Research note 009 — 2026-08-30 (expanded 2026-09-08)
The Talent for Learning to Love What's Bitter
Newborn babies grimace at bitterness. Rejecting bitterness — an alarm against poison — is the human default setting. And yet adults pay money to bring coffee, beer, and bitter gourd to their mouths — somewhere, a piece of learning must be rewriting that default setting. This note lays out what is known, and what remains contested, about the mechanism behind that rewrite.
bitterness is rejected more strongly in childhood than in adulthood: Newborns show a stereotyped rejection expression in response to bitter solutions. Children with bitter-sensitive TAS2R38 genotypes are more sensitive to PROP bitterness than adults with the same genotype.supported
Scope: US psychophysical studies and genotype-phenotype studies synthesized in the cited review
Note: That sensitivity changes with age even within the same genotype is the developmental background for 『獲得嗜好』 (in translation: "acquired taste"). The timing of the shift to the adult pattern (adolescence) is the review's summary phrasing, and the numbers from individual studies have not been confirmed.
Verbatim confirmed: newborns 'gape, wrinkle their noses, shake their heads, flail their arms, and frown when a bitter-tasting solution is placed in the oral cavity', and 'children with bitter-sensitive TAS2R38 genotypes were more sensitive to the bitter taste of PROP than were adults with the same genotype'. The review frames childhood bitter rejection as a defense against poison ingestion.
Limitations: A review-level synthesis. The dating of the shift to adolescence is the review's summary, and the numbers from primary studies are not yet in this database. The sentence 'declining to adult levels during middle to late adolescence' is about sweet preference, not bitterness.
Verbatim: "The taste of sucrose elicited homologous positive hedonic patterns of facial affective reactions from humans and other primates, whereas quinine elicited homologous aversive or negative affective patterns." 8 human newborns showed aversive affective reactions to quinine, independently confirming stereotyped newborn rejection of bitterness, and positioning it as phylogenetically conserved across 11 primate species.
Limitations: Supports only the newborn-rejection part of this claim; includes neither the child-adult psychophysical comparison nor TAS2R38 genotype data. The human sample is small (n = 8).
Show datasensation-bitter —is_rejected_more_strongly_in_childhood_than_adulthood→ "Newborns show a stereotyped rejection expression in response to bitter solutions. Children with bitter-sensitive TAS2R38 genotypes are more sensitive to PROP bitterness than adults with the same genotype."
Newborns' rejection response of grimacing at bitter liquids has been confirmed across multiple studies.
That liking for bitter vegetables and beverages rises with repeated intake has also been confirmed in a meta-analysis.
The reversal in which people more sensitive to coffee's bitterness drink more of it has been found in an analysis of genetic data.
Whether that reversal is the result of learning has not yet been confirmed.
What is established: repeated encounters breed liking
For children's vegetables, the evidence is thickest for simple repeated exposure. A meta-analysis confirms that both liking and intake rise. Even in adults, there is an experiment in which simply continuing to drink a bitter beverage for 7 days raised its pleasantness ratings. Flavors the mother consumed during pregnancy and breastfeeding increase an infant's acceptance during weaning.
repeated dietary exposure increases Liking of vegetables (SMD 0.35 [0.17, 0.53]) and intake (SMD 0.23 [0.07, 0.39]), compared with no exposure. About 0.6 points on a 9-point liking scalesupported
Scope: A meta-analysis of RCT comparisons (43 papers, 117 comparisons). Mainly infants and young children
Note: The effect size is small, and many of the original studies are at high risk of bias. Bitterness is mentioned as a major factor in vegetable rejection, but there is no analysis stratified by bitterness intensity.
Repeated exposure raised post-intervention liking compared with no exposure (SMD 0.35 [0.17, 0.53]; 17 comparisons, 820 participants), and also raised intake (SMD 0.23 [0.07, 0.39]; 21 comparisons, 1130 participants). This corresponds to about 0.6 points of liking on a 9-point scale, and about 10 g of vegetable.
Limitations: Most of the included studies are small and at high risk of bias. Heterogeneity is low (I2 25-29%), but the effect size is small.
Limits: The authors state verbatim: "Effect sizes, however, are small, and limited evidence suggests long-term benefits." Follow-up after the intervention was often limited.
