Differences in hedonic perception of retronasal odors between young children and adults

Abstract

Early childhood is a critical developmental period for the establishment of flavor preferences that in turn affect food and beverage consumption and health into adulthood. Flavor is a multisensory experience, combining taste and retronasal odor signals. However, while early life development of taste perception has received ample attention, there is limited knowledge of retronasal odor perception in early life. In the present cross-sectional study, we tested the hypothesis that hedonic perception of retronasal smell differs between children and adults. We used video analysis of facial expressions to taste and retronasal odor solutions in children and adults. Children ages 3 to 6 and one of their parents (n = 112 dyads) were asked to sample solutions containing either a taste or an odor compound. A subset of subjects (n = 84 dyads) also explicitly rated each solution on a pictorial liking scale. No differences between the 2 age groups were observed in responses to taste solutions. In contrast, responses to retronasal odor stimuli were less stimulus-specific in children compared with adults. Children showed fewer negative facial expressions to broccoli and pumpkin odors, and more negative facial expressions to apple and mango odors. Similar differences between the 2 age groups were observed in explicit hedonic ratings. These findings support our hypothesis that the hedonic value of retronasal odor components of flavor is not innate but differ between young children and adults.

1. Introduction

Between ages 2 and 6, children enter a critical developmental period when it comes to ingestive behaviors. During this time, children become independent in their food choices as they begin self-feeding and consuming foods similar to an adult diet [1, 2]. Early exposure to foods fosters the development of eating habits that persist into adulthood and nutritional intake during this period contributes to risk for a variety of diseases later in life, including heart disease, liver disease, and type 2 diabetes [3, 4]. Although a large variety of factors interact to govern food choices, flavor perception is a major contributor, and therefore a potential target for developing interventions aimed at improving eating habits in early life (Resnicow et al. 1997; Gibson et al. 1998; Sorensen et al. 2003; Brug et al. 2008; Beauchamp and Mennella 2009).

Flavor perception is a multisensory experience, combining taste and retronasal odor (among other qualities). During consumption, taste compounds stimulate the lingual epithelium and give rise to the following gustatory qualities: sweet, salty, umami, sour and bitter. At the same time, olfactory compounds released inside the oral cavity travel up the nasopharynx to stimulate the olfactory epithelium retronasally, giving rise to odor qualities such as fruity and vanilla that are often colloquially described as “taste”. Retronasal olfaction is a major contributor to flavor experience [21 to 24] as it constitutes the sensory qualities that uniquely identify most foods (e.g. the qualities that distinguish between peach and pear; pumpkin and sweet potato). Thus, flavor experience has contributions from multiple sensory systems that each may have different developmental profiles and changing contributions to flavor perception throughout the lifespan. Early life development of taste perception has received extensive attention (Steiner 1974; Liem and Mennella 2002, 2003; Beauchamp and Mennella 2009, 2011; Mennella et al. 2009; Schwartz et al. 2009; Forestell and Mennella 2015; Liem 2017; Fry Vennerod et al. 2018). Research on hedonic evaluation of taste stimuli indicates a strong innate component: liking sweet taste and disliking bitter taste are universal patterns that are present from birth and relatively stable throughout the lifespan (Maller and Desor 1973; Desor and Beauchamp 1987; Mennella 2008, 2014; Beauchamp and Mennella 2009; Ventura and Mennella 2011; Mennella et al. 2014). In contrast, few studies have directly investigated retronasal smell perception in early life (Konstantinidis et al. 2005; Altundag et al. 2014; Cayonu et al. 2014). To the best of our knowledge, no studies have explicitly characterized hedonic perception of retronasal odors in children when they become independent in their food choices.

Despite limited knowledge of retronasal smell perception in childhood, research on orthonasal olfaction in adulthood illustrates that smell perception is highly susceptible to experience-dependent plasticity (Huart et al. 2019; Maier and Zhang 2023). Exposure to orthonasal odors improves olfactory identification skills in healthy subjects (Morquecho-Campos et al. 2019) and olfactory function in patients with olfactory loss (Hummel et al. 2009). Adults are also able to quickly learn hedonic associations between odors and tastes (Stevenson et al. 1998; Li et al. 2008). This work suggests odor perception may undergo substantial changes throughout the lifespan. However, our knowledge of smell perception in early childhood and how it compares to adults is lacking, partly due to methodological challenges. Odor perception in adults is often measured using identification tasks, but previous work has shown that performance on such tasks improves with age and grade level, indicating that verbal and cognitive skills may confound results (Doty et al. 1984; Lehrner et al. 1999; Frank et al. 2004; Hummel et al. 2007; Monnery-Patris et al. 2009). Indeed, studies of orthonasal odor perception including the 3 to 6 yr old age range have yielded mixed results, with some studies highlighting differences between children and adults regarding sensitivity (Doty et al. 1984; Kobal et al. 2000; Monnery-Patris et al. 2009) and hedonic evaluation of smell (Rinck et al. 2011); others highlighting similarities between the 2 age groups for sensitivity (Lehrner et al. 1999; Frank et al. 2004; Hummel et al. 2007) and hedonic evaluation (Schmidt and Beauchamp 1988; Bensafi et al. 2007). In slightly older children (6 to 9 years of age), Armstrong et al. showed that facial EMG responses to orthonasal odors were similar between children and adults (Armstrong et al. 2007).

