Identity processes and food choice: predictors of dietary lapses among ethical and health vegans

ABSTRACT

An interesting finding in the literature on vegetarianism and veganism finds that vegetarians and vegans often report that they deviate from their diets from time to time. Work examining this phenomenon finds that these dietary lapses relate to many factors; however, little research examines how these factors collectively influence dietary lapses while also controlling for the relationships that may exist among factors. Here, I fill this gap by drawing from the unified model of vegetarian identity (UMVI) and identity theory (IT) to propose an inclusive model of dietary lapses. Structural equation model results from a sample of 488 vegans reveal differences in how identity and interactional processes relate to dietary lapses across ethical and health motivations. This work is important because it highlights how identities relate to dietary behaviors differently for ethical and health vegans; it also provides fruitful avenues for future work in this area.

KEYWORDS:

• Vegan
• dietary lapses
• identity
• social identity recognition
• social interaction
• ethical vegan
• health vegan
• dietary adherence

Introduction

What does it mean to be vegetarian or vegan? Some of the most well-known and defining features of veg*anism¹ involve eating patterns and food norms. For example, people across the dietary spectrum agree that vegetarians do not eat meat, while vegans do not eat meat, fish, eggs, or dairy products (see, e.g., North et al., 2021). Criteria for inclusion in these categories thus seem clear: to be a veg*an, one must not eat these foods. However, what if one still does? Can individuals claim to be vegetarian or vegan if they knowingly violate veg*an food norms and eat meat, eggs, and/or dairy?

Although the answer may seem obvious, the reality is not so straightforward. Across the burgeoning literature on veg*anism, a consistent finding is that many self-identified veg*ans report eating non-veg*an foods from time to time (Barr & Chapman, 2002; Jabs et al., 2000; Rosenfeld & Tomiyama, 2019; Rothgerber, 2015a, 2015b). Research exploring this curious finding shows that a range of factors are associated with these dietary lapses. Some work finds that personal factors, such as primary motivation for veg*anism, relate to stricter diet adherence and fewer lapses (Hoffman et al., 2013; Rosenfeld, 2019; Rothgerber, 2015a, 2015b). Other work finds relationships with demographic factors, such as gender (Rosenfeld, 2020). Still yet, some work finds fewer lapses among those who frequently interact with other veg*ans (Cherry, 2006) or among those for whom veg*anism is a more central part of how they see themselves (Plante et al., 2019).

This past work is important, but several limitations restrict the inferences that can be drawn about veg*ans’ dietary lapses. For example, much of this work exhibits omitted variable bias, notably omitting the role of social relationships (see, Cherry, 2006 for an exception), and/or treats predictors as uncorrelated. To address these limitations, I synthesize insights from two identity perspectives that outline and organize the range of factors identified by past work. I then test the inclusive model on a large dataset of self-identified vegans to provide a more comprehensive account of vegans’ dietary lapses across motivation for being vegan. This research is important because it sheds light on how different identity processes may influence food decisions across vegans with different primary motivations – information that can inform not only vegans’ wellbeing (see, e.g., Beezhold et al., 2015; Rosenfeld, 2019) but also the collective efficacy of vegan dietary patterns (see, e.g., Poore & Nemecek, 2018).

Literature review

Dietary lapses among vegans

According to The Vegan Society (2022), veganism is “a philosophy and way of living which seeks to exclude – as far as is possible and practicable – all forms of exploitation of, and cruelty to, animals for food, clothing or any other purpose.” This definition highlights that veganism involves specific food norms; however, it also implies that veganism is a lifestyle, encompassing many facets of life extending beyond food (Cherry, 2006, 2015). Although the “encompassing” nature of veganism may not describe all vegans – for example, ethical vegans may be more likely than health vegans to apply veganism beyond food (Greenebaum, 2012b) – norms surrounding food are more universally understood. For instance, vegans typically agree that food norms in veganism exclude meat, fish, dairy products, and eggs (North et al., 2021).

Despite food norms being one of the most predominant features of veganism, research consistently documents vegans’ dietary lapses – or when vegans report knowingly eating non-vegan foods. For example, Jabs et al. (2000) note that many vegans in their sample reported having eaten dishes made with cheese or eggs when they were in someone else’s home. Similarly, Hoffman et al. (2013) found that vegans in their sample varied in their dietary restriction, with some failing to select dairy products and eggs as foods that they “never” eat. Rosenfeld (2019, 2020) measured the degree of dietary pattern strictness among veg*ans in his sample by prompting them to answer agreement with items such as, “I can be flexible and sometimes eat foods that go against my dietary pattern” and “From time to time, I eat meat;” he found that strictness varied, with vegans reporting an average strictness level of about 5 on a 7-point scale. Rothgerber (2015a, 2015b) asked vegans to estimate the number of times they had knowingly violated their diets since they adopted it. Average violations were high, with ethical vegans reporting about 3 instances and health vegans reporting about 16 instances.

What predicts vegans’ dietary lapses? Several studies address this question, finding relationships with personal, demographic, interactional, and other social psychological factors. Regarding personal factors, Rothgerber’s (2015a, 2015b) and Rosenfeld’s (2019) results document a divide among vegans with different primary motivations for veganism: ethical vegans reported fewer dietary violations and greater dietary strictness than other vegans. Hoffman et al. (2013) corroborate this finding, with ethical vegans showing higher dietary restriction as well as greater lengths of time being vegan than health vegans. Rosenfeld (2020) finds a demographic difference, with vegan women reporting greater dietary strictness than vegan men. Cherry (2006) found social interactional differences, where vegans who were part of a “punk” community interacted with more vegans and had stricter food definitions based in ethical motivations than those who did not associate with other vegans. Finally, Plante et al. (2019) examined social psychological factors, including features associated with participants’ vegan identities. They found an association where vegans who saw their vegan identity as a more important part of how they see themselves exhibited greater dietary restriction (as did vegans who reported stronger animal rights motivations).

