Abstract
Marketers frequently style food to look pretty (e.g., in advertising). This article investigates how pretty aesthetics (defined by classical aesthetic principles, such as order, symmetry, and balance) influence healthiness judgments. The author proposes that prettier food is perceived as healthier, specifically because classical aesthetic features make it appear more natural. In a pilot, six main studies and four supplemental studies (total N = 4,301) across unhealthy and healthy, processed and unprocessed, and photographed and real foods alike, people judged prettier versions of the same food as healthier (e.g., more nutrients, less fat), despite equal perceived price. Even given financial stakes, people were misled by prettiness. In line with the proposed naturalness process, perceived naturalness mediated the effect; belief in a “natural = healthy” connection moderated it; expressive aesthetics, which do not evoke naturalness, did not produce the effect (despite being pretty); and reminders of artificial modification, which suppress perceived naturalness, mitigated it. Given that pretty food styling can harm consumers by misleading healthiness judgments for unhealthy foods, managers and policy makers should consider modification disclaimers as a tool to mitigate the “pretty = healthy” bias.
People view an estimated 4,013 food and 2,844 restaurant advertisements per year—that comes out to more than 18 food-related advertisements per day (Statista 2016). In these ads, they typically see foods that are extensively styled to look (unrealistically) pretty. Just imagine the beautiful pizza, with its picture-perfect bubbly crust, flawlessly allocated pepperoni, artfully scattered mushrooms, and glistening pillows of melted cheese, that you see every day driving by one of many billboards. The advertiser’s hope is, of course, that prettier foods will seem more appetizing and abundant. Indeed, neuroscience research suggests that viewing delectable food images activates the gustatory cortex, essentially simulating the food’s pleasurable taste (Simmons, Martin, and Barsalou 2005). But might pretty aesthetics have other, less obvious, and potentially perverse effects on food evaluations? Might the alluringly good-looking pizza actually seem healthier to you, by virtue of its aesthetics? This research examines whether and how prettiness influences judgments of the food’s healthiness.
Most people strive to eat healthily (International Food Information Council Foundation 2015), but it is notoriously difficult to accurately judge food’s healthiness, because nutrients and energy are invisible. Accordingly, consumers’ healthiness judgments are often shaped by arbitrary external factors, such as category membership (Rozin, Markwith, and Ashmore 1996), healthy descriptors (Irmak, Vallen, and Robinson 2011), and labels (Schuldt 2013). No research has examined how the prettiness of food affects healthiness judgments, perhaps because its direction and mechanism are not obvious. On the one hand, beautiful aesthetics may elicit notions of pleasure and hedonics. Activation of these concepts may lead people to view a food as more indulgent, causing them to judge prettier looking foods as unhealthier. On the other hand, beautiful aesthetics may elicit spontaneous inferences of naturalness which, in turn, may signal healthiness. I propose the latter, grounding this proposition in the link between classical aesthetic principles and patterns found in nature.
Across six experiments that use various different manipulations of prettiness, this research finds that consumers perceive prettier foods as more natural than less pretty foods, which causes them to infer that prettier foods are healthier. In line with the naturalness hypothesis, people exhibit strong implicit associations between prettiness and naturalness and between naturalness and healthiness. Furthermore, expressive aesthetics, which highlight human intent, do not produce the effect (because they do not evoke naturalness), and disclaimers reminding people that food has been artificially styled mitigate the effect (because they suppress naturalness). Across studies, prettiness only boosts inferences about select attributes, suggesting that its effect on perceived healthiness is not due to a general halo. Prettiness also enhances perceived healthiness regardless of perceived tastiness, suggesting the effect cannot be attributed to “wishful thinking.” Finally, other inferences that may covary with prettiness, such as freshness, care, and sophistication, do not better explain the effect, suggesting a distinct role of naturalness.
The findings have clear practical relevance. Evaluations of food’s healthiness have important downstream consequences, including food choice (Nikolova and Inman 2015) and serving size decisions (Suher, Raghunathan, and Hoyer 2016). Given both the prominence of food visuals and marketers’ ability to maximize food prettiness in myriad digital and nondigital ways, understanding how aesthetics guide eating decisions is key to promoting healthy choices. For instance, the results point to unnaturalness cues (e.g., reminders about processing) as an intervention to attenuate prettiness-based overestimation of healthiness. Furthermore, the finding that pretty aesthetics can boost perceived healthiness while enhancing (or preserving) perceived tastiness also suggests a way to sidestep the perceived health–taste trade-off that may deter people from healthy eating (Raghunathan, Naylor, and Hoyer 2006).
The insights also make important theoretical contributions. First, the findings contribute to the literature on lay beliefs by revealing a novel lay intuition: people take classical aesthetics as a signal of naturalness (e.g., being pure, unaltered). Beyond documenting this “pretty = natural” intuition, they also extend the finding that people believe natural entities to be categorically healthy (Rozin et al. 2004) by including objective measures of healthiness (e.g., calories, fat, nutrients).
Second, this article addresses the call for more process research in consumer aesthetics (Patrick and Peracchio 2010): an investigation of different drivers of the effect of pretty aesthetics on judgments offers consistent evidence for the role of naturalness. This finding suggests that classical aesthetics may lead to a host of unintended consequences known to spring from naturalness judgments and opens up novel avenues for aesthetics research, such as examining the impact of aesthetics on perceived efficacy. Finally, this finding has implications for any work using visual food stimuli, such as that on food choice, health communication, or vice–virtue conflict.
Relationship Between Food Aesthetics and Healthiness Judgments
I define “perceived healthiness” as a composite of being perceived as high in nutrients, low in fat, and low in calories, in addition to being viewed as good for your body and globally healthy, in line with current dietary guidelines and lay understanding of healthiness (International Food Information Council Foundation 2015; United States Department of Agriculture 2015). Because food is complex and the qualities that contribute to its nutritional value are intangible, people rely heavily on a variety of contextual cues to guide their healthiness judgments. For example, mere membership in stereotypically healthy categories enhances perceived healthiness (Rozin, Markwith, and Ashmore 1996). Similarly, healthy-sounding food names (Irmak, Vallen, and Robinson 2011) and healthy branding of restaurants (Chandon and Wansink 2007) boost the perceived healthiness of food offerings.
In terms of aesthetic factors and healthiness perceptions, existing research has almost exclusively studied isolated aspects, such as the color of labels (Schuldt 2013) or the shine of packaging (Ye, Morrin, and Kampfer 2020). Whether and how holistic food aesthetics (i.e., an overall beautiful, pleasing appearance) shape healthiness perceptions is unknown—a surprising gap, given the prevalence of food styling. Research on the holistic aesthetics of food is scarce in general, and the studies that have examined overall prettiness focus on its effect on food appeal. For example, in one experiment, people liked a food’s taste better when it was visually more appealing (Zellner et al. 2014). In another, diners expressed greater liking and willingness to pay (WTP) prior to eating when a meal was plated more artistically (Michel et al. 2015). In the field, too, consumers tend to avoid ugly-looking food (e.g., De Hooge et al. 2017), which has been attributed to ugly food undermining consumers’ self-esteem (Grewal et al. 2019).
This research aims to fill this gap. I specifically focus on prettiness in terms of classical aesthetics—one of two distinct dimensions of human aesthetic perception (Lavie and Tractinsky 2004). Classical aesthetics are defined by traditional principles of beauty, such as order, symmetry, balance, clarity, and pattern repetition (see Johnson 1994). I suggest that classical aesthetics signal naturalness, because they resemble patterns found in nature, and thereby enhance perceived healthiness. In contrast, expressive aesthetics are marked by the originality and creativity of a designer (Lavie and Tractinsky 2004) and thus are not expected to exert the same effect. Practitioners undoubtedly use aesthetics to make food appear more palatable, but prior research has generated multiple possible predictions regarding the unintended and counterintuitive consequences pretty aesthetics may have for perceived healthiness.