Here, we investigated hedonic perception of retronasal odors in early childhood. Specifically, we used a cross-sectional design to test the hypothesis that hedonic perception of retronasal smells differs between 3 and 6 yr-old children and adults. Alternatively, hedonic value of food odors could have a strong inborn component (as suggested by the extant literature on taste) or become established during fetal and infant developmental stages. The latter is suggested by findings that volatile molecules present in the maternal diet can be transmitted through amniotic fluid (Ustun et al. 2022) or breastmilk (Mennella et al. 2009), resulting in increased acceptance of the foods associated with those odors by the offspring. Another possibility is that children and adults differ in how they express hedonic valuation independent of how they value individual odor stimuli. To test our hypothesis, we asked 3 to 6 yr-old children and their parents to sample series of taste and odor solutions. We measured hedonic perception using 2 age-appropriate, nonverbal methods: video analysis of facial expressions (Ekman and Friesen 1978) and a pictorial rating scale (Chen et al. 1996; Visser et al. 2000; Yuan et al. 2016; Sorokowska et al. 2024). Measures obtained from children and adults were directly compared and revealed differences in hedonic perception of retronasal odors between the 2 age groups.

2. Methods

2.1. Subjects

Healthy children 3 to 6 years old (mean age: 5.02 years, 56% female) and one of their parents (mean age: 33.84 years; 81% female; total of n = 112 dyads; 76% white, 13% black, 11% biracial; 9.7% Hispanic/Latino) were recruited and tested at venues in the local community (children's museums and primary care clinics). Children were excluded from participating in the study if they did not speak English, had a known food allergy, or suffered from nasal congestion, impaired sense of taste and/or smell, motor and/or speech delay/difficulties, other developmental delays, or were diagnosed with an autism spectrum disorder. Parents were excluded if they met any of the previously listed exclusion criteria, were pregnant, or smoked. Prior to testing, parents completed a child eligibility form and provided informed consent and permission for their child to participate and for video recording; children provided verbal assent. The study was approved by the Wake Forest University School of Medicine Institutional Review Board (protocol #: IRB00050522). This study complies with the Declaration of Helsinki for Medical Research involving Human Subjects.

2.2. Stimuli

Both gustatory and retronasal olfactory compounds were presented in aqueous solution in plastic cups (25 mL of solution in 60 mL cup) with lids and tightly fitting straws (5 mm diameter) to prevent orthonasal detection. Each solution consisted of a food-grade tastant or odorant dissolved in purified water. Tastants (https://www.fishersci.com/) consisted of sucrose (0.3 M, sweet) and quinine-HCl (0.1 mM, bitter). The concentration of sucrose was selected based on high palatability in both children and adults (Mennella et al. 2015; Bobowski and Mennella 2017). The concentration of quinine was selected based on previous work showing that this concentration yielded maximum negative hedonic ratings in children ages 5 to 12 years without subjects refusing to taste the stimuli (Mennella et al. 2015). Odorants consisted of mango flavorant (0.2% vol/vol), honey crisp apple flavorant (0.2%), pumpkin flavorant (0.1%) and broccoli flavorant (0.1%), provided by Givaudan Flavors and Fragrances Corporation (https://www.givaudan.com/). Odors were selected to represent fruit and vegetable odors that vary in how often they are likely to be consumed in the local area. Odor concentrations were matched for intensity based on subjective evaluation by adults.

2.3. Testing procedure

Parents were asked if they wanted to participate in an experiment in which they and their child would taste different clear liquids containing flavors found in everyday foods. Children were asked if they wanted to help a scientist with an experiment. All subjects were told that they could say no or discontinue the experiment at any time. Testing took place in a single session (15 to 20 min total duration) in a closed room with minimal distractions. First, the child participant completed flavor sampling while the parent completed a set of questionnaires, facing away from the child to minimize social influences. Once the child completed sampling all solutions, the parent sampled while the child picked out a sticker reward with one of the investigators, facing away from the parent. The following procedures apply to both child and adult participants. Participants were seated at a table across from the experimenter and offered 25 mL of each of the taste and retronasal odor stimuli, one at a time. Participants were instructed that they could drink as much of the solution as they wanted. The order of stimulus presentation was randomized for each participant. Six seconds after participants removed the straw from their mouth, they were asked to rate the stimulus (see below) before moving on to the next stimulus. Facial responses were recorded using a video camera positioned directly across from participants. Subjects were not explicitly told that footage would be used to analyze facial expressions in order to minimize the influence of the test environment on facial responses. Parents received a $25 gift card for participating.

2.4. Pictorial ratings

A subset of subjects (n = 84 dyads) rated the stimuli after sampling. After sampling each test solution, subjects were first asked if it “tasted like water” or “different from water”. In case of the latter response, subjects were then presented with a pictorial rating scale printed on paper and asked to point to a smiley face or frowny face to indicate if the stimulus was “yummy” or “yucky”. Before testing, parents were asked to indicate a known liked food and known disliked food for their child, and the child was asked to point to the face for the known “yummy” and known “yucky” food before proceeding with the experiment to ensure that they understood the task. Subjects were also presented with plain water for reference before testing began and were told “this is plain water”. Three children were excluded because they gave the same rating for all stimuli, suggesting bias.