Taken together, this past research suggests that a range of factors relate to dietary lapses among vegans. However, three features of this work limit the inferences that can be drawn. First, this work tends to restrict the factors under consideration, examining only some of the factors identified as important across similar research. A second limitation is that, even when many factors are included, these factors are often examined as independent and uncorrelated.

To illustrate, Rothgerber (2015a, 2015b) as well as Plante et al. (2019) examined dietary violations in relation to personal factors as well as social psychological factors related to identity. However, neither of these studies control for demographic factors (first limitation). Rosenfeld (2020) also examined dietary strictness, personal factors, and social psychological factors related to identity. However, each of these factors was only examined separately in relation to gender, making it unclear how dietary behaviors are impacted by the interrelationships that may exist between demographic, personal, and/or social psychological factors (second limitation). To ensure that these relationships are not spurious, overestimated, or underestimated (see, e.g., MacKinnon et al., 2000), there is thus a need to more comprehensively examine how the range of factors identified by past research simultaneously influence dietary lapses while accounting for the relationships that might exist between them.

A third limitation of past research is that almost none of the work cited above examined the role of social relationships or social interactions as icorrelates to dietary lapses. This is a concern because eating is a social activity (Delormier et al., 2009). Eating together can bond and unite individuals, reinforcing communality by sharing the same food (Kemmer et al., 1998; Miller et al., 1998; Morrison, 1996). How this may relate to behavior is illustrated in Cherry’s (2006) work: vegans who had social relationships and social interactions with other vegans, including those in which vegans shared food together (e.g., at potlucks), ate less non-vegan food overall.

Food can also serve as a point of division and contention. If dietary differences preclude sharing food, this can create negative emotions and social distance which may feel especially uncomfortable if experienced between vegans and their non-vegan friends or family (Bresnahan et al., 2016; Cherry, 2015; Hirschler, 2011; Markowski & Roxburgh, 2019). Vegans’ dietary behaviors may be affected by such negative emotions and social distance (Jabs et al., 2000; Rosenfeld & Tomiyama, 2019); thus, the role of social relationships and social interactions should also be explored as key factors in the work on vegans’ dietary lapses. Below, I explore two frameworks that can aid in this endeavor because they outline and organize how factors, including social relationships, may be related to food behavior: the unified model of vegetarian identity (UMVI) and identity theory (IT).

Theoretical frameworks

The Unified Model of Vegetarian Identity (UMVI)

The unified model of vegetarian identity (UMVI) is a framework for understanding the vegetarian experience (Rosenfeld & Burrow, 2017). The framework takes a social identity approach, viewing “vegetarian” as a social category based on which attributions about in-group and out-group membership are made. From this perspective, “vegetarian” is a social identity with distinct experiences that relate to contextual, internalized, and externalized dimensions. Contextual dimensions “situate vegetarianism within historical, sociocultural, and lifespan contexts” (Rosenfeld & Burrow, 2017, p. 81) and underlie the internalized and externalized dimensions.

Internalized dimensions capture individuals’ self-perceptions of their identities and include concepts such as motivation and centrality. Motivation refers to the reasons for being vegetarian, such as ethical or health reasons, whereas centrality refers to the importance of the vegetarian identity to how individuals see themselves. Externalized dimensions, on the other hand, capture more social elements of individuals’ identities, such as how individuals present themselves in interactions. For example, social labels are the markers that individuals use to identify themselves to others. Whether the people with whom one interacts acknowledge individuals’ social labels, however, is called social identity recognition (Hopkins et al., 2019).

Although the UMVI outlines several important dimensions noted by past research on dietary lapses, two points are worth mentioning. First, the UMVI does not consider more descriptive aspects of social relationships, such as the frequency of interactions with similar others (Cherry, 2006). Second, the UMVI does not theorize about the interrelationships between dimensions. For example, under the UMVI, it is unclear if and how centrality, social identity recognition, and motivation are related to each other or to food choices, let alone to daily interactional realities. Given these limitations, it is useful to incorporate insights from additional work that not only outlines but also connects these dimensions.

Identity Theory (IT)

Identity theory (IT; Burke, 1991; Burke & Stets, 2009; Stryker, 1968, 1980) is derived from structural symbolic interactionism, a perspective that is encapsulated in the premise that “society shapes self, which shapes social interaction” (Serpe & Stryker, 2011). From this perspective, behavior is a product of individuals’ selves, which are reflections of the societies in which individuals belong as well as participate. Identity theory starts from this premise but centers the concept of identity for “self.” Here, identities refer to the meanings that individuals use to describe themselves as occupants of social roles, members of groups or social categories, or as individuals with characteristics that make them unique (Burke & Stets, 2009; Stets & Burke, 2000; Stets & Serpe, 2013).

IT describes identities according to their measurable properties, such as identity prominence, or how important the identity is to how a person sees herself² (McCall & Simmons, 1978). Properties like identity prominence theoretically and empirically influence individuals’ behaviors: if an identity is very important to how a person sees herself, she is more likely to behave in ways characteristic of the identity (Stryker & Serpe, 1994). At the same time, properties like identity prominence are themselves theoretically and empirically influenced by social structure and the structure of individuals’ identity-based relationships. For example, interactional commitment indexes individuals’ identity-based relationships and describes the number and frequency of interactions that occur with others the context of an identity (Hays, 2017; Stryker & Serpe, 1982). If individuals interact with many others because of an identity, the identity is likely to be higher in identity prominence (Hays, 2017; Yarrison, 2013).