Common Assumption: Prettiness Embodies Pleasure and Indulgence
On the one hand, people may intuit that prettier (vs. less pretty) food must be unhealthy because prettiness is associated with hedonic pleasure. Beautiful sights, by definition, elicit pleasure. Seeing beautiful designs directly activates the brain’s reward center (Reimann et al. 2010), and a primary function of visual art is aesthetic gratification (Hagtvedt and Patrick 2008). The notion of pleasure that is inherent in beautiful visual aesthetics may signal indulgence and, in turn, unhealthiness.
People tend to view hedonic attributes as incompatible with instrumental or moral attributes. This principle is reflected in research on self-control dilemmas and vice–virtue conflicts, which readily pits pleasurable options against sensible ones (Okada 2005; Shiv and Fedorikhin 1999). Some research suggests that the idea of a pleasure–merit trade-off may be especially salient in the domain of food: people appear to believe that unhealthy food provides greater pleasure, assuming that it is more satisfying (Finkelstein and Fishbach 2010; Suher, Raghunathan, and Hoyer 2016) and tastier (Raghunathan, Naylor, and Hoyer 2006). Thus, because pretty images are, by definition, pleasurable, prettier (vs. less pretty) versions of the same food may be perceived as less healthy.
Proposition: Prettiness Reflects Natural Patterns and Thereby Signals Healthiness
On the other hand, people may perceive prettier (vs. less pretty) food as healthier, given that prettier aesthetics often prompt more favorable evaluations on other, unrelated dimensions. For instance, prettier machines, such as ATMs, seem more usable (Tractinsky, Shoval-Katz, and Ikar 2000; for a review, see Hassenzahl and Monk [2010]); prettier financial reports can elicit higher stock valuation (Townsend and Shu 2010); and even for mundane, machine-made disposable products, such as napkins, people believe prettier versions result from greater effort (Wu et al. 2017).
An intuitive explanation for this pattern is a generalized halo effect (Nisbett and Wilson 1977; Thorndike 1920)—a process whereby an initial global positive affective reaction (elicited, e.g., by prettiness) spills over onto the judgment of other, entirely unrelated attributes. Applying this logic here, prettier (vs. less pretty) food may be evaluated more positively on all attributes, simply because people feel more positively about it on account of its pleasant looks. However, I suggest that pretty aesthetics provoke distinct inferences about specific attributes of food, which then give rise to perceived healthiness.
Specifically, I propose that prettiness in the form of classical aesthetics (i.e., order, symmetry, balance, and pattern repetition) induces the notion of naturalness (i.e., being natural, pure, and unprocessed: Rozin 2005), which in turn signals healthiness. Various disciplines have found that patterns that appear in nature are considered beautiful (for a review, see Palmer, Schloss, and Sammartino [2013]). For instance, approximate symmetry is both extremely prevalent in nature—be it the bilateral symmetry in the vast majority of modern animals (up to 99%; Finnerty 2005) or the radial, bilateral, or dissymmetry of most flowers (>90%; Neal, Dafni, and Giurfa 1998)—and considered beautiful (e.g., Jacobson and Höfel 2002; Zheng et al. 2009). Likewise, order and regularity are highly characteristic of nature, where they manifest in the patterns of self-similarity seen in trees, ferns, and other plants as well as coast and mountain profiles (Mandelbrot 1982) and in the pattern repetition exhibited by most animals that communicate through visual appearance (see Kenward et al. 2004). They, too, are considered beautiful (Spehar et al. 2003). Finally, balanced proportions between different elements or dimensions are common in nature. A prominent example are Fibonacci proportions, which give rise to the “golden spiral” found in most plants’ leaf arrangement (Mitchison 1977). Once again, balanced proportions are perceived as beautiful (e.g., Fechner 1871; Green 1995). All of these phenomena are interrelated: pattern repetition produces some order and often gives rise to symmetry; proportionality can induce self-similarity (as it does in logarithmic spirals); and symmetry generates some balance (reflected in the word “symmetry” itself, meaning “well-proportioned, well-ordered”).
In summary, it seems viable that classical aesthetics may elicit the concept of naturalness, and accordingly, I propose that prettier (vs. less pretty) food may be perceived as healthier because it seems more natural. Being natural or unprocessed does not necessarily imply anything about the fat, sugar, or calorie content of a food, but consumers appear to overextend the term to healthiness (Rozin et al. 2004).
Finally, note that the nature-like features of classical aesthetics are not the only way for objects to look pretty. Expressive aesthetics (e.g., creativity, originality), for example, refer to the abstract depiction of some imaginative idea—both decidedly human capacities—as opposed to features found in nature. Cubism or abstract expressionism are popular examples of expressive aesthetics. In line with the theorizing that prettiness enhances perceived healthiness by way of naturalness, not all prettiness should produce the effect. I propose that the effect of prettiness on perceived healthiness is elicited by classical aesthetics, which signal naturalness, but not by other forms (e.g., expressive aesthetics) that do not signal naturalness—despite being pretty.
Other Possible Effects of Prettiness
In addition to the proposed account, the prettiness of a food may inevitably bring to mind any number of other attributes beyond naturalness, and some of these attributes may also be tied to healthiness. For example, perhaps pretty food seems fresher or safer, and people may overgeneralize these qualities to healthiness in terms of calories and nutrients. Pretty food presentation may also signal food expertise or general refinement. Sophistication is linked with the idea of “upper class” (Aaker 1997), which people may associate with a healthier lifestyle, given that higher socioeconomic status correlates with lower body mass index in developed economies (McLaren 2007); mass media portrays success as tantamount to being slim and fit (Evans 2003); and consumers expect expensive foods to be healthier (Haws, Reczek, and Sample 2017). Relatedly, to the extent that pretty food presentation suggests diligent preparation, people may believe that a more caring agent made it. People stereotype others who choose healthy food as more conscientious (e.g., responsible, disciplined, caring; Oakes and Slotterback 2004) and may conversely assume conscientious food presentation is associated with particularly judicious, and healthy, ingredient choices. Neatness is also associated with following prescriptive norms (Vohs, Redden, and Rahinel 2013), which may likewise signal sensible food choices.
Across studies, I examine these possibilities as well as more general alternative mechanisms such as a generalized halo effect and motivated reasoning, but none appear to make compelling alternative explanations. I find robust evidence for a distinct effect of naturalness.
Similarly, because I study the effect of visual aesthetics of complex objects (i.e., food) as opposed to isolated elements (e.g., color swatches, dot patterns), the prettier and less pretty images will typically vary on more than one visual aspect. I have taken several precautions to alleviate a concern that the effect is idiosyncratic to a particular pair of images.
First, a pilot study employs stimulus sampling using images not selected by me. Second, two studies use nonvisual manipulations. In one study, all participants evaluate the identical photograph and perceived prettiness is manipulated only by way of their expectations to see pretty or ugly food, exploiting an assimilation effect. Another study documents the key associations in an Implicit Association Test using only verbal target stimuli. Third, the studies that do manipulate prettiness visually use various different images (or live food) and yet other stimuli in supplemental studies reported in the Web Appendix. Finally, I measure healthiness as nutrient, fat, and calorie content—variables that should be unaffected by many of the possible differences in the images.