2.5. Video analysis

Facial expressions were scored using the Facial Action Coding System (FACS) by a certified coder (SEC) (Ekman and Friesen 1978). FACS is an anatomically based coding system used to describe facial muscle movements. The minimally discernable action of a facial muscle is called an Action Unit (AU) (Ekman and Friesen 1978). The intensity of each AU is scored on a scale of A (slight movement) to E (maximum movement). FACS coding identifies facial muscle movements but does not indicate meaning (Greimel et al. 2006). Therefore, we used a subset of AUs and AU combinations from the FACS system that have been previously reported to be associated with positive and negative affect in the literature (Steiner 1973; Soussignan and Schall 1996; Horio 2003; Greimel et al. 2006; Mennella et al. 2009; Zeinstra et al. 2009). The following negative AUs were included: inner eyebrows raised (AU 1), outer eyebrows raised (AU 2), brows pulled together and lowered (AU 4), nose wrinkle (AU 9), eyes closed (AU 43), blink (AU 45), upper lip raiser (AU 10), lip corner depressor (AU 15), lower lip depressor (AU 16), lip stretch (AU 20), lip funneler (AU 22), lips pressed (AU 24), lips part (AU 25), gape (AU 25 + 26 + 27), and head shake (AU 84). Positive AUs included smile with cheeks raised (AU 6 + 12), lip corner puller (AU 12), smile with open mouth (AU 12 + 25), lip smack (AU 24 + 25), head nod (AU 81), and cheek sucking (AU 35). The first 6 s immediately after the subject removed the straw from the mouth were coded (Soussignan and Schall 1996; Zeinstra et al. 2009). The coder was blind to stimulus identity. AUs that occurred while the subject was talking or took additional sips were not counted. Videos for 4 children were excluded because the child moved outside of the frame or their face was not clearly visible due to wearing a hat or mask, resulting in a final sample of n = 112 adults and n = 108 children.

2.6. Data analysis

The proportion of subjects who showed different categories of AUs and individual AUs was compared between children and adults using chi-square tests. Two-tailed, independent sample t-tests were used to compare the average total counts of negative AUs for each stimulus between children and adults. Two-tailed, independent sample t-tests were also performed to compare the average differences in occurrence of positive or negative AUs between children and adults for each stimulus. To classify odors based on adult perception, we used a one-way ANOVA on the number of negative AUs per adult subject with factor odor identity. Based on perceptual patterns observed in adult subjects in the present study, we classified broccoli and pumpkin as negative odors; apple and mango as positive odors. Thus, this classification is specific to the context of the present study. We then ran 2 separate 2-way mixed ANOVAs with factors Odor Category (positive, negative) and Age Group (children, adults) on the frequency of occurrence for individual negative and positive AUs to compare how the 2 age groups respond to odors within the same category.

3. Results

The goal of the present study was to compare hedonic perception of retronasal odor stimuli between young children and adults. We accomplished this by recording facial expressions of 3 to 6 yr-old children and their parents while they sampled flavor solutions that contained either gustatory or olfactory compounds. Video recordings were analyzed using the FACS to detect individual action units (AUs). A subset of subjects was also asked to explicitly rate the stimuli as “yucky” or “yummy” using a pictorial rating scale.

3.1. Patterns of facial responses to taste solutions

We first analyzed video recordings of facial responses to solutions of the gustatory compounds sucrose and quinine. Sucrose (sweet) and quinine (bitter) were used because of their known positive and negative valence, respectively, for both children and adults [43, 44], thus providing a tool for confirming the validity of negative and positive AU categories to measure hedonic flavor perception. Figure 1a and b shows the frequency of occurrence for each AU in response to sucrose and quinine in adults and children. Both positive and negative AUs were expressed in response to both stimuli in both age groups. This indicates that the occurrence of an isolated AU does not indicate perceived valence of a stimulus. However, for both age groups, the number of negative AUs was higher in response to quinine than in response to sucrose (Adults: Χ² = 12.68, P < 0.01; Children: Χ² = 4.81, P < 0.05), while the number of positive AUs was higher in response to sucrose as compared to quinine (Adults: Χ² = 12.08, P < 0.01; Children: Χ² = 4.10, P < 0.05). This effect is more readily observed in Fig. 1c, showing the difference in occurrence between the 2 stimuli. This finding confirms that the frequency at which individual AUs are observed across the population reflects the hedonic value of the stimuli. Three negative AUs occurred more frequently for sucrose than quinine (AU 16, 22, and 25), suggesting that these AUs may not reliably signal negative affect in the current context. We therefore excluded these AUs from further analysis.

Fig. 1.

Facial response patterns to sucrose and quinine. The proportion of subjects who showed a given AU in response to sucrose or quinine are shown for adults (a) and children (b). (c) The difference in the proportion of subjects who showed a given AU in response to quinine minus sucrose for adults and children. Positive values indicate that the AU occurred more in response to quinine.

3.2. Overall responsiveness to flavor solutions

Next, we directly compared the occurrence of AUs between children and adults. Figure 2 shows occurrence of any facial expression across all stimuli (i.e. the proportion of subjects in each age group that showed at least 1 facial expression in response to the stimuli), broken down by AU categories. Table 1 column 2 shows the same but broken down into individual stimuli. Across all stimuli, a higher percentage of children (90%) than adults (71%) showed any AU (Χ² = 4.60, P < 0.05). The same pattern was observed for each individual stimulus. Figure 2 shows that this difference was primarily driven by positive AUs. The occurrence of any positive AU was 1.56 times higher in children compared with adults (Χ² = 14.08, P < 0.001). The occurrence of any negative AU did not differ significantly between children and adults, with a slightly higher frequency in children (1.09-fold increase; Χ² = 3.15, P = 0.07).