IT argues that identity processes should unfold similarly despite the type of identity under consideration. This means that the theoretical relationships outlined above should hold for veganism as a category-based social identity as they do for role or person identities (see, Stets & Burke, 2014). Indeed, a recent study applying identity theory to the vegan social identity confirmed that other identity processes operate as expected (Davis et al., 2019). However, this study did not assess if the society-self-interaction path model explained above holds for vegans.

Additionally, IT omits an aspect of social identities outlined as important to social identity approaches and the UMVI: whether interaction partners acknowledge the social labels that individuals use to define themselves. For example, a person’s vegan identity may be highly prominent, or important to how she sees herself, but it may also be important to her that, in social interaction, other people acknowledge the vegan social label that defines her as a member of the vegan category and socially recognize her vegan social identity (Hopkins et al., 2019). Some work suggests that this dimension of the vegan experience may also be associated with identity-based relationships and identity-consistent behavior (see, e.g., Greenebaum, 2012b). Thus, identity theory would benefit from incorporating this concept because it would increase the theory’s explanatory power for behavior related to food-based (as well as other) social identities.

Current research: synthesizing IT and the UMVI to understand vegans’ dietary lapses

In this paper, I synthesize insights from both the UMVI and IT. Specifically, I connect the factors identified as important by the UMVI and other veg*anism research under an IT framework while also incorporating a new concept for IT. Important factors include motivations for veganism, properties of the vegan identity including identity prominence and social identity recognition, and interactions with other vegans. Under an IT framework, interactions with vegans should positively influence identity prominence and social identity recognition; identity prominence and social identity recognition should then negatively influence dietary lapses as an identity-inconsistent behavior. Although IT theorizes that the effect of interactions with vegans influences behavior indirectly through identity properties, some research supports a direct link (Cherry, 2006). Thus, I also test this direct effect, where interactions with vegans should negatively influence dietary lapses. Formally, then, I predict:

H1: Interactions with vegans will be positively associated with identity prominence.

H2: Interactions with vegans will be positively associated with social identity recognition.

H3: Interactions with vegans will be negatively associated with dietary lapses.

H4: Identity prominence will be negatively associated with dietary lapses.

H5: Social identity recognition will be negatively associated with dietary lapses.

How do motivations for veganism fit into the expanded IT framework? A key emphasis in IT is the meanings individuals associate with their identities. In the context of veganism, motivations may indicate the main category of meanings individuals associate with being vegan. For example, an individual identifying as vegan for health reasons may think her vegan self as “healthy” and “nutrition-conscious,” whereas an individual identifying as vegan for moral or ethical reasons may think of her vegan self as “just” and “virtuous.” These differing meanings likely carry implications for how identity and other interactional processes unfold. For instance, ethical vegans may lose social ties for espousing moral/ethical concerns viewed as “radical” by non-vegans, whereas health vegans may preserve social ties for espousing health concerns viewed as more conventionally palatable (Greenebaum, 2012a; Markowski & Roxburgh, 2019). As a result, any ties that ethical vegans forge and maintain with fellow vegans may be especially supportive ones (Radnitz et al., 2015) that highly influence self-evaluation and behavior (see, e.g., M. Gallagher et al., 2022). This would mean that links theorized above may be stronger for ethical vegans compared to health vegans. To assess this possibility, I test the inclusive IT model among ethical and health vegans separately, generally expecting stronger associations for ethical vegans.

Method

Data and sample

This study uses data from a 2016 cross-sectional survey distributed electronically to two non-probability samples. Non-probability samples are useful for identifying and recruiting individuals from hard-to-reach populations, such as vegans, who are widely dispersed geographically (Simmons & Bobo, 2015). Although non-probability samples are not appropriate for generalization to larger populations, this strategy generated a large quantitative dataset, which is essential for testing complex theoretical relationships.

The first non-probability sample was provided by a survey vendor, where individuals opt-in to the survey vendor participant population. If during the opt-in process the individual indicated that they identify as vegan, they were solicited to participate in this study through an e-mail with a link to the survey. The second non-probability sample included vegans from two popular social media platforms at the time: Facebook and Tumblr. On Facebook, individuals were solicited for participation through posts on vegan-related Facebook groups. These groups varied in their primary motivation for being vegan (i.e., ethics or health, as indicated in their group title or group description) and their central location in the U.S. (e.g., Portland, Chicago, Boston). On Tumblr, individuals were solicited for participation through blog posts that were published by three vegan influencers. All posts included a standardized description of the study and a link to the survey. Across the two samples, individuals were eligible to participate if they indicated through pre-screening questions that they currently self-identified as a vegan who was 18 years or older.

The two non-probability samples reflect important differences that may exist across vegans. Vegans who are active on social media in vegan-specific groups may be more motivated to participate in research related to veganism, and thus, may represent specific vegan experiences. By contrast, even though the survey vendor sample consists of individuals who may be more motivated to participate in research more generally, this strategy likely captures a wider range of vegans than those who voluntarily join vegan social media sites. Examining both samples crosscuts a wide range of vegan experiences, serving as a strength of this study.