Pilot Study: Suggestive Evidence of the Pretty = Healthy Link
A pilot study (see Web Appendix A) explored whether prettier food is indeed perceived as healthier. Eight hundred three participants were asked to search online for an image of a “pretty” or an “ugly” version of a food (ice cream sundae, burger, pizza, sandwiches, lasagna, omelet, or salad), upload the image, and judge the extent to which the food pictured was “healthy,” “nutritious,” “good for me,” “fatty,” and “high in calories.” The latter two items were reverse-coded and all five items were combined into a “perceived healthiness” scale (α = .89).
As predicted, participants perceived the food as healthier when it was pretty (M = 3.74, SD = 1.60) than when it was ugly (M = 2.71, SD = 1.19; F(1, 787) = 209.41, p < .001, d = .73; see Table 1; for means by food, see Web Appendix A). Gender did not interact with prettiness on any key variable in this or any other study and is thus not discussed further.
Results of All Main Text Studies.
Notes: Results are means (SD) unless otherwise indicated. For Studies 4a and 4b, means in the same row with different superscripts are significantly different from each other at p < .05; means that share the † symbol are significantly different from each other at the p < .1 level; means that share the same superscript letter do not differ from each other.
Of course, the images could have varied in many ways, but this finding provides initial impetus for subsequent experimentation, suggesting a link between prettiness and healthiness with naturalistic stimuli. Study 1a moves to a controlled setting to test the idea that pretty (vs. ugly) food is perceived as healthier.
Study 1a: Prettiness Enhances Perceived Healthiness via Perceived Naturalness
Study 1a manipulates the prettiness of an otherwise identical food by varying the presence of pattern repetition in its presentation but holds constant the setting and photographic quality. Furthermore, to ensure that everyone is equally aware of all food components, the stimuli contain only one salient ingredient, a list of ingredients is provided, and the food name is displayed with the picture.
Method
Four hundred Prolific Academic panelists in the United States (54.0% women; Mage = 35.46 years, range = 18–77 years) were randomly assigned to evaluate either a pretty or an ugly version of an avocado toast (made from the identical ingredients; see Figure 1).
Stimuli used in Studies 1–3.
As theories of classical aesthetics predict, in a pretest (see Web Appendix B), people found the avocado toast significantly prettier when it featured pattern repetition and order (M = 5.24, SD = 1.45) than when it lacked pattern and thus presented no salient order (M = 3.39, SD = 1.58; p < .001).
Participants learned that the researchers were interested in opinions about food products and that they would view a food image and rate it on various dimensions. Before viewing the food, participants read that they would see an “avocado toast, made from 1 slice of wheat bread and 1/2 of an avocado.” The ingredient text was boldfaced, and participants were required to stay on the page for 10 seconds to ensure that they would not skip over these important details.
In addition, on each of the subsequent pages, “Avocado Toast” was displayed under the image alongside a cost of “∼$2.” Participants provided all subsequent ratings on seven-point scales anchored at 1 = “not at all” and 7 = “very much.”
Perceived healthiness
First, participants rated the avocado toast on the perceived healthiness scale (α = .79; note that this and all subsequent studies use “low in fat” and “low in calories” instead of the reverse-coded items used in the pilot). Then perceived naturalness and perceived tastiness were measured, in randomized order.
Perceived naturalness
To capture the proposed driver of the effect, perceived naturalness, participants rated the extent to which they thought the avocado toast was “natural,” “pure,” and “unprocessed.” The items were combined into a “perceived naturalness” scale (α = .87). (Note that principal component analysis showed that naturalness and healthiness are statistically distinct constructs, suggesting discriminant validity; see Web Appendix C.)
Equality check for perceived tastiness
Food styling is done expressly to make the food look pretty and appetizing. Research has shown that consumers believe that unhealthy food is tastier (Raghunathan, Naylor, and Hoyer 2006), so if they also believe the reverse, then styled, pretty food may appear less healthy merely on the basis of this lay theory. For the ideal test, tastiness should be equal across conditions. Participants rated to what extent they thought the avocado toast was “tasty,” “flavorful,” and “delicious.” These items were combined into a “perceived tastiness” scale (α = .97).
Equality check for perceived price
Although equal cost information was provided, to rule out any effects of prettiness on perceived healthiness being due to differences in perceived price (Haws, Reczek, and Sample 2017), participants rated the extent to which they thought the avocado toast was “pricey” and “expensive.” These items were combined into a “perceived price” composite (r = .92, p < .001).
Finally, participants reported demographics, which included education level and household income. Neither interacted with prettiness on perceived healthiness or naturalness, so these variables are not discussed further (although details and results are available upon request).
Attention check
Finally, participants responded to an open-ended question asking what the most prominent component of the food was to ensure all participants were equally aware that the toast topping was avocado. In this free recall, 98.75% (395 out of 400) answered “avocado;” the five nonavocado responses were distributed equally across conditions (χ2(1) = .21, p = .645).
Results
Table 1 displays the means for all dependent variables collected.
Basic effect: Perceived healthiness
Participants rated the avocado toast as significantly healthier when it was pretty (M = 5.03, SD = .98) than when it was ugly (M = 4.70, SD = 1.07; F(1, 398) = 10.04, p = .002, d = .32).
Proposed mediator: Perceived naturalness
Participants rated the avocado toast as significantly more natural when it was pretty (M = 5.39, SD = 1.12) than when it was ugly (M = 4.95, SD = 1.37; F(1, 398) = 12.33, p < .001, d = .35).
Equality checks: Perceived tastiness and perceived price
As intended, the two avocado toasts were perceived as virtually equal in terms of tastiness (F(1, 398) = 2.07, p = .151) and perceived price (F(1, 398) = .524, p = .470).
Mediation
Simple mediation (model 4; 10,000 samples) indicated that perceived naturalness mediated the effect of prettiness on perceived healthiness (B = .084, SE = .026; 95% confidence interval [CI] = [.036, .137]; robust to controlling for tastiness and price).
Discussion and Replication Study
People thought the same portion of food was healthier when it was pretty than when it was ugly (despite being given a list of exact ingredients), and this effect was driven by perceived naturalness. Tastiness, in contrast, was unaffected by prettiness.
Replication
I also conducted a conceptual replication (see Web Appendix D) to ensure the generality of the effect and test several alternative explanations. Manipulating prettiness via order and symmetry of an almond butter and banana toast (AB&B), it replicated the naturalness-mediated effect of prettiness on perceived healthiness as well as on calorie estimates but found no effect of prettiness on perceived tastiness, freshness, amount, or price. Dieting did not moderate the effect.
Together, Study 1a and its replication rule out tastiness, freshness, and amount as alternative explanations. It is undoubtedly possible that any of these variables may, for certain stimuli, exert an additional positive effect on perceived healthiness beyond that of perceived naturalness, but the results attest that they do not supersede, nor are they necessary for, the mediating effect of naturalness. In addition to ruling out these specific alternative mechanisms, the lack of an effect of prettiness on other important attributes (e.g., tastiness, freshness) also challenges a general halo effect as a generic alternative explanation. To demonstrate robustness and eliminate the possibility that any real differences between pretty and ugly images drive the pretty = healthy relationship, Study 1b tests the effect using a nonvisual manipulation of prettiness.
Study 1b: Prettiness Enhances Perceived Healthiness via Perceived Naturalness
In the pilot study and Study 1a, participants evaluated pretty and ugly images that actually differed from each other, which may raise concerns about other attributes that may have covaried with prettiness. To circumvent this issue, Study 1b uses the identical image in both the pretty and the ugly condition and manipulates perceived prettiness of the same image solely by exploiting a biased assimilation effect, whereby the expectation of encountering a certain type of stimulus biases the subsequent perception of the stimulus in the direction of the expectation (Sherif, Taub, and Hovland 1958; for a review of biased assimilation, see Lord and Taylor [2009]). This type of manipulation eliminates all other objective differences between the foods and varies prettiness in the cleanest fashion.