Fig. 2.

Overall responsiveness. Proportion of subjects who showed any AU across all stimuli separated by type of AU (all, positive, negative). Children (n = 108); adults (n = 112). The proportion of subjects who showed any AU (Χ² = 4.60, P < 0.05), as well as any positive AU (Χ² = 14.08, P < 0.001), was higher for children compared with adults.

Table 1.

Proportion of subjects who showed any AU overall, any positive AU, or any negative AU.

	n (Adults, Children)	Proportion of subjects who showed any AU (Adults, Children)	Proportion of subjects who showed any positive AU (Adults, Children)	Proportion of subjects who showed any negative AU (Adults, Children)
Sucrose	112, 108	0.75, 0.86*	0.28, 0.35	0.52, 0.52
Quinine	83, 79	0.81, 0.91*	0.08, 0.21**	0.78, 0.72
Apple	85, 83	0.67, 0.97***	0.24, 0.38*	0.43, 0.65**
Mango	56, 54	0.48, 0.85***	0.12, 0.18	0.35, 0.61**
Broccoli	85, 83	0.79, 0.92*	0.16, 0.26	0.72, 0.69
Pumpkin	56, 54	0.66, 0.87**	0.12, 0.25	0.55, 0.59

*P<0.05, **P<0.01; ***P<0.001.

3.3. Differences in facial response patterns between children and adults

The analysis presented above demonstrates that the frequency at which individual AUs are observed across the population reflects the hedonic value of gustatory stimuli in both children and adults. However, there also appear to be differences in the frequency of occurrence for individual AUs between the 2 age groups that are independent of the value of the stimulus being sampled. For example, comparing Fig. 1a and b, the occurrence of negative AU 45 in response to both taste stimuli is higher in adults, while negative AU 10 is more common in children; positive AU 6 + 12 occurs more frequently in children. To test for the possibility that children and adults differ in the AUs they express in response to our stimuli, we directly compared frequency of occurrence for each AU across all 6 solutions (i.e. both taste and odor). Frequency of occurrence for each AU is shown in Fig. 3 for both children and adults. Children showed overall more positive AUs, as shown in Fig. 2. Specifically, we observed significantly more zygomatic smiles (AU 6 + 12) and smiles with mouth open (AU 12 + 25) in children compared with adults. Regarding negative AUs, we observed a less consistent pattern: some negative AUs occurred more frequently in children, while others occurred more frequently in adults. Thus, although the overall frequency of negative AUs is the same in children and adults (Fig. 2), the 2 age groups differ in the specific negative AUs they use.

Fig. 3.

Facial response patterns for children and adults. The proportion of subjects who showed a given AU across all stimuli for children and adults. Some AUs occurred more frequently in children: AU 4 (Χ² = 10.6, P < 0.01), AU 10 (Χ² = 9.69, P = 0.001), AU 24 (Χ² = 8.89, P < 0.01), and AU combination 25 + 26 + 27 (Χ² = 7.99, P < 0.01). Other AUs occurred more frequently in adults: AU 1 (Χ² = 5.00, P < 0.05), AU 2 (Χ² = 8.61, P < 0.01), AU 20 (Χ² = 5.89, P = 0.01), AU 43 (Χ² = 27.2, P < 0.001), AU 45 (Χ² = 9.76, P < 0.01), and AU 84 (Χ² = 13.36, P < 0.001).

3.4. Differences in facial response patterns to odor solutions between children and adults

3.4.1. Occurrence of any negative AU

The observation reported above shows that children and adults use different expressions (i.e. children and adults differ in how they express). However, children and adults do not differ in what they express in response to the gustatory stimuli sucrose and quinine. That is, both age groups express more negatively in response to quinine than sucrose. To test whether the 2 age groups differ in what they express in response to retronasal odors, we compared facial expressions between children and adults in response to individual odor stimuli. Table 1 column 3 compares the proportion of subjects who showed any positive AU to individual stimuli. Consistent with Figs. 1 and 3, the proportion of positive AUs was higher for children compared with adults for each individual stimulus. Table 1 column 4 compares the proportion of subjects who showed any negative AU in response to individual odor stimuli. Children showed a higher proportion of negative AUs to apple and mango odors compared with adults. There were no differences in the proportion of negative AUs in response to quinine, sucrose, broccoli and pumpkin. This finding demonstrates the existence of stimulus-specific differences in the expression of negative AUs between children and adults. However, only considering the presence/absence of any negative AU does not capture more subtle differences in expressiveness. For example, subjects who show multiple AUs in response to a stimulus are not distinguished from subjects who show only a single AU. Thus, we considered the average number of negative AUs expressed per subject.

3.4.2. Average number of negative AUs

Figure 4 shows the average number of total negative AUs expressed by each child and adult in response to the individual stimuli. In line with Table 1 column 4, the number of negative AUs in response to mango and apple was higher for children compared with adults. In addition, the number of negative AUs in response to broccoli was higher for adults.

Fig. 4.

Negative facial responsiveness to retronasal odors. Total number of negative AUs observed for each subject, averaged across children) and adults. Error bars represent standard error of the mean. Adults showed fewer total negative AUs than children in response to apple (t = −3.21, P < 0.001) and mango (t = −1.97, P < 0.05), and more negative AUs in response to broccoli (t = 2.14, P < 0.05).