A total of 739 vegans completed the survey, including 535 survey vendor members and 204 social media recruits. For this analysis, I use data from 488 vegans, of which 376 were panel members and 112 were social media recruits, and of which 325 were ethical vegans and 163 were health vegans. A total of 251 participants were excluded due to one or more of the following reasons: they indicated that their primary residence was not in the United States (66 people), they indicated in the demographic questions that they were under 18 years of age (2 people), they did not plan to be vegan in the future (10 people), they had been vegan for less than 12 months (75 people), or they identified a primary motivation for being vegan other than ethical or health reasons (98 people).³

Compensation

Vegans recruited on social media were given the option to enter a prize drawing for one of three $25 gift certificates. For vegans recruited through the survey vendor, I did not offer compensation because participants accrued compensation directly from the survey vendor.

Measures

Dietary lapses

The dependent variable in this study is dietary lapses. Participants indicated how frequently they ate the following non-vegan items within the past 12 months: red meat, white meat, fish, shellfish, eggs, milk, cheese, and other dairy products. Response options included: “Never,” “One or twice a year,” “Once a month,” “Less than once a week,” “Once a week,” “Several times a week,” and “Daily.” Dummy variables were created with 0 reflecting items reported as “Never” eaten and 1 reflecting items reported as eaten more than “Never” (Hoffman et al., 2013). Items in this scale were reliabile for ethical vegans (Cronbach’s alpha = .855) and health vegans (Cronbach’s alpha = .852). For the analysis, I summed the number of items that participants reported eating more than “Never” (see, e.g., Hoffman et al., 2013).

Primary motivation for being vegan

Participants indicated their most important motivation for being a vegan out of an exhaustive list, including: “Moral/ethical reasons,” “Animal rights,” “Environmental reasons,” “World hunger,” “Political reasons,” “Health reasons,” “Weight loss,” “Religious or spiritual beliefs,” “Family/friends,” “To save money,” “Taste preferences,” “I was born and raised a vegan,” and “Other reason not listed.” Participants selecting “Moral/ethical reasons” or “Animal rights” were grouped as “Ethical” vegans, and participants selecting “Health reasons” or “Weight loss” were grouped as “Health vegans” (Hoffman et al., 2013). These reasons comprised majority of both samples, as less than 11% indicated a different primary motivation.

Interactions with vegans

Interactions with vegans was assessed through one item where participants indicated how frequently they talked with vegans about veganism. Response options included: “Never,” “One or twice a year,” “Once a month,” “Less than once a week,” “Once a week,” “Several times a week,” and “Daily.” This measure is akin to one of Hays’s (2017) interactional commitment items. Higher scores reflect more frequent identity-specific interaction with similar others.

Identity prominence and social identity recognition

Identity prominence was assessed through four items related to the importance of the vegan identity to how an individual sees herself. Participants indicated their agreement on a scale of 1 (“Strongly disagree”) to 4 (“Strongly agree”) with the following items: “I have a strong sense of belonging to veganism,” “My veganism is an important part of my self-image,” “I have come to think about myself as a vegan,” and “Being vegan is an important reflection of who I am.” These items were based on Brenner et al. (2014, 2018), though I made slight changes in wording to reflect the vegan identity context.

Social identity recognition was assessed through three items related to the importance of the vegan identity being acknowledged by others in social interaction. Participants indicated their agreement on a scale of 1 (“Strongly disagree”) to 4 (“Strongly agree”) with the following items: “It is important to me that other vegans know about me as a vegan,” “It is important to me that other non-vegans know about me as a vegan,” and “It is important to me that others in general know about me as a vegan.” These items were adapted from Brenner et al. (2014, 2018) to reflect how important it is to the individual that others see her as vegan.

To assess if the items reflected separate constructs, an exploratory factor analysis was performed. Results revealed two factors (see, Table 1). These factors reflect the expected identity prominence and social identity recognition dimensions. Identity prominence scale items were reliabile for ethical vegans (Cronbach’s alpha = .930) and health vegans (Cronbach’s alpha = .880), as were social identity recognition items (Cronbach’s alpha = .941 and .946, respectively).

Table 1. Factor loadings from factor analysis of prominence and social identity recognition items.

Items	Factor 1 “Identity Prominence”	Factor 2 “Social Identity Recognition”
I have a strong sense of belonging to veganism.	.76	.38
My veganism is an important part of my self-image.	.81	.35
I have come to think about myself as a vegan.	.76	.23
Being vegan is an important reflection of who I am.	.86	.31
It is important to me that non-vegans know about me as a vegan.	.27	.89
It is important to me that other vegans know about me as a vegan.	.36	.80
It is important to me that others in general know about me as a vegan.	.28	.88
Eigenvalue	2.84	2.62

Background variables

Several additional variables were assessed in the analysis. These include age, race/ethnicity, gender, income, length of time spent as a vegan, and sample type. Participants indicated their age in years. They indicated if they were “White,” “Black or African American,” “American Indian or Alaskan Native,” “Asian,” “Native Hawaiian or Other Pacific Islander,” “Multi-racial,” or “Other.” They also indicated if they were of Hispanic, Latino, or Spanish origin. From this, a dummy variable for race/ethnicity was created, where non-Hispanic/Latino White was coded as 1, and all other races/ethnicities were coded as 0. For gender, participants indicated if they identified as any of the following: “Woman,” “Man,” “Transgender,” “No gender,” or “Other.” Only 20 participants selected the latter three categories; however, all 20 were eliminated from the analysis due to one or more reasons outlined above. A dummy variable for gender was created, where “Man” was coded as 1, and “Woman” was coded as 0. Income was assessed by asking participants to select their personal annual income before taxes. Response options ranged from “Less than $9,999” to “Above $75,000” and increased in increments of $4,999 up to $49,999, where final increments included “Between $50,00 and $59,999” and “Between $60,000 and $74,999.” Participants indicated in years the length of time they have been a vegan. Finally, the sample variable coded the survey vendor sample as 0 and the social media recruits as 1.