Study 1b also includes foods at three different general healthiness levels. The proposed naturalness account does not make direct predictions about any moderating effect of baseline (un)healthiness. Yet it is possible that very unhealthy categories pose a boundary condition. For example, unhealthy food categories may involve more obviously manipulated components (e.g., processed meats, compounded ingredients such as sauces), limiting the extent to which the foods can be construed as natural. Understanding the scope of the effect and its potential for harm are practically important to assess the need for regulation.
Method
Eight hundred one Prolific Academic panelists in the United States (49.90% women; Mage = 32.70 years, range = 18–76 years) were randomly assigned to evaluate food images that they expected to be either pretty or ugly and then, having established this expectation, rated three food replicates, creating a 2 (prettiness: pretty vs. ugly) × 3 (food replicate: AB&B toast, spaghetti marinara, cupcake) mixed design. Three different foods were included to test whether the effect of prettiness varies systematically across general healthiness levels, and a within-subjects design and large sample size were used to offset the rather subtle manipulation.
Participants learned that the researchers were interested in opinions about food products and that they would view three food images and rate them on various dimensions. To hold the actual food image constant and only change perceived prettiness of the exact same image, I manipulated participants’ expectations about whether they would see pretty aesthetic features or ugly aesthetic features by providing the following instructions at the beginning of the study: This study is about PRETTY [UGLY] FOOD. The food in the image we will be showing you will be very pretty [ugly] (based on ratings from previous MTurk participants and food design professionals). The foods will be orderly [disorderly], they will look symmetrical [lopsided], and the proportions will be balanced [unbalanced].
Then, all participants evaluated identical images of an AB&B toast, a plate of spaghetti marinara, and a cupcake, presented in randomized order (see Figure 1). Because moderate or ambiguous stimuli are most amenable to manipulation via expectations (see Herr 1989; Sumer and Knight 1996), I selected images that had been rated near the midpoint of the prettiness scale in pretests. Again, the food’s name and cost were shown under each image.
For each food, participants first rated the perceived healthiness scale (αAB&B = .83; αspaghetti = .86; αcupcake = .91) and then the perceived naturalness scale (αs = .89/.90/.87). (Note that principal component analyses showed that naturalness and healthiness are statistically distinct constructs, suggesting discriminant validity; see Web Appendix E.)
Manipulation check for perceived prettiness
To assess the effectiveness of the manipulation, participants rated to what extent they thought the food was “beautiful,” “pretty,” and “good looking.” The items were combined into a “perceived prettiness” scale (αs = .93/.93/.94). Finally, participants completed an attention check and reported demographics.
Results
Table 1 displays the means for all dependent variables collected. I conducted repeated-measures analyses of variance with prettiness as the between-subjects factor and food replicate as the within-subject factor. Prettiness did not interact with food replicate on any of the dependent variables (prettiness: p = .146; healthiness: p = .382; naturalness: p = .368), so I collapsed across the three replicates (for results by food, see Web Appendix E).
Manipulation check: Perceived prettiness
As theories of classical aesthetics and biased assimilation predict, participants found identical food images significantly prettier when they expected to see orderly, symmetrical, and balanced food presentation (M = 4.19, SD = 1.20) than when they expected to see disorderly, lopsided, and unbalanced food presentation (M = 3.66, SD = 1.18; F(1, 799) = 40.29, p < .001, d = .45).
Basic effect: Perceived healthiness
Participants rated the identical food images as significantly healthier when they perceived them as pretty (M = 3.56, SD = .73) than when they perceived them as ugly (M = 3.39, SD = .69; F(1, 799) = 11.96, p = .001, d = .24).
Proposed mediator: Perceived naturalness
Participants rated the identical food images as significantly more natural when they perceived them as pretty (M = 3.60, SD = .92) than when they perceived them as ugly (M = 3.35, SD = .86; F(1, 799) = 16.38, p < .001, d = .28).
Mediation
Because the independent variable of interest, prettiness, was manipulated between-subjects and the lack of an interaction between prettiness and food replicate allowed me to collapse across replicates, I conducted regular mediation. Simple mediation (model 4; 10,000 samples) indicated that perceived naturalness mediated the effect of prettiness on perceived healthiness (B = .07, SE = .017; 95% CI = [.033, .100]). Results hold when conducting a separate mediation for each replicate.
Discussion
People judged identical foods as healthier when they looked pretty (vs. ugly) to them, because they perceived them as more natural, thus replicating the results from Study 1a. The effect of prettiness on perceived naturalness and healthiness was not moderated by the food’s general healthiness level. That is, the bias extends to those foods for which it is the most problematic.
Importantly, the effect was produced purely by focusing people on order, symmetry, and balance (vs. disorder, asymmetry, and imbalance) to manipulate perceived prettiness, without any changes in the food picture itself—an especially conservative manipulation. This finding suggests that the effect of prettiness on perceived healthiness does not rely on true differences between the food images but rather on people perceiving the food as featuring classical aesthetics, isolating the psychological underpinnings of the effect beyond any image differences.
While this nonvisual manipulation of prettiness enables a clean test of both the key effect and the proposed mechanism, it is also a fairly weak prettiness manipulation (manipulation check d = .45). To test the pretty = natural = healthy framework more effectively, in subsequent studies I return to manipulating prettiness via real visual differences in classical aesthetics between images while holding constant or controlling for other relevant food features. Studies 2 and 3 explore behavioral consequences of the pretty = healthy effect.
Study 2: Prettiness Increases Willingness to Pay for Real Food via Perceived Healthiness
Study 2 investigates the effect of prettiness-induced perceived healthiness on a consequential behavior: WTP. Studies 1a–b and the replication held price constant, as people may infer healthiness from market price (Haws, Reczek, and Sample 2017). Yet it is possible that people value foods they perceive as healthier more highly, irrespective of price (for the price–value distinction in the aesthetics domain, see Sevilla and Townsend [2016]). The goal of this study is to quantify the extent to which the shift in healthiness judgments contributes to shifts in WTP, not to demonstrate the general phenomenon that prettiness enhances WTP (e.g., Grewal et al. 2019). Study 2 manipulates prettiness via symmetry and balance of a whole food. It also uses the most unequivocally healthy food (i.e., produce) and ensures that the effect is robust when people interact with real food instead of photographs.
Method
Volunteers were recruited for a product evaluation study at University of Southern California in conjunction with “move-in day” over a period of three days (nine hours) for $1 in cash. Eighty-nine student and nonstudent passersby (57.3% women; Mage = 23.61 years, range = 17–52 years) were randomly assigned to evaluate either a pretty or an ugly bell pepper (see Figure 1).
Participants completed the study behind a trifold privacy screen, where a white tray with the ugly or pretty produce had been placed according to condition. They were asked to visually examine the product without touching it. Then they rated, in randomized order, an abbreviated perceived healthiness scale (α = .91; using only “healthy,” “nutritious,” and “good for me,” to manage time constraints) and the perceived tastiness scale (α = .93).
Willingness to pay
I elicited WTP using a version of the BDM procedure (Becker, DeGroot, and Marschak 1964). Participants bid some amount of their $1 on the pepper in front of them, anticipating that a die roll would determine its selling price and that if they bid more or equal to the selling price, they would buy the pepper and receive the rest of their $1 in change, but if they bid below the selling price, they would not buy it and instead receive the whole $1 in cash. They chose a bid between $0 and $1 (in $.10 increments) and displayed this reservation price to the researcher, who then rolled a ten-sided die to determine the selling price.
Manipulation check for perceived prettiness
Before receiving their cash and/or (a different) pepper, participants rated the perceived prettiness scale (α = .97). Finally, they reported demographics, including income. Income had no main effect and did not interact with prettiness on perceived healthiness or naturalness, so this variable is not discussed further (although details and results are available upon request).