Limitations: Follow-up periods were inconsistent; the authors themselves acknowledge limitations in adherence and reporting.
8 daily tasting sessions of sweet red pepper significantly increased liking (P = 0.006) and intake (P = 0.03) compared with a no-treatment control. The sticker-reward condition was intermediate, not differing significantly from either group.
Limitations: A single vegetable (sweet red pepper, not strongly bitter). Small sample. Abstract-level reading only.
Show dataprocess-repeated-exposure —increases→ "Liking of vegetables (SMD 0.35 [0.17, 0.53]) and intake (SMD 0.23 [0.07, 0.39]), compared with no exposure. About 0.6 points on a 9-point liking scale"
repeated dietary exposure increases Adult hedonic ratings of a bittersweet beverage (+68% after once-daily consumption at home for 7 days). Meanwhile, a control stimulus tasted only in the laboratory did not changesupported
Scope: Adult volunteers, one commercially available bittersweet beverage (Stein 2003)
Note: A single study. The beverage's composition (presence or absence of caffeine) and n are not reported in the abstract, and the mechanistic attribution (post-ingestive association) remains the authors' speculation.
After consuming a commercially available bittersweet beverage once daily at home for 7 days, hedonic ratings of that beverage rose by 68%, while ratings of a control stimulus tasted only in the laboratory did not change; health-related information did not affect the change in hedonic ratings.
Limitations: The abstract shows neither the beverage's composition (caffeine content) nor the sample size; a single study; the mechanistic attribution is the authors' speculation.
Context: In 20 moderate caffeine consumers, pleasantness ratings of a novel flavored beverage paired with caffeine (CS+) rose gradually across 8 breakfast conditioning trials, while the pleasantness of a beverage paired with placebo (CS-) did not change. The authors conclude that this change 'are best explained in terms of Pavlovian associations between drink flavour and the postingestive effects of caffeine', providing direct experimental support for the post-ingestive mechanism that Stein 2003 could only speculate about.
Limitations: Since the abstract does not describe this beverage as bitter, this does not reproduce the result for bittersweet beverages, but rather contextualizes the acquisition mechanism for caffeinated beverages in general. Despite the change in pleasantness, overall preference did not change.
Limits: Health-related information did not affect the perceptual (hedonic) change accompanying exposure itself, but tended to somewhat increase the behavioral measure of acceptability. The authors interpret this as information potentially acting more strongly on behavior than on hedonic evaluation.
Limitations: Stops at a 『傾向』 (in translation: "tendency"), and the abstract does not report statistical significance. The definition and measurement method of the behavioral measure (acceptability) are also absent from the abstract. Full text not reached (Elsevier/Appetite, Unpaywall closed, no repository copy).
Show dataprocess-repeated-exposure —increases→ "Adult hedonic ratings of a bittersweet beverage (+68% after once-daily consumption at home for 7 days). Meanwhile, a control stimulus tasted only in the laboratory did not change"
perinatal flavor exposure increases Infant acceptance of an exposed flavor during weaning: groups exposed either prenatally or postnatally showed fewer negative facial expressions to carrot-flavored cereal than to plain cerealsupported
Scope: Randomized 3-group design, 47 mother-infant pairs who completed the study. During the last trimester of pregnancy or the first 2 months of breastfeeding, 300 mL of carrot juice 4 days a week for 3 weeks (Mennella 2001)
Note: The target is carrot flavor, not bitterness (no mention of bitter in the text). The group × flavor interaction for intake and feeding duration was not significant. Positioned as contextual material connecting to the acquired-taste line as a foundation for flavor learning in general.
Infants exposed to carrot flavor prenatally (paired t(14) = 2.88, P =.01) or via breast milk (paired t(16) = 2.41, P =.03) showed fewer negative facial expressions to carrot-flavored cereal than to plain cereal, whereas the unexposed control group did not. A group × flavor interaction was found in mothers' enjoyment ratings (F(2,43) = 3.37, P =.04).
Limitations: The groups are small. 7 mothers weaned before testing. Transmission of carrot flavor via amniotic fluid is inference, not demonstration.