Analysis

In this analysis, I first use t-tests and chi-squares to assess differences across ethical and health vegans on the main variables of interest. I then assess relationships among the variables with correlations. Finally, I use structural equation modeling to test the multivariate structure of relationships. The structural equation model tested here is a groups model that estimates equations separately but simultaneously for ethical and health vegans. The structural equation model includes a measurement model and structural model. For the measurement model, both identity prominence and social identity recognition were included as latent constructs; all other variables were observed. All background variables were entered into the model as exogenous predictors on all other variables. Invariance testing across groups revealed that all pathways in the measurement model could be constrained as equal across groups. For the structural model, invariance testing revealed that all pathways involving the background variables could be constrained as equal, as could the pathways between interactions with vegans and social identity recognition as well as identity prominence and dietary lapses. Thus, pathways that distinctly vary across ethical vegans and health vegans include interactions with vegans on identity prominence, interactions with vegans on dietary lapses, and social identity recognition on dietary lapses.⁴ I employed maximum likelihood with missing values (a full-information maximum likelihood technique) as the estimation strategy to preserve sample size.

Results

Sample descriptive statistics

Table 2 presents descriptive statistics for all variables included in this analysis by primary motivation for being vegan. On average, health vegans were older (M = 41.81, SD = 14.81) than ethical vegans (M = 34.51, SD = 13.89) (t[445] = −5.08, p < .001), though more ethical vegans were white (78.68%) than health vegans (67.08%) (X² [1] = 7.64, p < .01). More health vegans were men (38.51%) than ethical vegans (16.77%) (X² [1] = 27.54, p < .001), and they earned more on average (M = 8.37, SD = 3.82, corresponding to $40,000-$49,999, compared to M = 6.78, SD = 4.07, corresponding to $30,000-$39,999, respectively) (t[440] = −4.01, p < .001). There were no differences in the length of time being vegan; on average, all vegans reported being vegan for more than 6 years (M = 6.83, SD = 8.03).

Table 2. Descriptive statistics and differences by motivation.

	Ethical				Health				Differences	Effect Size
Variable	M	SD	Range	N	M	SD	Range	N	t-test	Cohen’s d
Age	34.51	13.89	18–93	302	41.81	14.86	18–75	145	t(445) = −5.08***	0.51
Income	6.78	4.07	1–12	286	8.37	3.82	1–12	156	t(440) = −4.01***	0.40
Years Vegan	7.05	8.10	1–40	314	6.37	7.87	1–35	155	t(467) = 0.87, ns	0.09
Dietary Lapses	1.26	2.00	0–8	317	2.89	2.57	0–8	158	t(474) = −7.58***	0.71
Vegan Interaction	4.08	2.00	1–7	322	3.26	1.81	0–7	158	t(478) = 4.36***	0.43
Identity Prominence	3.48	0.69	1–4	318	3.04	0.73	1–4	157	t(473) = 6.51***	0.62
Social Identity Recognition	2.76	0.92	1–4	304	2.47	0.91	1–4	156	t(458) = 3.19***	0.32
Variable	%			N	%			N	chi-square
White	78.68%			319	67.08%			161	X2 (1) = 7.64**	–
Man	16.77%			316	38.51%			161	X2 (1) = 27.54***	–
Social Media Sample	32.31%			325	4.29%			163	X2 (1) = 48.17***	–

Notes.*p < .05, **p < .01, ***p < .001

Regarding the main variables, on average, health vegans reported eating more non-vegan food items (M = 2.89, SD = 2.57) than ethical vegans (M = 1.26, SD = 2.00) (t[474] = −7.58, p < .001). Ethical vegans interacted more frequently with other vegans (M = 4.08, SD = 2.00) than health vegans (M = 3.26, SD = 1.81) (t[478] = 4.36, p < .001). Additionally, both identity prominence and social identity recognition were higher among ethical vegans (M = 3.48, SD = 0.69 and M = 2.76, SD = 0.92) than health vegans (M = 3.04, SD = 0.73 and M = 2.47, SD = 0.91) (t[473] = 6.51, p < .001 and t[458] = 3.19, p < .001). Overall, Cohen’s d effect sizes for these significant differences are moderate to high, ranging from a low of 0.32 for social identity recognition to a high of 0.71 for dietary lapses.

Correlations

Table 3 presents correlation coefficients among all variables. Many relationships are consistent with past work. For example, interactions with vegans is positively related to identity prominence (r = .37) and social identity recognition (r = .41) but negatively related to dietary lapses (r = −.22). Identity prominence and social identity recognition are also negatively related to dietary lapses (r = −.26 and r = −.11) but positively related to one another (r = .61). Last, health motivation is negatively related to interactions with vegans (r = −.22), identity prominence (r = −.29), and social identity recognition (r = −.15) but positively related to dietary lapses (r = .33). Other relationships of note include positive relationships between dietary lapses and age (r = .19), being male (r = .24), income (r = .19), and surprisingly, length of time being vegan (r = .07), as well as between the social media sample and age (r = .24), gender (r = .17), race/ethnicity (r = −.16), income (r = −.19), and health motivation (r = −.14).

Table 3. Correlation coefficients.