Results
Table 1 displays the means (or medians) for all dependent variables collected.
Manipulation check: Perceived prettiness
As theories of classical aesthetics predict, participants rated the pepper as significantly prettier when it was symmetrical and balanced (M = 5.42, SD = 1.31) than when it was asymmetrical and unbalanced (M = 3.20, SD = 1.86; F(1, 87) = 43.49, p < .001, d = 1.37).
Basic effect: Perceived healthiness
Participants rated the pepper as significantly healthier when it was pretty (M = 6.03, SD = 1.02) than when it was ugly (M = 5.08, SD = 1.49; F(1, 87) = 12.96, p = .001, d = .74).
Perceived tastiness
Participants rated the pepper as tastier when it was pretty (M = 5.44, SD = 1.22) than when it was ugly (M = 4.26, SD = 1.68; F(1, 87) = 14.64, p < .001, d = .80).
Willingness to pay
Willingness to pay in the ugly condition was right-skewed (Shapiro–Wilk W (40) = .869, p < .001; for the distribution, see Web Appendix F), so I applied a Mann–Whitney U test. People bid significantly more real money to buy the pepper when it was pretty (Mdn = $.50; M = $.47, SD = .27) than when it was ugly (Mdn = $.20; M = $.30, SD = .29; U = 1,365, z = 3.20, p = .001). Results remain unchanged when applying a parametric test (F(1, 87) = 8.96, p = .004, d = .61).
Mediation
I tested several mediation models. Separate simple mediations (model 4; 10,000 samples) indicated that perceived healthiness mediated the effect of prettiness on WTP (B = .022, SE = .012; 95% CI = [.004, .051]), as did perceived tastiness (B = .031, SE = .014; 95% CI = [.009, .063]). Serial mediation with both healthiness and tastiness revealed that a healthiness–tastiness sequence mediated the effect on WTP (B = .012, SE = .008; 95% CI = [.001, .030]), but the reversed tastiness–healthiness sequence did not (B = .005, SE = .067; 95% CI = [−.005, .021]).
Discussion and Replication Study
Again, people thought food was healthier when it was pretty (vs. ugly), and this inference raised their willingness to pay real money for the item. Although WTP is undoubtedly multiply determined, perceived healthiness contributed significantly to the increase in valuation.
Replication
I also conducted a conceptual replication online with photographs of peppers and apples (see Web Appendix G). This replication also measured perceived naturalness and found that a naturalness–healthiness sequence serially mediated the effect of prettiness on WTP. It also replicated the finding that healthiness appears to give rise to tastiness: a naturalness–healthiness–tastiness sequence serially mediated the effect of prettiness on WTP, but reversed sequences (i.e., naturalness–tastiness–healthiness and tastiness–naturalness–healthiness) did not.
Study 2 and its replication show that prettiness boosts WTP by increasing perceived healthiness. To demonstrate the effect with a different type of consequential behavior, Study 3 tests whether the prettiness-induced differences in perceived healthiness are powerful enough to shift people’s choices even when they are motivated to choose healthy food.
Study 3: Prettiness Influences Healthiness Judgments Even When Accuracy Is Incentivized
Study 2 showed that greater perceived healthiness boosts WTP, but an arguably equally important downstream consequence is choice. People who try to identify healthy options may be misled by unhealthy options (believing them to be the healthy choice) simply by virtue of those unhealthy options looking pretty. Study 3 tests whether prettiness biases choice even when people are explicitly incentivized to make a healthy selection.
Method
Three hundred Prolific Academic panelists in the United States (53.3% women; Mage = 30.80 years, range = 18–78 years) were asked to identify which of two foods had fewer calories. They were randomly assigned to a choice set that contained either a pretty version or an ugly version of a target food in addition to a reference food, creating a one-factor (target food prettiness: pretty vs. ugly) design. The task was consequential, with a bonus given for the correct choice.
Participants learned that the researchers were interested in how people think about nutrition, that they would be asked to identify the lower-calorie food of two options, and that correct answers would be rewarded with a $.25 bonus (doubling their base compensation). Then they saw two foods and chose which one they thought had fewer calories.
Choice
Participants chose between an AB&B toast and an avocado toast, presented in random order. The avocado toast (reference food) was the same for all participants, but for the AB&B toast (target food) they were randomly assigned to see either a pretty version or an ugly version of it (made from the identical ingredients; see Figure 1).
As theories of classical aesthetics predict, in a pretest (see Web Appendix B), people found the toast significantly prettier when it was orderly and symmetrical (M = 4.14, SD = 1.68) than when it was disorderly and asymmetrical (M = 3.22, SD = 1.74; p = .008).
Again, each food’s name was displayed under the respective image along with equal cost information. Participants chose either the AB&B toast (coded as 1) or the avocado toast (coded as 0) as the lower-calorie food and reported demographics. The objectively accurate answer was the avocado toast (which contained 270 calories, whereas the AB&B contained 380 calories). The relevant test is whether people’s propensity to (mistakenly) identify the AB&B toast as the lower-calorie option is greater when it looks pretty versus ugly.
Results
Table 1 displays the choice shares for the dependent variable. A logistic regression with target food prettiness as the predictor (ugly serving as the reference category) and choice of which food contained fewer calories as the dependent variable (the AB&B toast serving as the target criterion) was significant (Wald χ2(1) = 6.63, B = .613, SE = .238, p = .010). Specifically, significantly more people (falsely) identified the AB&B toast as the lower-calorie food when they saw its pretty version (48.0% [72 of 150]) than when they saw its ugly version (33.3% [50 of 150]). The odds of choosing the AB&B toast increased by a factor of 1.846 for the pretty version compared with the ugly version (signified by Exp[B]). That is, people were more likely to misidentify the higher-calorie food as the lower-calorie food merely because it looked prettier.
Discussion
These results extend the support for the basic hypothesis. In line with pretty foods being judged as healthier in the previous studies, in a consequential task, people were more likely to miscategorize a food as a lower-calorie option when it looked pretty than when it looked less pretty. The fact that this effect occurred even when people were financially incentivized to judge accurately suggests that it is not due to demand, low-effort thinking, or motivated reasoning (which is attenuated by accuracy motivation, see Hart et al. [2009]). The phenomenon is robust—people cannot ignore prettiness, even when monetary rewards are at stake.
Cognitive Nature of the Documented Associations and Implications for Moderators
Moving beyond demonstrating the effect of prettiness on perceived healthiness in various ways, the subsequent studies shift focus to moderators. To make predictions about potential moderation, it is important to first understand the cognitive nature of the associations. For instance, if the effect is rooted in strong implicit associations, interventions targeting the explicit level (e.g., verbal information) will likely require hefty information and deliberation to override said implicit associations. Therefore, I conducted a supplemental study to explore how the pretty = natural and natural = healthy associations operate psychologically on the explicit and implicit levels (see Web Appendix H).
A combination of Implicit Association Tests and surveys revealed that both the pretty = natural and the natural = healthy association are grounded in strong implicit associations. Notably, both associations appear to be much stronger at the implicit level than the explicit level (based on effect sizes). Indeed, for the pretty = natural association, people do not seem to hold a salient explicit belief in either direction. These findings have implications for viable moderators.
First, even people with weak or absent explicit beliefs exhibited implicit associations. Accordingly, while belief strength may moderate the mediating effect of perceived naturalness, weak beliefs alone are unlikely to turn off the effect completely. Indeed, a supplemental moderation study, which manipulated prettiness via order and symmetry using spaghetti marinara, corroborated this hypothesis: a weaker natural = healthy belief weakened the mediation by naturalness, but did not eliminate the pretty = healthy effect (see Web Appendix I).