3.4.3. Individual negative AU occurrence for each retronasal odor

Next, we further broke down the data, comparing the occurrence of each negative AU between adults and children for individual stimuli. Figure 5 shows the result of this analysis for each odor stimulus. Critically, this analysis reduces the impact of highly frequent AUs (e.g. AU 4, 10, 15) by considering only the difference in occurrence between age groups. For each stimulus, there was an overall strong, positive correlation between the occurrence of individual negative AUs between children and adults. However, occurrence values were biased in a stimulus-specific manner. For apple and mango odors, negative AUs occurred more frequently in children; for broccoli and pumpkin odors, negative AUs occurred more frequently in adults. Figure 5e summarizes these findings by showing the average difference in occurrence of individual negative AUs between children and adults. Table 2 provides a quantitative comparison for each individual stimulus using t-test. In addition, we considered an effect of odor category. Apple and mango are considered less negative by adults than broccoli and pumpkin, as evidenced by the finding that the adult subjects showed significantly fewer negative AUs in response to mango and apple as compared to broccoli and pumpkin (one-way ANOVA on the number of negative AUs per subject with factor Odor Category [positive, negative]: F = 32.67, P < 0.001; see coral bars in Fig. 4). Based on perceptual patterns observed in adult subjects (Figs. 4 and  6), broccoli and pumpkin were classified as negative odors, and mango and apple were classified as positive. Thus, we ran a 2-way mixed ANOVA with factors Odor Category (positive, negative) and Age Group (children, adults) on the occurrence values for individual negative AUs. We observed a significant main effect of Odor Category (F = 19.78, P < 0.001) and a significant interaction between Odor Category and Age Group (F = 8.14, P < 0.01), indicating that differences in the expression of negative AUs between children and adults depend on odor identity. For comparison, Fig. 5f shows the results of the same analysis performed on positive AUs. For positive AUs, we only observed a main effect of Age Group (F = 5.17, P < 0.05), again confirming that positive AUs occurred more frequently in children compared with adults regardless of stimulus identity.

Fig. 5.

Facial response patterns to retronasal odors. a to d). Scatterplots showing occurrence of individual negative AUs in children versus adults for broccoli a), pumpkin b), apple c) and mango d). Occurrence of individual AUs was highly correlated between children and adults for all stimuli, but the distribution of individual AUs varied by odor category. e, f) Average difference in occurrence of individual negative AUs e) and positive AUs f) between children and adults. Error bars represent standard error of the mean. Negative odors evoke more individual negative AUs in adults while positive odors evoke more individual negative AUs in children (F = 19.78, P < 0.01). Children showed more individual positive AUs than adults in response to both negative and positive odors (F = 5.17, P < 0.05).

Table 2.

Average difference in occurrence of individual AUs between adults and children.

	Difference in negative AU occurrence	t-test (df = 11)	Difference in positive AU occurrence	t-test (df = 4)
Sucrose	0.002 ± 0.02	t = −0.13, P = 0.89	−0.03 ± 0.03	t = 1.11, P = 0.32
Quinine	0.06 ± 0.02	t = −2.15, P = 0.05	−0.04 ± 0.01	t = 2.41, P = 0.07
Apple	−0.06 ± 0.02	t = 2.88, P = 0.01	−0.30 ± 0.03	t = 1.23, P = 0.28
Mango	−0.05 ± 0.02	t = 2.38, P = 0.03	−0.01 ± 0.01	t = 0.99, P = 0.37
Broccoli	0.05 ± 0.02	t = −1.93, P = 0.07	−0.03 ± 0.02	t = 1.43, P = 0.22
Pumpkin	0.03 ± 0.01	t = −1.87, P = 0.08	−0.03 ± 0.01	t = 2.01, P = 0.11

Fig. 6.

Explicit ratings of taste and retronasal odor stimuli. Proportion of subjects who rated a given stimulus as “yucky” for taste (a) and retronasal odor (b) solutions. Children (n = 81); adults (n = 84). Compared with adults, a higher proportion of children rated mango as yucky (Χ² = 7.27, P < 0.01) and a lower proportion of children rated broccoli (Χ² = 7.84, P < 0.01) and pumpkin (Χ² = 7.57, P < 0.01) as yucky.

To gain further insight into how negative expressions in particular change between stimuli, we directly compared occurrence of individual negative AUs between stimuli within age groups. Neither age group showed significant differences in expression of any of the negative AUs between stimuli of the same category (positive: apple versus mango; negative: broccoli versus pumpkin). We then grouped occurrence of individual negative AUs across stimuli within category and compared expression patterns between negative and positive categories. Table 3 shows the results of this comparison: adults showed differences in occurrence of multiple negative AUs (AU 1, 2, 4, 10, 15, 20, 43, and 84 are more commonly expressed in response to negative odors). In contrast, no differences in expression between the 2 odor categories was observed in children. Together with the findings reported in Figs. 4 and 5, this suggest that the hedonic value of retronasal odors differs between children and adults, and that this difference between the 2 age groups is due to an increase in specificity of the expressions elicited by different odorants with age.

Table 3.

Proportion of individual negative AUs in response to positive and negative odors.