	1.	2.	3.	4.	5.	6.	7.	8.	9.	10.
1. Interactions with Vegans	–
2. Identity Prominence	.37***
3. Social Identity Recognition	.41***	.61***
4. Dietary Lapses	−.22***	−.26***	−.11*
5. Age	−.20***	−.06	−.04	.19*
6.White	.03	−.00	−.00	−.03	.08
7. Man	−.06	−.04	.04	.24***	.15**	−.10*
8. Income	−.01	−.11*	−.07	.19***	.29***	−.05	.27***
9. Years Vegan	−.06	.06	.06	.07*	.29***	.01	.11*	.22***
10. Social Media Sample	.35***	.20***	.17**	−.35***	.24***	.17***	−.16***	−.19***	−.14**
11. Health Reason	−.20***	−.29***	−.15**	.33***	.23***	−.13**	.24***	.19***	−.04	−.31***

Notes.*p < .05, **p < .01, ***p < .001

Structural equation model results

The correlations support the idea that significant relationships exist between dietary lapses and all the main variables of interest. But how relatively important is each variable for explaining dietary lapses across ethical and health vegans, and how is this affected by the relationships among the variables?

To address these questions, I turn to results from the main analysis, consisting of a structural equation model that estimates relationships across ethical and health vegans. First, standardized measurement model results are presented in Table 4. All items load onto their respective constructs, and this is consistent across primary motivation. Second, results for the structural model⁵ are presented in Table 5. Coefficients can be interpreted as the standard deviation unit change associated with a one standard deviation unit increase in the predictor. Standardized results facilitate direct comparisons among coefficients both within and between the ethical and health vegan solutions.

Table 4. Measurement model results (Standardized).

	Ethical	Health
Item	Identity Prominence
I have a strong sense of belonging to veganism.	.91***	.85***
My veganism is an important part of my self-image	.85***	.76***
I have come to think about myself as a vegan.	.81***	.73***
Being vegan is an important reflection of who I am.	.85***	.79***
Item	Social Identity Recognition
It is important to me that other non-vegans know about me as a vegan.	.86***	.91***
It is important to me that other vegans know about me as a vegan.	.95***	.95***
It is important to me that others in general know about me as a vegan.	.88***	.88***

Notes.All measurement coefficients were constrained to be equal across groups, but the standardized solution produces estimates separately by group.

*p < .05, **p < .01, ***p < .001

Table 5. Structural equation groups model results (Standardized).

	→ Interactions with Vegans	→ Identity Prominence	→ Social Identity Recognition	→ Dietary Lapses
Predictors
Interactions with Vegans
Ethical	–	.32***	.42***	.01
Health	–	.49***	.38***	−.20*
Identity Prominence
Ethical	–	–	–	−.25***
Health	–	–	–	−.20***
Social Identity Recognition
Ethical	–	–	–	.10
Health	–	–	–	.25**
Background Variables
Age
Ethical	−.14**	.06	.02	.06
Health	−.17**	.06	.02	.05
White
Ethical	−.01	−.06	−.02	.05
Health	−.02	−.07	−.03	.05
Man
Ethical	.00	.03	.09*	.13**
Health	.01	.04	.12*	.13**
Income
Ethical	.11*	−.09	−.11*	.08
Health	.11*	−.08	−.11*	.06
Years Vegan
Ethical	−.00	.06	.11*	−.02
Health	−.00	.16*	.11*	−.01
Social Media Sample
Ethical	.34***	.06	.07	−.30***
Health	.16***	.02	.03	−.10***

Notes.N = 488; X² (128) = 144.76 p = .147; RMSEA = .023; CFI = .995; Total CD = .194; Ethical CD = .299; Health CD = .169

Bold indicates structural pathways that could not be constrained as equal across groups. All other coefficients were constrained to be equal across groups, but the standardized solution produces estimates separately by group.

*p < .05, **p < .01, ***p < .001

First, although interactions with vegans varied across motivation, interactions with vegans shows significant and positive coefficients on identity prominence (beta = .32 and .49 for ethical and health vegans, respectively), supporting H1. Similarly, interactions with vegans shows positive coefficients on social identity recognition (beta = .42 and .38 for ethical and health vegans), though the effect could be constrained as equal across vegans. This supports H2. Interactions with vegans on dietary lapses varied across motivation; however, only the coefficient for health vegans was significant (beta = −.20). This supports H3 among health vegans but not ethical vegans. Identity prominence shows a negative coefficient on dietary lapses (beta = −.25 and −.20 for ethical and health vegans), but the effect could be constrained as equal across vegans. This supports H4. Finally, socially identity recognition on dietary lapses varied across motivation, though again, only the coefficient for health vegans was significant. That said, the coefficient is positive instead of the predicted negative association (beta = .25), failing to support H5.

The coefficients for several background variables were also significant. Age is negatively associated with interactions with vegans (beta = −.14 and −.17 for ethical and health vegans), though income shows a positive association (beta = .11 for both). Income is negatively associated with social identity recognition (beta = .11 for both), but length of time being vegan as well as identifying as a man are positively associated with social identity recognition (beta = .11 for time and beta = .09 and .12 for gender, respectively). Identifying as a man is also positively associated with dietary lapses (beta = .13 for both). Being in the social media sample is associated with greater interactions with vegans (beta = .34 and .16 for ethical and health vegans) and fewer dietary lapses (beta = −.30 and -.10, respectively). Finally, length of time being vegan is positively associated with identity prominence for health vegans only (beta = .16).

On a final note, the model provides an excellent fit for the data overall (non-significant X², RMSEA = .023, CFI = .995). This instills confidence that the relationships presented here describe the patterns exhibited in the data well. At the equation level, the model explains about 30% of the variance in dietary deviance for ethical vegans and about 17% for health vegans.