Second, the best shot at mitigating the robust effect of prettiness on perceived healthiness may therefore lie in turning off the naturalness inference directly. This approach is particularly promising due to the absence of any strong explicit belief about the pretty = natural relationship. One way to curtail naturalness inferences is to manipulate prettiness in a manner that does not elicit the notion of naturalness in the first place. My theory proposes that expressive aesthetics will achieve just that. Another may be reminding people that pretty food has been artificially modified. Studies 4a and 4b investigate these theoretically predicted moderators and boundaries.
Study 4a: Only Classical Aesthetics Enhance Perceived Naturalness and Perceived Healthiness
I theorize that classical aesthetics, which resemble patterns found in nature, boost perceived healthiness specifically because they elicit the notion of naturalness. Per this logic, the effect of prettiness on naturalness should not extend to expressive aesthetics, which lack classical aesthetics features and are instead characterized by creativity. Study 4a tests this theoretical prediction. The predicted pattern would also negate perceived sophistication and care as alternate drivers, to the extent that they are heightened for expressive aesthetics, and contradict a general halo, as a halo should arise from the generic positive affect from any prettiness, not just specific types.
Method
Six hundred one Amazon Mechanical Turk panelists in the United States (49.1% women; Mage = 36.46 years, range = 19–78 years) were randomly assigned to evaluate one of three photographs: an AB&B toast that was ugly; the same AB&B toast that was pretty by virtue of classic aesthetic features; or an AB&B toast that was pretty by virtue of expressive aesthetic features (made from the same ingredients; see Figure 2).
Stimuli used in Studies 4a and 4b.
In a pretest (see Web Appendix B), people found the AB&B toast significantly prettier when it was symmetrical and orderly (classical aesthetics; M = 4.69, SD = 1.69) and when it displayed basic shapes assembled into a playful scene (expressive aesthetics; M = 4.65, SD = 1.70) than when it was disorderly and asymmetrical (ugly; M = 3.33, SD = 1.76; ps < .001). However, they found classical aesthetics and expressive aesthetics equally pretty (p = .81).
Importantly, the AB&B toast was rated as higher on classical aesthetics in the pretty–classical condition (M = 5.87, SD = 1.04) than in the pretty–expressive condition (M = 4.05, SD = 1.50; p < .001) and the ugly condition (M = 3.80, SD = 1.58; p < .001). Conversely, the AB&B toast was rated higher on expressive aesthetics in the pretty–expressive condition (M = 5.43, SD = 1.28) than in the pretty–classical condition (M = 4.29, SD = 1.58; p < .001) and the ugly condition (M = 3.14, SD = 1.66; p < .001). As such, the images meet the requirements for testing the hypothesis that the pretty = healthy effect emerges only for prettiness based on classical aesthetics, because they signal naturalness, but not for equally high prettiness based on expressive aesthetics, because they do not.
Before viewing the food, participants read that they would see an “almond butter and banana toast, made from one slice of wheat bread, a scoop of almond butter, and half of a banana.” Again, equal cost information was shown under each image.
Then they completed, in randomized order, the perceived healthiness scale (α = .87), the perceived naturalness scale (α = .85), and measures of perceived sophistication and care. However, neither the results for sophistication nor those for care aligned with the results for perceived healthiness, so measures and detailed results are reported only in Web Appendix J (for means, see Table 1). Finally, participants reported demographics.
Results
Table 1 displays the means for all dependent variables collected.
Basic effect: Perceived healthiness
Condition had a significant effect (F(2, 598) = 10.82, p < .001). Compared with the ugly condition (M = 4.83, SD = 1.13), participants rated the toast as significantly healthier in the pretty–classical condition (M = 5.31, SD = 1.09; t(598) = −4.24, p < .001, d = .43), but not any healthier in the pretty–expressive condition (M = 4.88, SD = 1.17; t(598) = −.44, p = .663, d = .04). They also rated the toast as significantly healthier in the pretty–classical condition than in the pretty–expressive condition (t(598) = −3.79, p < .001, d = .38).
Proposed mediator: Perceived naturalness
Condition had a significant effect (F(2, 598) = 14.53, p < .001). Compared with the ugly condition (M = 4.90, SD = 1.27), participants rated the toast as significantly more natural in the pretty–classical condition (M = 5.45, SD = 1.19; t(598) = −4.23, p < .001, d = .45), but no more natural in the pretty–expressive condition (M = 4.80, SD = 1.39; t(598) = .80, p = .425, d = .08). They also rated the toast as significantly more natural in the pretty–classical than in the pretty–expressive condition (t(598) = −5.02, p < .001, d = .50).
Mediation
Simple mediation (multicategorical independent variable with indicator coding, the ugly condition serving as the reference category; model 4; 10,000 samples) with perceived naturalness as the mediator returned the following results. For the comparison between the ugly and the pretty–classical aesthetics conditions, naturalness mediated the effect of prettiness on perceived healthiness (B = .292, SE = .066; 95% CI = [.164, .428]). For the comparison between the ugly and the pretty–expressive conditions, naturalness did not mediate (B = −.055, SE = .070; 95% CI = [−.194, .082])—as expected, given that these groups differed on neither naturalness nor healthiness. An additional mediation with pretty–expressive serving as the reference category showed that for the comparison between the pretty–expressive aesthetics and the pretty–classical aesthetics conditions, naturalness also mediated the effect of prettiness on perceived healthiness (B = .347, SE = .072; 95% CI = [.205, .487]).
The patterns of the perceived sophistication and perceived care results across the three conditions did not align with the perceived healthiness results, making them unlikely mediators. Nonetheless, I also tested parallel mediation with naturalness, care, and sophistication simultaneously. Central to testing the theory about the role of perceived naturalness, all mediation patterns for naturalness remained unchanged; that is, naturalness mediated for all three comparisons above and beyond the other variables. Care also had some additional explanatory value for some comparisons; sophistication had none. Because the key insight is that naturalness is a unique driver of the effect, these ancillary results are reported only in Web Appendix J.
Discussion
These results not only provide more support for the proposition that prettiness in the form of classical aesthetics boosts perceived healthiness by signaling naturalness but also show that neither sophistication nor care are viable alternative explanations. Furthermore, the finding that both the classically pretty and expressively pretty toast were rated as equally pretty speaks against general positive inferences from general attractiveness as an alternative explanation and contradicts the idea of prettiness merely casting a general halo.
Study 4b: Artificial Modification Disclaimer Eliminates the Effect of Prettiness
This research proposes that prettiness enhances perceived healthiness by way of intuitively being perceived as more natural. Accordingly, reminders that the food was artificially styled should mitigate the effect of prettiness on perceived healthiness, insofar as they suppress naturalness perceptions. Study 4b tests this theoretical prediction. The anticipated pattern would provide further evidence of process by moderation and also point to an effective intervention. Study 4a manipulated (classical) aesthetics chiefly via symmetry and order. Study 4b manipulates prettiness via the presence of pattern repetition, while again listing ingredients and displaying the food name with the picture.
Method
Three hundred one Prolific Academic panelists in the United States (55.8% women; Mage = 34.17 years, range = 18–79 years) were randomly assigned to evaluate one of three photos: an ugly avocado toast, a pretty avocado toast, or the pretty avocado toast plus a disclaimer that the food had been artificially modified (all made from the identical ingredients; see Figure 2).
In a pretest (see Web Appendix B), people found the avocado toast significantly prettier both when it featured pattern repetition (M = 4.57, SD = 1.57) and when it featured pattern repetition plus the disclaimer (M = 4.41, SD = 1.56) than when it lacked salient pattern (M = 3.06, SD = 1.46; ps < .001). However, they found the two pretty conditions equally pretty (p = .591).