AU	Proportion of adults (Positive odor, Negative odor)	Χ 2	P	Proportion of children (Positive odor, Negative odor)	Χ 2	P
1	0.07, 0.21	11.65	<0.001	0.13, 0.09	1.34	0.24
2	0.06, 0.14	4.79	<0.05	0.09, 0.06	0.48	0.48
4	0.16, 0.31	9.57	<0.01	0.34, 0.36	0.14	0.70
9	0.01, 0.03	0.40	0.68	0.01, 0.01	0.00	1.00
10	0.18, 0.39	14.56	<0.001	0.38, 0.44	0.96	0.32
15	0.24, 0.40	8.58	<0.01	0.36, 0.39	0.14	0.71
20	0.12, 0.24	6.91	<0.01	0.17, 0.15	0.24	0.62
24	0.06, 0.10	0.75	0.38	0.12, 0.15	0.51	0.47
25  +  26  +  27	0.00, 0.01	0.50	0.48	0.04, 0.01	0.58	0.44
43	0.04, 0.13	7.69	<0.01	0.03, 0.01	0.17	0.68
45	0.04, 0.09	2.90	0.08	0.02, 0.03	0.00	1.00
84	0.02, 0.12	9.09	<0.01	0.04, 0.01	0.58	0.44

3.5. Differences in explicit ratings of flavor solutions between children and adults

Our findings based on analysis of facial expressions suggest differences in how children perceive retronasal odor stimuli as compared to adults. However, as for any single measure of perception, facial expressions provide limited insight: not all subjects produce facial expressions; adults and children differ not only in what they express but also in how they express; and stimulus-specific differences between children and adults were observed only for a subset of negative expressions. We therefore used a pictorial rating scale as a second, independent measure of hedonic perception in a subset of subjects (n = 81 children, 84 adults). After sampling each solution, subjects were first asked whether it tasted “like water” or “different from water”. Table 4 shows the proportion of subjects who indicated that the solution tasted “different from water” for each stimulus. Compared with adults, fewer children rated odor solutions as “different from water”. If subjects indicated that the solution tasted “different from water”, they were then asked to point to a face to indicate whether the solution was “yummy” or “yucky”. Overall, ratings were consistent with facial expressions. Subjects who gave “yucky” (as compared to “yummy”) ratings were more likely to show any negative AU (Χ² = 20.17, P < 0.001) and less likely to show any positive AU (Χ² = 11.44, P < 0.001). This pattern held true for children, adults, taste stimuli and odor stimuli. Figure 6 shows hedonic ratings for taste (Fig. 6a) and retronasal odor solutions (Fig. 6b). The vast majority of subjects rated quinine as “yucky”; sucrose as “yummy”. There were no differences in ratings for these tastes between children and adults. For retronasal odor stimuli, rating patterns mimicked the pattern obtained from facial expressions: compared with adults, children showed fewer “yucky” ratings for broccoli and pumpkin (Χ² = 14.55, P < 0.001), and more “yucky” ratings for apple and mango (Χ² = 10.30, P < 0.001). These results are thus in line with the idea that hedonic valuation of retronasal odor stimuli is less stimulus-specific in children as compared to adults.

Table 4.

Proportion of subjects who detected each stimulus as different from water.

	Proportion of subjects who rated stimuli as “different from water” (Adults, Children)	Χ 2	P-value
Sucrose	1.00, 0.92	4.51	0.03
Quinine	0.98, 0.93	0.32	0.56
Apple	0.97, 0.68	23.78	<0.001
Mango	0.98, 0.75	9.53	<0.01
Broccoli	0.98, 0.79	13.43	<0.001
Pumpkin	1.00, 0.70	13.55	<0.001

4. Discussion

In this cross-sectional study, we used facial expressions and a pictorial rating scale as age-appropriate, nonverbal measures of hedonic perception of gustatory and retronasal odor solutions in 3 to 6-year-old children and adults. The results show that facial responses and pictorial ratings evoked by retronasal odor stimuli differ between children and adults in a stimulus-specific manner, supporting our hypothesis that hedonic perception of odor components of flavor is not inborn but differs between children and adults. We also observed overall differences in the specific facial responses used by children and adults, indicating expressive differences that are independent of stimulus identity.

The observed perceptual patterns for odor solutions in young children are characterized by a lack of stimulus-specificity compared with adults. Adults consistently displayed more negative facial responses and ratings to broccoli and pumpkin odors as compared to apple and mango odors; children show less differentiated responses between odors. This finding stands in contrast with a previous study using orthonasal odors showing that children 3 years of age exhibit adult-like preferences for wintergreen, floral, strong cheese/vomit, and strawberry odors (Schmidt and Beauchamp 1988). In this study, Schmidt and Beauchamp used a rating procedure where subjects played “a smell game”. They were instructed to give good odors to Big Bird and bad odors to Oscar the Grouch from “Sesame Street”. Although they concluded that odor preferences were similar between the 2 age groups, they did observe differences in liking for some odors, such as banana. The lack of adult-like hedonic perception of odors observed in the present study is consistent with research indicating that odor preferences are not innate. Soussignan et al. showed that newborns did not exhibit consistent responses to pleasant and unpleasant odors (Soussignan et al. 1997), while another study by the same investigators demonstrated that 5 to 12-year-old children showed predictable responses and ratings for fishy, fecal, fruity, and floral odors (Soussignan and Schall 1996). However, these results were based on orthonasal odors and a limited sample of subjects (N = 16 to 50 children). The fact that odors in the present study were presented in the context of consumption may explain differences in the development of hedonic odor perception reported in the literature. The same odor can be experienced differently via the orthonasal versus the retronasal route (Pierce and Halpern 1996; Hannum et al. 2021; Pellegrino et al. 2021), and retronasal odors activate a unique neural system that includes the gustatory cortex (de Araujo et al. 2003; Landis et al. 2005; Small et al. 2005; Blankenship et al. 2019). Future work will direct compare behavioral responses to the same odors presented ortho- versus retronasally in different age groups. It is also unknown how our findings generalize to other odorants. Our design—and work with young children in general—precludes testing a large number of stimuli, and we limited our stimulus set to a few common food odors. Future work will test a larger variety of food and nonfood odors that vary in chemical structure and familiarity.