Discussion

In this paper, I tested an inclusive model of dietary lapses among a sample of 488 vegans. This model synthesized insight from the unified model of vegetarian identity (UMVI) and identity theory (IT) by organizing each of the factors identified by previous studies as important while also accounting for the relationships that exist between factors. I found several relationships that are consistent with past work. Specifically, ethical vegans reported fewer dietary lapses than health vegans (Hoffman et al., 2013; Rosenfeld, 2019; Rothgerber, 2015a, 2015b), who reported less frequent interaction with other vegans and lower identity prominence (Cherry, 2006; Plante et al., 2019). Health vegans also reported lower social identity recognition, or the view that it is important for the people with whom one interacts to acknowledge that the individual belongs to the vegan social category.

However, results from a structural equation model that separately but simultaneously estimated the model across ethical and health vegans revealed important differences. Although three out of the five predicted relationships were found as expected (with one pathway being stronger in magnitude for health vegans), different variables related to dietary deviance, and in different ways, across vegans. Interactions with vegans, identity prominence, and social identity recognition were important for health vegans, though interactions with vegans and prominence were associated with reduced dietary lapses while social identity recognition was associated with increased dietary lapses. This contrasts with ethical vegans, for whom only identity prominence was associated with reduced dietary lapses.

An important implication of these results is that the differences in eating behavior documented across types of vegans may be due to differences in identity processes and not solely due to differences in individual, interpersonal, and/or sociodemographic variables. These differences seem to hinge on the role of identity prominence versus social identity recognition. Interestingly, the finding for social identity recognition among health vegans suggests a paradoxical relationship in that the importance of having others acknowledge the individual as a member of the vegan category relates to behavior that is inconsistent with the vegan label; however, a similar finding has been reported by at least one other study (Plante et al., 2019).

Why would greater social identity recognition increase dietary lapses among health vegans and not ethical vegans? Several explanations are possible. For example, viewing one’s social identity as an important social marker likely leads individuals to use that identity as a frame for organizing interactions (see, Serpe, 1991; Stryker & Burke, 2000). In doing so, this frame may provide a sense of commonality among those sharing the label but simultaneously lead individuals to differentiate from those that do not fall within the social category (Leonardelli et al., 2010). In the vegan context, this means that high social identity recognition may influence vegans to largely define themselves in opposition to non-vegans present in social situations, thus building social distance between them and non-vegans. However, a result of social distance may be emotional strain, especially if the distance is built between family or friends (Bresnahan et al., 2016; Cherry, 2015; Hirschler, 2011; Markowski & Roxburgh, 2019). Ultimately, this strain may serve as a motivating factor to reduce distance and preserve social bonds. One way to do this is to deviate from one’s diet by sharing in the non-vegan food that non-vegans are eating (Kemmer et al., 1998; Miller et al., 1998; Morrison, 1996).

Similar scenarios have been described by recent qualitative work. Rosenfeld and Tomiyama (2019) analyzed written responses from 108 vegetarians who described the circumstances surrounding instances of diet deviations. A common finding was that individuals reported being most likely to eat meat at family gatherings, on special occasions, and to make social situations flow more smoothly. For example, many did not want to disrupt family tradition or be perceived as rude, and instead, wanted to convey appreciation for hosts’ efforts (see, also Jabs et al., 2000). Some reported feeling pressure from others to eat non-vegan food, a finding commonly reported by other studies (Cherry, 2015; Hirschler, 2011; Markowski & Roxburgh, 2019). Each of these circumstances reflects an interest in preserving social bonds and reducing social distance. Resulting dietary lapses may thus occur when high social identity recognition leads the vegan identity to be a defining feature of a situation for vegans, producing uncomfortable social distance that one desires to reduce.

As to why this might impact health vegans but not ethical vegans, the present work as well as past work finds that, compared to health vegans, ethical vegans affiliate with and thus know and interact with more vegans (Cherry, 2006). Thus, ethical vegans may be more likely to have fellow in-group members present in such situations, such as romantic partners who are also vegan (Nezlek et al., 2020; Potts & Parry, 2010). This may provide solidarity, reinforce in-group bonds, and make social distance with non-vegan out-group members feel less isolating. It would also make deviation from one’s diet more costly because one’s vegan identity and social group membership could be called into question by the fellow in-group member (Branscombe et al., 1999; Hornsey & Jetten, 2003; Rosenfeld & Burrow, 2017; Rothgerber, 2014). By contrast, if health vegans are less likely to have fellow vegans present in such situations due to having fewer vegan social contacts, social distance may feel especially severe. In such cases, dietary lapses may feel less costly, as in-group members ostensibly cannot pass judgment about behaviors to which they are not privy (see, Jabs et al., 2000).

In this study, I unfortunately lacked information about the context surrounding dietary lapses, so I was unable to confirm if lapses happened in social situations, with unsupportive family or friends, and/or in response to social processes, like the above explanations suggest. Importantly, it is also possible that lapses happened alone or isolated from others. Such situations may enable dietary lapses because neither in-group nor out-group members are present, meaning that neither membership nor bonds are at risk. In this case, the connection between dietary lapses and social identity recognition may be weaker, and in the extreme, reversed. For example, in their study, Rosenfeld and Tomiyama (2019) also examined how vegetarians rationalized their dietary deviations. One of the most common rationalizing tactics following a dietary lapse was that participants made a personal resolve to not deviate from their diet again in the future. This tactic clearly reaffirms social identity boundaries and may increase social identity recognition.