Before viewing the food, all participants read that they would see an “avocado toast, made from 1 slice of wheat bread and 1/2 an avocado.” The ingredient text was boldfaced, and participants were required to stay on the page for 10 seconds to ensure that they would not skip over these important details. Participants in the pretty + disclaimer condition also read that the food they would see had been “artificially modified for advertising” and “as a result, the food may strike [them] as unnatural.” Then participants moved on to viewing the food.
Again, the food’s name and equal cost information were shown under each image. Participants in the pretty + disclaimer condition also saw a statement reiterating that the food was “artificially modified for advertising” and that it was “not a natural representation” (see Figure 2). The Federal Trade Commission requires that advertising disclaimers be “clear and conspicuous” (www.ftc.gov). Thus, presenting the statement near the relevant content resembles how it would likely be displayed in the real world.
They completed the perceived healthiness scale (α = .83), and then, in randomized order, the perceived naturalness scale (α = .86) and the perceived tastiness scale (α = .95). They also completed the perceived price measure (r = .87, p < .001). As in Studies 1a–b and the replication, tastiness and price were equal across conditions, so detailed results are reported only in Web Appendix K (for means, see Table 1).
They also rated a measure of perceived amount. However, the pattern did not align with that for perceived healthiness, so the measure and detailed results are reported only in Web Appendix K (for means, see Table 1).
Finally, participants reported demographics, which included education level and household income. Neither interacted with prettiness on perceived healthiness or naturalness, so these variables are not discussed further (although details and results are available upon request).
Attention check
Finally, participants recalled the most prominent component of the food as in Study 1a. In this free recall, 98.67% (297 out of 301) responded “avocado”; the four other responses were distributed equally across conditions (χ2(2) = .51, p = .773).
Results
Table 1 displays the means for all dependent variables collected.
Basic effect: Perceived healthiness
Condition had a significant effect (F(2, 298) = 4.57, p = .011). Compared with the ugly condition (M = 4.58, SD = 1.22), participants rated the avocado toast as significantly healthier in the pretty condition (M = 5.07, SD = 1.13; t(298) = –2.99, p = .003, d = .42), but not any healthier in the pretty + disclaimer condition (M = 4.76, SD = 1.16; t(298) = −1.11, p = .267, d = .15). They also rated the toast as marginally healthier in the pretty condition than in the pretty + disclaimer condition (t(298) = −1.87, p = .062, d = .27).
Proposed mediator: Perceived naturalness
Condition had a significant effect (F(2, 298) = 8.78, p < .001). Compared with the ugly condition (M = 4.89, SD = 1.27), participants rated the toast as significantly more natural in the pretty condition (M = 5.52, SD = 1.04; t(298) = −3.38, p = .001, d = .54) but no more natural in the pretty + disclaimer condition (M = 4.81, SD = 1.60; t(298) = .45, p = .656, d = .06). They also rated the toast as significantly more natural in the pretty condition than the pretty + disclaimer condition (t(298) = −3.83, p < .001, d = .53).
Mediation
Simple mediation (multicategorical independent variable with indicator coding, ugly serving as the reference category; model 4; 10,000 samples) with perceived naturalness as the mediator returned the following results. For the comparison between the pretty and the ugly conditions, naturalness mediated the effect of prettiness on perceived healthiness (B = .285, SE = .080; 95% CI = [.136, .447]). For the comparison between the pretty + disclaimer and the ugly conditions, naturalness did not mediate (B = –.038, SE = .093; 95% CI = [−.221, .147])—as expected, given that the groups differed neither on naturalness nor on healthiness. An additional mediation with pretty + disclaimer serving as the reference category showed that for the comparison between the pretty and the pretty + disclaimer conditions, naturalness also mediated the effect of prettiness on perceived healthiness (B = .323, SE = .089; 95% CI = [.155, .504]). All results are robust to controlling for tastiness, amount, and price.
Discussion
These results are informative from both a theoretical and a managerial or policy perspective. First, they provide process-by-moderation evidence for the idea that prettiness increases perceived healthiness specifically by signaling naturalness: as this theory predicts, if the pretty = natural link is disrupted, prettiness no longer exerts its effect. Similar to Study 1b, this pattern also rules out several alternative explanations, such as actual differences in food visibility or amounts or prototypicality of the food presentation between the visuals, given that the two pretty conditions used exactly the same photograph, yet still produced healthiness differences based on the artificial modification disclaimer. Beyond the moderation pattern, tastiness was unaffected by prettiness, challenging both a motivated reasoning and a general halo explanation.
Second, these results offer important directions to marketers or policy makers who want to protect consumers from drawing false inferences about healthiness. Evidently, consumers can be inoculated against the misleading effect of aesthetics, specifically by highlighting that food in advertising pictures has been artificially modified. Disclaimers placed near food images may be an actionable way to mitigate the illusory healthiness boost that prettiness otherwise induces.
General Discussion
People constantly encounter food that is styled to look pretty. The literature makes multiple compelling, but divergent, predictions: pretty aesthetics may elicit notions of pleasure and thereby induce lower healthiness judgments, or pretty aesthetics may give rise to intuitions about specific attributes that lead to higher healthiness judgments. I propose the latter and suggest that perceived naturalness may be a key driver. Six high-powered experiments (N = 2,492), a pilot study (N = 803), and four supplemental studies (N = 1,006) demonstrate that people perceive prettier (vs. less pretty) versions of the same food as healthier because they seem more natural. This effect materializes with naturalistic and controlled stimuli; with visual and nonvisual manipulations of prettiness; with photographed and live, unhealthy and healthy, and processed and whole foods. It misleads people’s choices even when they have financial stakes and emerges irrespective of prettier food looking tastier or larger. Prettiness affects only select attributes, and only classical aesthetics produce the effect, contradicting a general halo. The effect is independent of tastiness and it persists when accuracy is incentivized, contradicting motivated reasoning. In all studies, the effect emerges despite equal price perceptions.
Theoretical Contribution
This research is the first to explore the role of aesthetics in healthiness judgments. I systematically tested different possible processes that could give rise to the observed effect of aesthetics and found that the effect of aesthetics on perceived healthiness has a cognitive basis (lay intuitions) rather than an affective (halo effect) or motivational (wishful thinking) one.
These findings add to a body of work that documents how lay theories of food, nutrition, taste, and dieting govern many everyday food decisions (Finkelstein and Fishbach 2010; Haws, Reczek, and Sample 2017; McFerran and Mukhopadhyay 2013; Raghunathan, Naylor, and Hoyer 2006). First and foremost, they reveal the novel lay intuition that classical aesthetics are a sign of naturalness. They also support and extend work on the basic link between naturalness and healthiness, which has shown that more natural entities are ranked as generally healthier than less natural entities (Rozin et al. 2004) by quantifying health perceptions in a more nuanced way (e.g., calorie and fat content). Furthermore, this research identifies a moderator of the unhealthy = tasty belief. Prior work, in which foods were either not depicted or visually identical, found that describing foods in unhealthy (vs. healthy) terms may enhance expected and experienced tastiness (Mai and Hoffmann 2015; Raghunathan, Naylor, and Hoyer 2006). Yet the current research suggests that classical aesthetics can make food seem simultaneously healthier and tastier. This qualification is critical, as it offers a solution to overcome the (perceived) health–taste trade-off that may pose a barrier to healthy choices.
