A robust association exists between fruit and vegetable (F&V) consumption and reduction in all-cause mortality and in the occurrence of several chronic diseases such as cancer and cardiovascular disease (Wang et al., 2014; Aune et al., 2017). The United Kingdom (National Health Service (NHS), 2018) recommends that adults (≥18 years old) should consume at least five portions (5 × 80g) of F&V daily. Despite the introduction of the national “5-a-day” campaign (National Health Service, 2018), the latest surveys show national consumption of F&V falls short of the recommended amount (Rose, 2018; NHS Digital, 2019). The age cohort who consume the least F&V are young adults (16–24 years), who, in England, consume approximately 3.3 portions daily (NHS Digital, 2019), while intake by their Scottish counterparts is lower at 3.2 portions daily (Rose, 2018).

Young adults’ low F&V consumption is concerning as it is the period in which eating habits begin to form, after which resistance to change of established habits increases with age (Gall et al., 2000; Lien et al., 2001). Young adults’ eating behavior is predominantly influenced by peers (Stok et al., 2015; König et al., 2017). Therefore, harnessing social influences may be an effective approach to improving F&V intake (Higgs and Thomas, 2016; Stok et al., 2016).

Social norms are defined as behavioral standards that indicate appropriate and correct behavior (Aronson et al., 1998) and can be used in models as determinants of intentions and behavior. For example, the Theory of Planned Behavior (TPB; Ajzen, 1991) posits that intentions are determined by one’s attitudes, perceived behavioral control, and subjective norms (i.e., one’s norm perceptions) and predicts subsequent behavior from intentions (Ajzen and Madden, 1986). Norms may reflect what the group should be doing, i.e., perceived approval about a behavior (injunctive norms) or what the group is actually doing, i.e., perceived behavior (descriptive norms; Cialdini et al., 1990). Exposure to descriptive norm messages has been found consistently to alter eating behaviors in field experiments (Mollen et al., 2013; Thomas et al., 2017), experimental laboratory studies (Stok et al., 2012, 2014b), systematic reviews, and meta-analyses (Robinson et al., 2014b; Stok et al., 2016 for reviews). Recent research also indicates that descriptive norms are often more successful in increasing F&V intake than conventional messages highlighting the health implications of consuming sufficient F&V (Croker et al., 2009; Mollen et al., 2013; Robinson et al., 2014a).

Prior studies have used non-norm-based messages as a comparator to descriptive norm-based messages. However, de Bruijn et al. (2015) argue that such control messages lack validity. When it comes to “real-life” normative content, it is norms regarding unhealthy behavior—problem behavior—that are most frequently conveyed by mass media and health campaigns (Schultz et al., 2007; Stok et al., 2012; Niederdeppe et al., 2014). de Bruijn et al. (2015) have investigated whether desired descriptive norms were effective when compared with undesired or “problem” descriptive norms on eating intentions (fruit) and behavior, but this small study focused on older adults who may be less sensitive to normative influences than young adults (16–24 years old; Steinberg and Monahan, 2007). Therefore, the impact of descriptive norms highlighting desired behaviors (i.e., sufficient intake) compared with those that focus on problem behaviors (i.e., insufficient intake) on eating intentions and behavior is unclear.

When individuals identify with a group they are more motivated to adhere to in-group norms than out-group norms (Higgs, 2015; Reynolds et al., 2015; Tarrant et al., 2015). Young adulthood is the period throughout which individuals acquire a range of identities (e.g., student) and are motivated by their need to belong (Baumeister and Leary, 1995; Arnett, 2000). Within an eating behavior context, Louis et al. (2007) were the first to investigate the association between group identification strength and perceived eating norms in a longitudinal predictive study. They found high identifiers reported group-congruent intentions, whereas low identifiers did not. However, in a two-week follow-up, identification strength was not found to predict behavior (Louis et al., 2007). Further evidence suggests that the effect of identification strength on the influence of norms on eating behavior is not conclusive (Dempsey et al., 2018), with recent studies suggesting that high identification may result in both norm-divergent behavior (Banas et al., 2016) and convergent behavior (Liu et al., 2019).

The aim of this exploratory research was to investigate whether a descriptive norm message communicating a sufficient F&V intake norm is effective in improving F&V intake intentions and subsequent intake compared with a message communicating an insufficient F&V intake norm. Additionally, we explored whether the influence of the descriptive norm messages depends on the strength of student identification. To examine the effect of identity strength, an identity manipulation was included to categorize participants into distinct “low” and “high” identifier groups.

This study compared the effect of a descriptive norm message communicating the “sufficient” F&V intake of in-group members with an “insufficient” F&V intake message, on immediate F&V intake intentions and subsequent intake at a two-day follow-up. Whether the impact of descriptive norms was dependent on student identification strength was also investigated by employing a manipulation to categorize students as “low”/“high” identifiers. The manipulation was not fully successful in creating distinct “low”/“high” identifiers, and participants in the “low” identification group displayed relatively high identification, which is frequently observed in studies (Stok et al., 2012; Banas et al., 2016; Liu and Higgs, 2019). We found that participants in the “low” identification group intended to and acted norm-congruently, while participants in the “high” identification group intended to and acted against the presented norms.