One explanation for age-related differences in hedonic retronasal odor perception is experience. Such an explanation is generally consistent with the idea that odor perception undergoes extensive experience-dependent plasticity throughout life (Li et al. 2008; Beauchamp and Mennella 2009; Hummel et al. 2009; Poncelet et al. 2010; Huart et al. 2019; Honnens de Lichtenberg Broge et al. 2023; Maier and Zhang 2023). In the context of consumption, limited exposure to a wide variety of foods in children could explain differences in perception between the 2 age groups observed here. However, it remains unclear how experience may shape retronasal odor perception. Mere exposure to odorants in the context of amniotic fluid and breast milk or formula is known to shape hedonic perception of food odors during fetal and early postnatal life, suggesting continuity between these developmental phases (Forestell and Mennella 2007; Mennella et al. 2017; Ustun et al. 2022). However, plasticity mechanisms at play during fetal or infant developmental stages may differ from those in toddlers and young children. Once children gain autonomy over food intake, the specific external context in which retronasal odors are experienced may become an important factor in shaping perception. For example, while mango and apple flavors are often consumed in smoothies and juices, broccoli and pumpkin flavors are rarely encountered in pure liquid form. Thus, consumption over the course of a lifetime may have rendered broccoli and pumpkin unpleasant in this context for adults. Another possibility is that retronasal odors apple and mango are often experienced in the context of sugar and may have become associated with sweetness in adults. It is to date unknown when and how taste-odor association learning develops across the lifespan. It is also important to note that “positive” and “negative” labels applied to these odors are specific to the current (liquid) context as they are based on perceptual patterns observed in our adult sample; they should not be taken to be universal labels for these odors or the foods they are associated with. Future research will focus on examining the role of experience and context in shaping hedonic perception of retronasal odors, as well as determining how experience is shaped by cultural differences and interacts with genetic factors. Parental eating habits and foods offered at home should be considered as factors impacting experience with food odors. Cultural differences may influence familiarity with and liking for specific food odors (Gotow et al. 2021). For orthonasal odors, there is evidence that odors show consistent pleasantness across cultures (Arshamian et al. 2022; Oleszkiewicz et al. 2022), but whether this extends to retronasal odor perception and is generalizable across odorants remains unknown. Genetic factors, such as TAS2R38 polymorphisms, could also play a role in individual differences in retronasal odor perception. For example, the genetic ability to perceive the bitter compounds 6-n-propylthiouracil (PROP) and phenylthiocarbamide (PTC) may influence how people perceive odors associated with bitter foods (Mastinu et al. 2023). Finally, a change in retronasal odor sensitivity may contribute to the age-related differences observed here. Although our stimuli were above threshold for the vast majority of subjects, children gave a higher percentage of “tastes like water” responses compared with adults for odor solutions in particular, suggesting that children may have lower retronasal odor sensitivity. The literature comparing orthonasal odor sensitivity between children and adults is mixed (Cain et al. 1995; Monnery-Patris et al. 2009; Cameron and Doty 2013). Methods used to assess odor sensitivity tend to rely on verbal and cognitive skills, which improve with age, making it difficult to distinguish differences in sensitivity from differences in cognitive development. We may observe more “tastes like water” responses in children due to cognitive differences.

Perceptual patterns observed for odor solutions differed markedly from those observed in response to taste solutions. Regarding taste solutions, our findings are consistent with prior work demonstrating that the hedonic value of taste qualities is inborn and relatively stable across the lifespan (Steiner 1973; Rosenstein and Oster 1988; Beauchamp and Mennella 2011; Aziza et al. 2024). Together with the present findings, this demonstrates that hedonic evaluation of taste and odor components of flavor undergo different developmental trajectories, with taste perception having a strong innate component whereas retronasal odor perception appears more plastic.

The results we obtained from facial expressions were consistent with the results obtained from explicit ratings. Consistency of these 2 independent measures of perception provides strong support for our conclusion that hedonic odor perception differs between young children and adults. Previous work has successfully used facial expressions to assess taste, orthonasal odor, and multisensory flavor preferences in infants (Steiner 1974; Soussignan et al. 1997; Schaal et al. 2000; Mennella et al. 2009), children ages 5 to 13 (Tolia et al. 2004; Zeinstra et al. 2009; De Wijk et al. 2012; Galler et al. 2022) and adults (Weiland et al. 2010; He et al. 2014; Leitch et al. 2015). Pictorial rating scales have also been used to measure gustatory and olfactory perception in 3 to 6 yr old children (Chen et al. 1996; Visser et al. 2000; Yuan et al. 2016; Sorokowska et al. 2024). Our results add further support for the validity of both methods for measuring flavor perception across the lifespan. An added advantage is that both methods are readily implemented at low cost. However, it is possible that immediate, reflexive responses (facial expressions) and explicit, deliberate judgement (rating scale) capture to some extent different aspects of perception. Future work will identify similarities and differences in the particular aspects of perception captured by each method.