Again, as to why this might impact health vegans but not ethical vegans, the explanation may involve other identity considerations unable to be examined here. For example, IT highlights that identity meanings may vary in flexibility across individuals (Burke, 2020; Cantwell, 2016). Based on this, ethical vegans may associate strong moral meanings with their vegan identities, and these meanings may fall within a strict, narrow range (e.g., it is “never” okay to be immoral; Cherry, 2006, 2015; Greenebaum, 2012b). By contrast, health vegans may actively distance their vegan identities from moral meanings (Rosenfeld & Tomiyama, 2019), and the resulting meanings may be more flexible or fall within a wider range (e.g., it is “sometimes” okay to indulge in non-vegan food; Fox & Ward, 2008; C. T. Gallagher et al., 2021). If so, infrequent dietary lapses (alone or in social situations) may be regarded as a reasonable course of action for health vegans, increasing the likelihood that this action (and possible subsequent rationalization) will occur. Yet, the same course of action may be seen as unacceptable under any circumstance for ethical vegans, prompting them to find alternative courses of action, especially socially (e.g., making, bringing, and sharing vegan food with others) (Cherry, 2015; Greenebaum, 2012a; Paxman, 2021).

Another possibility exists. In their study, Plante et al. (2019) examined social identity as a motivation for veg*anism alongside ethical, health, environmental, and religious reasons. According to the authors, social identity motivations involve individuals being ve*gan because they like fitting in and being part of the veg*an social category. Consistent with the main finding presented here, Plante et al. (2019) found that social identity motivations positively predicted dietary lapses. They posited that individuals may derive self-esteem by affiliating with the “vegan” social category and that individuals with flexible meanings may see it as especially beneficial for their self-image to do so (see, Yeh, 2014). This may also explain the findings seen here, but whether these differences fall along the ethical versus health distinction remains an open question.

Limitations

Several limitations to this study exist. Although this study used a unique dataset with many vegans, an even greater sample size would have allowed for more in-depth analyses involving more complex intersections, such as primary motivation by gender, the latter being a major characteristic associated with meat consumption (or lack thereof; Adams, 1990; Greenebaum & Dexter, 2018; Mycek, 2018; Rosenfeld, 2020; Rothgerber, 2013). A larger sample would also likely improve the diversity of vegans included. For example, although this sample included vegans that do vary substantially across variables like length of time being vegan and identity properties, the sample is notably white and majority women. It is important to ensure that a heterogeneous range of vegans are included in studies on veganism to disrupt the marginalization of vegans of color (Greenebaum, 2018). The analysis also exclusively examined those citing ethics or health as their primary motivation for being vegan since the sample included so few vegans citing other motivations.

Another limitation is that the data are cross-sectional, meaning that I was unable to determine causality, or how eating behavior and identity processes unfold for vegans over time. Given the likely reciprocal relationship between social identity recognition and dietary lapses, longitudinal data would be a distinct advantage for research in this area. One possible way to collect this data while assessing social context is to employ ecological momentary assessment (Rosenfeld & Burrow, 2017). This strategy allows for the capture of real-time information as it unfolds in individuals’ real-life environments, meaning that important contextual information can be captured and assessed in relation to behavior (Shiffman et al., 2008).

Conclusion

This study contributes to the literature in several ways. First, this study provides a more comprehensive account of vegans’ dietary lapses across motivation by simultaneously examining the range of factors identified by past work as separately related to dietary lapses among vegans. My analysis showed the pivotal role of identity processes, controlling for a variety of other factors. This asserts the importance of examining veganism as a social identity, especially when examining eating patterns. Second, this study sheds light on how identity processes may vary across vegans with different primary motivations. This contributes to the discussion on patterns in dietary lapses documented by recent work, providing several possibilities that can be used to structure future work in the area. Last, this study bridges together work on identity, including social identity approaches and the UMVI as well as IT. This is important because it provides theoretical structure to the UMVI while also continuing to expand IT to better examine social identities (see, Davis et al., 2019). The result is an inclusive framework on how individual, interpersonal, and structural elements of social life combine to influence dietary behavior among vegans, a framework that other scholars may find fruitful when examining other food-based social identities.

Open Scholarship

This article has earned the Center for Open Science badges for Open Data and Open Materials through Open Practices Disclosure. The data and materials are openly accessible at https://doi.org/10.7910/DVN/DFBEOM..

Acknowledgement

The author would like to thank the Graduate Student Senate and University Research Council at Kent State University for the support that made this work possible.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data described in this article are openly available in the Open Science Framework at https://doi.org/10.7910/DVN/DFBEOM

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Additional information

Funding

The research was funded by the Graduate Student Senate and University Research Council at Kent State University.

Notes on contributors

Kelly L. Markowski

Kelly L. Markowski is a Senior Researcher in the Government Resource Center (GRC) at the Ohio State University, where her work evaluates central aspects of the Ohio Department of Medicaid's Next Generation of Managed Care program. She also holds affiliate status in the Rural Drug Addiction Research (RDAR) center at the University of Nebraska-Lincoln. Her independent research spans identity theory, social networks, veganism, and substance use.

Notes

1. This term is used to describe vegetarianism as well as veganism.

2. This definition is identical to the UMVI’s definition of centrality. However, since recent identity theory work outlines the distinction between this shared definition and definitions of “centrality” in other social psychological work (Markowski & Serpe, 2021), this idea is hereafter referred to as identity prominence.

3. Among these 98 people, primary reason for being vegan was varied, as 23 indicated environmental reasons, 6 indicated because of world hunger, 3 indicated political reasons, 10 indicated religious reasons, 6 indicated because of family or friends, 2 indicated in order to save money, 31 indicated taste preferences, 11 were born and raised vegan, 4 selected “Other,” and 2 refused to answer.

4. This pathway yielded a marginally significant difference (p = .07) across ethical and health vegans. Because this difference approaches conventional levels of significance and because the goodness of fit slightly decreases when constrained as equal, I retain separation of this pathway across vegans in the analysis.

5. Four error covariances were estimated but are not presented here. Results are available upon request.