Limits: The group × flavor interaction was not significant for cereal intake (F(2,43) = 1.09, P =.34) or feeding time (F(2,43) = 0.88, P =.42), and since the paper does not characterize carrot as bitter, extending this flavor-learning result to bitter foods is inference beyond the paper.
Limitations: The acceptance effect is carried by facial expressions and maternal ratings, not by measured intake.
Show dataprocess-perinatal-flavor-exposure —increases→ "Infant acceptance of an exposed flavor during weaning: groups exposed either prenatally or postnatally showed fewer negative facial expressions to carrot-flavored cereal than to plain cereal"
This rewriting may begin even before birth. In a longitudinal study in which mothers in late pregnancy repeatedly consumed capsules of bitter kale flavor, fetuses tended to show more negative facial expressions to the kale flavor than to the non-bitter carrot flavor. The same children were followed up to age 3, and when made to smell kale and carrot again, negative facial expressions to the odor they had been exposed to prenatally were fewer than to the odor they had not been exposed to — a within-subject comparison. 12 children could be followed.
perinatal flavor exposure increases Across multiple measurement points from the fetal period (32 and 36 weeks of pregnancy) through after birth (3 weeks and 3 years), the bitter kale-flavored stimulus tended to elicit significantly more negative facial expressions (cry-face gestalt) than the non-bitter carrot-flavored stimulus (p<0.001).supported
Scope: Stratum: food. A single flavor-capsule stimulus (kale or carrot) to the fetus at 32 and 36 weeks of pregnancy, repeated intake from week 36 to delivery, and longitudinal follow-up through 3 weeks and 3 years after birth (the number of codable data points varies by measurement point; final follow-up n=12).
"fetuses in the third trimester exhibit differential facial reactions to a bitter (kale) and a non-bitter (carrot) flavor... Across all measurement points, there is a general tendency for kale flavor to produce more cry-face gestalts (p < 0.001; Table 6)."
Limitations: The number of participants who could be followed up to age 3 had decreased to 12, a small sample size. The authors report that maternal mental health status (anxiety, depression, stress) can also affect facial expressiveness.
Show dataprocess-perinatal-flavor-exposure —increases→ "Across multiple measurement points from the fetal period (32 and 36 weeks of pregnancy) through after birth (3 weeks and 3 years), the bitter kale-flavored stimulus tended to elicit significantly more negative facial expressions (cry-face gestalt) than the non-bitter carrot-flavored stimulus (p<0.001)."
perinatal flavor exposure reduces When kale and carrot odors were presented at age 3 to 12 children of mothers who had repeatedly consumed kale (bitter) flavor in late pregnancy, within the same children, negative facial expressions to the odor of the flavor they had been repeatedly exposed to prenatally were significantly fewer than negative facial expressions to the odor of the flavor they had not been exposed to (p<0.001). The authors interpret this as evidence that fetal chemosensory memory persists for years after birth.supported
Scope: Stratum: food. Repeated intake of kale- or carrot-flavored capsules in late pregnancy (from week 36 to delivery); an olfactory presentation test with a within-subject comparison presenting both odors at age 3 (n=12).
"the 3-year-olds exhibited a significantly reduced rate of negative facial expressions in response to the odor they had been repeatedly exposed to in utero (p < 0.001)... we observed a significant reduction in facial expressions, which could be interpreted as rejection of that vegetable odor when the child had been exposed to it repeatedly in the late stages of pregnancy."
Limitations: Only 12 participants could be followed up to age 3, a small number. The comparison is a within-subject comparison presenting both odors to the same child, and the number of participants per exposure group is not reported in the text.
Show dataprocess-perinatal-flavor-exposure —reduces→ "When kale and carrot odors were presented at age 3 to 12 children of mothers who had repeatedly consumed kale (bitter) flavor in late pregnancy, within the same children, negative facial expressions to the odor of the flavor they had been repeatedly exposed to prenatally were significantly fewer than negative facial expressions to the odor of the flavor they had not been exposed to (p<0.001). The authors interpret this as evidence that fetal chemosensory memory persists for years after birth."
What is contested: 「慣れ」 (in translation: "habituation") or 「ご褒美」 (in translation: "reward")?