Apart from differences in the perception of individual odors between the 2 age groups, we observed differences in facial expressiveness that were independent of flavor identity. Children generally showed more facial expressions than adults, an effect that was primarily attributed to an increase in positive facial expressions. One possible explanation for the finding that positive AUs are more common in children is that this age group overall valued the stimuli more positively than adults. However, we did not observe a parallel decrease in occurrence of negative AUs. Together, these findings suggest that children (compared with adults) are more likely to express positive affect, but not negative affect. The difference in positive expressiveness between children and adults may at least partially be explained by suppression of positive affect in adults. A previous study reported that adults responded more to unpleasant compared with pleasant orthonasal odors (Weiland et al. 2010), and that positive facial expressions to liked juices were suppressed by adults unless they were explicitly instructed to make facial expressions (Danner et al. 2014a), or unaware that their facial expressions were being recorded (Danner et al. 2014b). Subjects in our study were not instructed to produce facial expressions and were aware that they were being recorded, although they were never explicitly told that recordings would be used to examine facial expressions. Children in our study may also have been differentially influenced by the presence of the experimenter, resulting in more positive facial expressions to please the experimenter or produce the “desired” response, as children in our sample's age range begin to develop the ability to mask responses to unpleasant stimuli by smiling (Soussignan and Schaal 1996; Soussignan and Schall 1996). This is consistent with the research demonstrating that children will also show more positive facial expressions in other social situations, such as in response to disappointment (Cole 1986). Regarding positive facial expressions in children, a previous study by Zeinstra et al. examined facial responses to juices, milk, and gustatory solutions in a small sample of children ages 5 to 16 (N = 6) (2009). They reported that while disliked stimuli yielded more negative than positive expressions, liked flavor stimuli yielded equal numbers of positive and negative expressions. Thus, together with the extant literature, our findings suggest that positive facial expressions are not an effective tool for comparing flavor perception between age groups and may be a poor indicator of hedonic perception in general.

Regarding negative AUs, we show that when collapsing over all stimuli, some negative AUs occurred more in adults, while others occurred more in children. With the exception of AU 4 (eyebrow lower), all negative AUs that occurred more frequently in children affected the inferior facial muscles, while all AUs that occurred more frequently in adults involved the temporal facial muscles and movement of the head. These differences may reflect differences in anatomy, social development or perception between the 2 age groups. Soussignan and Schaal found that in 5 to 12 yr old children, there was a difference in facial responses to orthonasal odors depending on whether the children were alone or in the presence of a female researcher (1996). They noted that the muscles of the lower/middle face were more controlled in a social condition, while muscles of the upper face were more difficult to control. In the present study, subjects sampled flavors in the presence of a female researcher. We may see more negative AUs of the lower face in children compared with adults because adults are more able to control these movements in a social setting. Upper facial movements may be observed more in adults because they are difficult to control even in the social setting, combined with the fact that muscles in adults are generally more developed. AUs can reflect other cognitive processes besides preference. For example, AUs that occur more in adults (e.g. AU 1, AU 2, and AU 45 [inner and outer eyebrow raises; blink]) may reflect surprise or attention to a novel stimulus rather than disgust (Delplanque et al. 2009; Gosselin et al. 2010; Miao et al. 2023). Adults may show more surprise because there is no color or orthonasal odor to create an expectation of the flavor they are about to experience. In addition, movements that are differentially expressed by the 2 age groups may reflect differences in familiarity or sensitivity, as discussed above.

Our findings regarding the development of multisensory aspects of flavor perception have important implications for designing interventions aimed at altering food choice behavior, identifying the olfactory system as a particularly promising target. Unlike taste, hedonic perception of the retronasal odor component of flavor appears to be malleable based on differences between children and adults. Repeated exposure is often recommended to parents as a way to increase food acceptance in children (Appleton et al. 2018). Interventions have shown success in increasing vegetable consumption through exposure. Forestell and Mennella implemented an 8 d intervention for 4 to 8 mo infants to increase vegetable acceptance (Forestell and Mennella 2007). One group was fed green beans each day, while another group was given green beans followed by peaches. After the intervention, green bean intake increased for both groups, but only the group that was fed green beans followed by peaches showed fewer negative facial expressions during consumption, possibly due to association with the sweet taste of peaches. A longer exposure period or association with another positive context than sweet taste might increase liking for children fed vegetables alone. Dazeley and Houston-Price conducted a 4-week intervention for 1 to 3-year-old children which explored the effect of exposure to nontaste sensory qualities of unfamiliar fruits and vegetables during playtime (Dazeley and Houston-Price 2015). Following the intervention, children tasted more of the vegetables they had been familiarized with during playtime. This further demonstrates how exposure to sensory characteristics of foods can improve acceptance. Designing targeted interventions across a larger range of ages will require a more complete understanding of the plasticity mechanisms that shape hedonic retronasal odor perception at different phases of development.

5. Conclusion

Retronasal odor perception in the context of consumption differs between 3 to 6 yr old children and adults. Future work will examine the mechanisms that contribute to these differences, including experience, contextual, cultural and genetic factors, as well as perceptual and cognitive processes besides hedonic evaluation.

Funding

This project was supported by the National Institute on Deafness and Other Communication Disorders, F31DC022149 (SEC) and R01DC020212 (JXM), and the National Center for Advancing Translational Sciences of the National Institutes of Health under Award Number UL1TR001420 (JXM).

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.