This research shows that classical aesthetics elicit a sense of naturalness. This association is somewhat counterintuitive—achieving the maxims of classical aesthetics (i.e., order, symmetry, and balance) often requires effort and artificial manipulation. Yet it raises novel implications for aesthetics research. A new way to think about “what is beautiful is good”–type effects may rather be that “what is natural is good:” naturalness may underlie other positive aesthetics effects attributed to a “beauty halo.” The results also highlight that aesthetics research may benefit from distinguishing between classical and expressive aesthetics, as they may operate differently. I offer initial evidence that the pretty = natural effect may be limited to classical aesthetics, but additional empirical examination of the different types of aesthetics is warranted.
The pretty = natural effect in particular opens up new avenues for aesthetics research. Not only may the effect generalize from food to other entities (e.g., products, people), it also points to novel related mediators (e.g., authenticity, talent) and downstream effects of aesthetics in other domains. For example, it is conceivable that interior design emphasizing classical aesthetic principles may cast products sold or services rendered in a space as more wholesome and health-promoting. It also prompts questions about how aesthetics influence perceived product efficacy. Some research suggests that outcomes achieved naturally are perceived as higher quality (e.g., Tsay 2016), but other work hints that more natural products are perceived as weaker (e.g., Luchs et al. 2010). How aesthetics, by way of influencing perceived naturalness, shape consumers’ beliefs about product success or strength should be an important line of inquiry for marketers.
Marketing and Policy Implications
The finding that the same food appears healthier when it looks more aesthetically pleasing than when it looks less aesthetically pleasing has practical implications for marketers and policy makers. First, it means that many food advertisements and restaurant menus may, by virtue of depicting heavily styled foods, be promising more than tastiness alone. They may be (falsely) heralding greater levels of healthiness along with it. This research shows that the pretty = healthy effect extends to unhealthy foods (e.g., pastries). This finding is disconcerting because a large proportion of visually advertised food is unhealthy food. For example, approximately 72% of restaurant ads viewed per year are those of fast-food restaurants, and the most advertised food brand in the United States is McDonald’s (Statista 2016). If fast food is consistently portrayed in ways that increase perceived healthiness, even consumers who are motivated to select foods based on healthfulness may become more likely to make unhealthy choices, such as considering fast food as an option at all, choosing it more frequently, selecting larger portions of seemingly not-so-unhealthy foods, or underestimating how much “balancing out” fast food requires (e.g., via exercise). Transcending unhealthy choices at individual decision points, the persistent subtle boost in perceived healthiness may, over time, even promote overly optimistic general beliefs about unhealthy food types, for instance, that fast food overall is fairly healthy.
Second, food advertisers and restaurant marketers may strategically leverage aesthetics as a signal of healthiness in times when health-related marketing language is increasingly under scrutiny. Since 2016, the Food and Drug Administration has begun curtailing which foods can be labeled “healthy,” but healthfulness is a valuable attribute (with losses as much as $3.82 million in monthly revenue after dropping health claims; Rao and Wang 2017). Firms may seek more surreptitious ways to elevate perceived healthiness among their increasingly health-conscious customer base. The potential use of aesthetics as a (deceptive) signal that misleads “reasonable” people (i.e., college students, campus passersby, online panelists) warrants close consideration by policy makers. Even companies that do not wish to actively deceive customers may inadvertently lead people astray in their healthiness judgments simply by presenting food in a (classically) pretty fashion. Indeed, findings from this research indicate that consumers do not seem to hold a salient belief that prettier food is more natural, and this lack of awareness may make them particularly vulnerable (and marketing managers oblivious) to the bias.
On the upside, this research also identifies an effective intervention to protect consumers from being deceived by foods’ aesthetics. I found that a statement that explicitly reminds people that a pretty food was artificially modified for depiction can mitigate the effect. Given that it is not viable to prohibit firms from depicting food products in a basic aesthetically pleasant manner altogether, disclaimers may be the most practical solution.
Limitations and Future Research
This research was designed to test the influence of food aesthetics on perceived healthiness and food choice. Food evaluation and choice are critical for a healthy lifestyle, but of course, subsequent food consumption decisions, such as portion size selection or intake, play a role as well. I did not examine consumption, and given the complexity of eating decisions I expect multiple judgments to interact with each other. For instance, people may eat less of a high (vs. low) aesthetic food for fear of destroying its beauty (Wu et al. 2017, Study 2) in spite of its apparent healthiness. At the same time, people often overcompensate and increase consumption when food seems healthy (Suher, Raghunathan, and Hoyer 2016) or when context factors reduce guilt for unhealthy eating (Hagen, Krishna, and McFerran 2017), so they may eat more of a prettier food because it seems healthier. Future research may investigate these competing forces directly.
The findings also cannot fully illuminate whether the effect of prettiness is equally strong in the very low versus the very high prettiness range. I found the effect with stimuli at various levels of prettiness (see Web Appendix B), which suggests that the effect is not limited to either the positive or the negative realm. Nonetheless, it is conceivable that prettiness has a stronger effect in the below-average domain (i.e., very ugly vs. medium) than the above-average domain (i.e., medium vs. very pretty), akin to patterns found in person perception research (Griffin and Langlois 2006).
Relatedly, these studies demonstrate the effect of prettiness on perceived healthiness across a variety of foods, but there are likely boundary conditions. For example, the effect emerged with unhealthy food categories (e.g., pizza, cupcakes) but may not extend to extremely unhealthy food (e.g., pretty butter pats); and it appeared with highly processed foods (e.g., frosting) but may well be thwarted by more blatantly unnatural components (e.g., neon dyes, edible glitter).
Finally, this research documents the intuitions that people have, but not their origin—are they merely overgeneralizations from otherwise true correlations in the real world or utter misconceptions about the food system? It is unrealistic that prettier food really is more natural on average, given how much preparation and additional ingredients go into making food look pretty (e.g., color preservatives). Instead, an extrapolation from nature-like patterns that define classical aesthetics to naturalness seems reasonable. Likewise, greater naturalness does not necessarily boost nutritional value (Smith-Spangler et al. 2012), but it is plausible that consumers frequently observe a co-occurrence of naturalness signals (e.g., organic label) and healthiness signals (e.g., marketers’ emphasis of a healthy lifestyle). Content analysis and survey data may be more fruitful approaches to the complex issue of how lay beliefs develop than experimental methods.
Conclusion
This research exposes a novel effect of aesthetics in the domain of food and reveals an unrecognized influence on healthiness judgments. People perceive the same food as more natural when it happens to look prettier and believe that this naturalness implies healthiness—both in terms of the presence of positive elements (e.g., nutrients) and the absence of negative elements (e.g., calories). The investigation used primarily laboratory experiments and focused chiefly on causes of consumers’ perceptions of healthiness rather than on the consequences of these perceptions. However, the studies document that perceived healthiness affects real WTP and that the pretty = healthy bias is robust even when consumers are (financially) motivated to choose the healthy option. While the origins of lay intuitions about nutrition are not always clear, other research has shown that they can meaningfully influence outcomes ranging from food choice to consumption to weight status (McFerran and Mukhopadhyay 2013). Given the prevalent use of food visuals in marketing practice and research alike, understanding the effect of food aesthetics offers important insights into the thought processes of consumers trying to make pretty healthy choices.
Supplemental Material
Supplemental Material, WebApp_JM.18.0435Final - Pretty Healthy Food: How and When Aesthetics Enhance Perceived Healthiness
Supplemental Material, WebApp_JM.18.0435Final for Pretty Healthy Food: How and When Aesthetics Enhance Perceived Healthiness by Linda Hagen in Journal of Marketing
Footnotes
Acknowledgments
The author thanks Ed O’Brien, Brent McFerran, Kristin Diehl, and Kirk Kristofferson for helpful comments on previous versions of this article, as well as Elisa Solinas for assistance collecting data for Study 2. Supplementary materials are included in the Web Appendix. OSF link to data and material: 
Associate Editor
Kelly Haws
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
