by children in the reward group during the treatment sessions was always higher than necessary to achieve the reward, suggesting that their behaviour was not controlled entirely by the promise of reward. In interpreting these results it is important to remember that being rewarded for consumption ipso facto involves exposure—the two cannot be separated. This means that the reward condition is in practice a ‘reward plus exposure’ condition. Given the linear ordering in the change in liking over the three groups (exposure>reward>control), it is reasonable to propose that, rather than having an entirely adverse effect on liking, reward acts to limit the positive effects of exposure. This pattern of findings is consistent with the idea that offering rewards for consumption might have two effects. First it should promote consumption in the short term, thereby achieving exposure, which enhances liking. At the same time, being rewarded for eating a food might modify children's view of the acceptability of the food, along the lines suggested by the overjustification hypothesis (Lepper & Greene, 1978). It is worth noting that, where success has been claimed for reward-based interventions (Horne et al, 1995, 1998), the main outcome measure has been consumption. In contrast, studies reporting detrimental effects of reward (eg Birch et al, 1982) have typically used verbal statements of preference. At least in the short term, there may be differential effects of reward on liking and consumption, and the inclusion of both measures in the present study allowed us to look in more detail at these effects.
Surprisingly, the control group also showed a statistically significant increase in consumption over the study period, despite receiving no intervention. No corresponding increase in liking was observed. It may be that a single exposure is sufficient to reduce the typical neophobic response of reluctance to taste, but that further exposures are required to effect changes in liking.
We also analysed the findings in the two treatment groups on a session-by-session basis over the whole treatment period, in an attempt to shed light on the mechanisms involved in the changes in liking. The patterning of results over the eight treatment sessions suggested that the exposure procedure (‘Please have a taste of this red pepper. You can eat as much as you like’) resulted in a slow, steady increase in both liking and consumption, as evidenced by the significant linear trends over the treatment sessions. In contrast, the reward procedure (‘Please have a taste of this red pepper. If you eat at least one piece, you may choose one of these stickers. You can eat as much as you like’) produced a swift increase from the pre-test to treatment session one, but after that no significant increase in consumption over the treatment sessions. The consumption effect was paralleled by a less steady increase in liking in the reward group. These findings suggest that reward is contra-indicated as a tech-nique for producing a steady increase in consumption, although it may be highly effective in the short term. It is worth noting, however, that the reward used (stickers) may not have been equally valued across participants in group R and that this may have diminished their effect.
A further problem that has been identified with the offering of rewards concerns their eventual withdrawal. Failing to receive an expected reward may lead to extinction of the previously reinforced behaviour—in this case eating the food. The present study used only a single non-rewarded post-treatment test and therefore had limited opportunity to test this proposition. Looking at the present results, liking continued to increase in the exposure group but declined in the reward group; however, this finding was not matched in the intake data. It is possible that further un-rewarded sessions would have seen a greater decline in consumption in the reward group. Alternatively, insofar as the child continues to eat some of the red pepper, the beneficial effects of exposure alone could take over. Horne et al (1995, 1998) found that consumption of vegetables was still above baseline levels in many participants at 6-month follow-up. These results therefore suggest that a longer-term follow-up of the effectiveness of rewarded changes in consumption is crucial.
Whilst the present study has demonstrated the efficacy of an exposure-based intervention in increasing children's acceptance of an unfamiliar vegetable, the underlying mechanisms of change remain unclear. That ‘mere exposure’ produced increases in liking and consumption was expected, and is supported by the findings of previous research in this area (Birch & Marlin, 1982; de Silva, 1988; Pliner, 1982; Sullivan & Birch, 1994). The finding that rewards did not decrease liking raises new questions. However we have argued that, as reward necessarily involves exposure, the way in which these two factors interact requires further investigation.
There are a number of limitations to the present study which could be usefully addressed in future research. In common with many published studies in this area, the sample size was comparatively small, and would limit generalization to a wider population. In addition, it was beyond the scope of this study to control the length of time that had elapsed since participants had last eaten, although none would have had the opportunity to eat since breakfast. Furthermore, the setting of the study (in schools), and the implementation of the intervention by an unfamiliar researcher, meant that the social situation in which the children found themselves was far from naturalistic. Since most children's meals are served to them at home by a parent, there is a strong case for locating future studies in home settings and for teaching parents themselves how to apply the exposure approach. There seems little doubt that parental co-operation would be good given the level of concern they commonly express about their children's intake of vegetables. Finally, the lack of long-term follow-up in the present study means that we are unable to determine whether the positive effects of either exposure or reward are lasting and indeed whether any effects generalize to similar foods over time.
One of the most important longer-term contributions of this line of research may be to improve the advice given to parents on how to manage food rejections in their children. Our findings suggest that repeatedly inviting the child to taste a small amount of a rejected or disliked food, without great emphasis on how much they eat, is a good strategy for promoting liking. The element of repetition may be particularly crucial since the process of adaptation may be slower than parents expect, leading some parents to give up before the exposure has had a chance to work. We know that most parents will, at some point, resort to using rewards since it seems to them to be common sense. The results of the present study indicate that the offer of reward limits the effectiveness of exposure. However, there may be an argument for the use of rewards, to encourage a child who flatly refuses to taste the food at all, since a food must be tasted for exposure to be effective (Birch et al, 1987). It seems likely that both parents and health professionals would welcome scientifically based guidance on how to increase children's acceptance of vegetables.