Some social norms research asserts that higher identification predicts norm-congruent behavior (Louis et al., 2007; Stok et al., 2012, 2014a). For example, in a similar study, Liu et al. (2019) investigated the relationship between descriptive norms and identification strength on participants’ F&V intake. The researchers primed one group’s student identity and showed participants a flyer or a poster either communicating descriptive norms about most students consuming over 5 portions of F&V daily or a health message conveying the benefits of eating 5-a-day. They found the primed group consumed 40 g more F&V from a buffet than the non-primed group. In contrast, the present data show that only the “low” identification group participants’ fruit intake intentions and vegetable intake were norm-congruent. Several explanations may be attributable to the differences in the current findings and Liu et al.’s (2019) experiment. Crucially, Liu et al.’s (2019) sample consumed 2 F&V portions at baseline, whereas our sample reported 4.5 portions. Previous studies have indicated substantial differences in the effect of norms on “high” and “low” F&V consumers, with “low” consumers being more prone to match norms (Schultz et al., 2007; Robinson and Higgs, 2012; Robinson et al., 2014a; Verkooijen et al., 2015). Therefore, the contrasting findings could be attributed to baseline F&V intake. Additionally, the difference between the identity manipulations used by Liu et al. (2019) and in the present study may also account for the observed contradictory results.

The present data also suggest that participants in the “high” identification group actually diverged from norms. This manifested in the present study in two ways. Firstly, participants in the “high” identification group intended to eat more fruit and increased their vegetable intake upon receiving the “insufficient” descriptive norm, although this was not significant. This may be explained by their desire not to be associated with a group which has unfavorable norms (Berger and Heath, 2007; Berger and Rand, 2008), leading to a compensatory behavior. Secondly, participants in the “high” identification group intended to eat fewer fruit portions and (non-significantly) decreased their vegetable intake by half a portion upon receiving the “sufficient” descriptive norm. The finding corroborates Banas et al. (2016), who demonstrated that “high” identifiers chose calorific food items from an online menu when presented with “healthy” descriptive norms, indicating the presence of ironic effects. Banas et al. (2016) suggested that observed ironic effects could be explained by vicarious licensing. Vicarious licensing posits when high identifiers perceive their in-group members making progress in achieving a goal (e.g., eating healthily), they may give an individual license to themselves (e.g., choose unhealthy food; Kouchaki, 2011). This ironic effect has primarily been associated with hedonic consumption, where one is offered an alternative choice (Wilcox et al., 2009; De Witt Huberts et al., 2012). However, as the current study did not investigate vicarious licensing, nor offer an alternative choice, a definitive conclusion cannot be drawn as to whether this is the underlying mechanism for the findings. Taken together, the data suggest that understanding of the moderating effects of identification on responses to eating norms requires further investigation.

In the present study, an approximate half-portion (~ 40 g) difference (non-significant) was consistently observed between descriptive norm conditions, which is noteworthy, given that long-term school-based dietary interventions for children (5–12 years) can only demonstrate an increase in F&V intake by an average of one-quarter to one-third of a portion (~20–30 g; Evans et al., 2012). The half-portion difference is clinically relevant given the dose-related relationship between F&V intake and diseases such as cardiovascular disease and cancer (Aune et al., 2017), and evidence indicating that each additional serving of fruit or vegetable a day is associated with 5–6% reduced risk of all-cause mortality (Wang et al., 2014).

It is important to note that as the manipulation was not fully successful to create distinct “low”/“high” identifiers, participants in the “low” identification group displayed relatively high identification, which is frequently observed in studies (Stok et al., 2012; Banas et al., 2016; Liu and Higgs, 2019). Consequently, the two identification groups could be regarded as “identifiers” and “extreme identifiers,” whereby “identifiers” perceive descriptive norms relevant, and thus act norm-congruently, a well-documented finding (Cruwys et al., 2012; Stok et al., 2012; Reynolds et al., 2015). Our findings showed that the intention and consumption of “identifiers” were poorer in the “insufficient” condition, which may be attributed to the “backlash” effect (Cialdini, 2003), which is an undesired behavioral outcome following exposure to undesired norms conveying problem behaviors about one’s in-group.

Our findings in relation to the “sufficient”/“desired” descriptive majority norms compared with “problem”/“insufficient” norms on intention and behavior show that when identity is not taken into account, there were no differences in their impact. These findings support the only similar investigation conducted to date by de Bruijn et al. (2015), who found desired descriptive norms had no effect on fruit intake intentions and intake when compared with undesired, “problem norm” content. These findings are broadly consistent with available field research on drinking behavior (Foxcroft et al., 2015). However, an explanation for the non-significant differences between descriptive norms may also lie in participants’ norm recall rates. Participants in the “insufficient” condition recalled norms more successfully than those in the “sufficient” condition, suggesting the bogus “sufficient” norm was perhaps perceived as inaccurate. In support of this suggestion, research shows students are generally perceived to be “unhealthy” (Tarrant and Butler, 2011) with students often overestimating peer’s poor health behaviors (Neighbors et al., 2006).

Although descriptive norms offer a cost-effective and simple approach to improve F&V intake intentions and behavior, and are successful when compared with no-norm controls and health messages (Robinson et al., 2014a), their effectiveness has not yet been demonstrated compared with undesired normative content in an eating behavior context. However, descriptive norms influenced fruit intake intentions and vegetable intake when investigated for their interaction with the identification manipulation, with participants in the “low” identification group acting norm-congruently, and participants in the “high” identification group diverging from the presented norms. The latter potentially suggests the ironic effects of high identification on behavior. Whether the findings generalize to other health behavior contexts, and to the general young adult population who would benefit from F&V intake improvement, remains subject to further investigation.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by the University Teaching and Research Ethics Committee at the University of St Andrews (MD14242). The participants provided their written informed consent to participate in this study.

All authors listed have made a substantial, direct, and intellectual contribution to the work and approved it for publication.

The funding for open access was supported by the University of St Andrews.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.