So is coming to like coffee's bitterness also just 「慣れ」 (in translation: "habituation")? Experiments in the caffeine domain paint a different picture — a new flavor comes to be liked when it is repeatedly paired with caffeine, and that liking only shows up in a state where caffeine has worn off. In other words, it may not be bitterness itself that becomes liked, but the flavor becoming liked as a signal foretelling a reward. Vegetables, on the other hand, respond to exposure alone, with no reward required. Whether 「獲得嗜好に報酬は必要か」 (in translation: "does acquired liking require a reward?") is, at this research lab, still contested.
caffeine conditions liking for paired flavors: In moderate caffeine consumers who abstained overnight, an increase in liking for a novel beverage flavor repeatedly paired with caffeinesupported
Scope: Double-blind human flavor-conditioning studies (University of Sussex group)
Note: The experimental basis for the 『連合学習』 (in translation: "associative learning") theory of coffee and beer preference. Note that replication remains within the same research group.
Liking for a novel-flavored caffeinated beverage, consumed after 12 hours of caffeine abstinence, increased across 4 days of training in 36 moderate caffeine consumers, and was maintained only in the group that continued to receive caffeine during the abstinence state.
Limitations: Moderate consumers only. Prior research cited in the abstract shows that conditioning does not occur in low consumers. A single research group.
Comfort/pleasantness ratings of a novel beverage containing 100 mg of caffeine rose over 4 days of testing in consumers who maintained caffeine intake, but fell significantly in consumers who abstained from caffeine for 4 weeks.
Limitations: An abstract-level reading; part of the same Sussex research lineage as Yeomans 2000, and not a fully independent replication.
Using an 「生態学的に妥当な」 (in translation: "ecologically valid") procedure — overnight caffeine abstinence plus 70 mg at breakfast (equivalent to one cup of coffee) — moderate consumers developed a relative dislike for the caffeine-free beverage. Overnight abstinence lowered mood (less lively, clearheaded and cheerful), and caffeine significantly improved it. Low consumers showed no significant effect on either preference or mood. The authors conclude that "caffeine reinforcement can occur as the result of the alleviation by caffeine of the adverse effects of overnight caffeine abstinence (negative reinforcement)". A foundational report using the same-type paradigm preceding Yeomans 2000 and Tinley 2003, also directly showing dependence on consumption history (no conditioning in low consumers).
Limitations: Abstract only confirmed. n and number of training days are not reported in the abstract. The Rogers group at Reading later co-authored Yeomans 2000, so this is not a fully independent lab replication from the Sussex lineage (an earlier report from the same lineage at a different institution).
Context: Circumstantial support from the genetic epidemiology side: shows that variants in caffeine metabolism/sensitivity systems are more strongly linked to coffee taste preference (including presence or absence of milk/sweetener) than variants in the bitter-perception system, and the authors give the same interpretation to population genetic data as laboratory flavor-conditioning studies (citing the Yeomans lineage): "Taste preferences and physiological caffeine effects intertwine in a way that is difficult to distinguish for individuals which may represent conditioned taste preferences".
Limitations: An interpretation from observational genetic data, not a direct test of the conditioning process. A discussion that does not go beyond "may represent".
Show datacompound-caffeine —conditions_liking_for_paired_flavors→ "In moderate caffeine consumers who abstained overnight, an increase in liking for a novel beverage flavor repeatedly paired with caffeine"
caffeine produces conditioned liking that depends on Caffeine deprivation state: the conditioned preference is not expressed when tested in a non-deprived state, is extinguished without caffeine, and reverses in consumers who are fully withdrawnsupported
Scope: The same conditioning paradigm. The 4-week maintenance/withdrawal manipulation in Tinley 2003
Note: Supports the negative reinforcement theory. A strong limitation on the naive interpretation of 『苦味自体を安定的に好きになる』 (in translation: "coming to stably like bitterness itself").
Both groups tested in the non-deprived state showed a marked decline in pleasantness across all 4 test days compared with the final training day, and the deprived group without caffeine showed extinction. The authors conclude that the conditioned flavor preference 'may not be expressed in the absence of the relevant motivational state (caffeine deprivation)'.
Limitations: Training is short (4 days). It is the flavor of a novel drink, not coffee itself.
Giving a novel caffeinated beverage to consumers placed in full withdrawal lowered pleasantness. The authors read this as showing that caffeine's reinforcement of flavor liking 'is driven by the alleviation of withdrawal symptoms among habitual caffeine consumers' (negative reinforcement theory).
Limitations: This does not directly test long-term coffee drinkers' liking for coffee's bitterness. An abstract-level reading.
Shows that the establishment of the conditioned preference itself is explained as 「一晩のカフェイン断ちによる負の状態のカフェインによる緩和」 (in translation: "caffeine's alleviation of the negative state caused by overnight caffeine abstinence") (negative reinforcement), and that low consumers, who do not enter a withdrawal state, show neither a preference change nor a mood effect. A report from a different institution, preceding Tinley 2003's 4-week withdrawal manipulation, showing the core of this claim — that the conditioned preference requires the motivational state of caffeine withdrawal.
Limitations: Abstract only confirmed. The demonstration of the difference in expression from the deprived/non-deprived state manipulation at test is not this study's design but Yeomans 2000's; this report shows state-dependence on the acquisition side.
Context: A double-blind crossover from a lab (Johns Hopkins) fully independent of the Sussex/Reading conditioning lineage. Under about 19 hours of intake restriction, moderate consumers (mean 379 mg/day) showed a choice of forgoing a mean of $2.51 to avoid placebo (i.e., withdrawal), while paying only a mean of $0.38 to reacquire caffeine. Headache intensity correlated significantly with the placebo-avoidance amount. The authors conclude "choice of caffeine is more potently controlled by avoiding withdrawal than it is by the positive effects of caffeine", confirming in a different paradigm (drug choice) that negative reinforcement is the main driver of caffeine reinforcement.
Limitations: Abstract only confirmed. n is not reported in the abstract. Taste/flavor preference was not measured, so this is not a replication of the conditioned preference itself (circumstantial support on the mechanism side). Intake was via capsule, not in a beverage context.
Show datacompound-caffeine —produces_conditioned_liking_dependent_on→ "Caffeine deprivation state: the conditioned preference is not expressed when tested in a non-deprived state, is extinguished without caffeine, and reverses in consumers who are fully withdrawn"
acquired liking requires Not mere exposure alone, but the association of flavor with a post-ingestive reward (nutritional or pharmacological)contested
Scope: The acquisition of liking for bitter foods and novel foods in humans
Note: For caffeinated beverages, reward association plus motivational-state dependence is supported, while for vegetables, liking rises from reward-free repeated exposure alone, and no additional benefit from flavor-nutrient conditioning is found (intake even shows a lower SMD of -0.45). The phenomenon (whether liking can be acquired) is separated out into each claim, and this claim addresses only the mechanism. The mechanism may differ by domain (presence or absence of a pharmacological reward).
Context: The authors suggest that the hedonic increase arose 'perhaps through a learned association of flavor with post-ingestive consequences', but the design does not separate mere exposure from association with nutritional/pharmacological consequences.
Limitations: It cannot be determined whether this is mere exposure or post-ingestive reinforcement. An abstract-level reading only.
In the domain of caffeinated beverages, the acquisition and expression of the increase in liking tracked the contingency between caffeine and the fasting/abstinence state, consistent with reward association rather than mere exposure.
Limitations: Domain-specific (pharmacologically active beverages); may not generalize to foods without a drug reward.
Contradicts: Simple repeated exposure without a contingent reward raised liking for vegetables (SMD 0.35), while flavor-nutrient conditioning produced lower intake than repeated exposure (SMD -0.45 [-0.87, -0.03], P = 0.04; 3 comparisons, 101 participants). In this domain, a post-ingestive nutritional reward added no benefit.
Limitations: The flavor-nutrient comparison relies on only 3 comparisons. Vegetables are not stratified by bitterness. A nutritional reward is pharmacologically different from caffeine.
Show dataconcept-acquired-liking —requires→ "Not mere exposure alone, but the association of flavor with a post-ingestive reward (nutritional or pharmacological)"
The biggest surprise: the more sensitive you are, the more coffee you drink
「苦味に敏感な人は苦い飲み物を避ける」 (in translation: "people sensitive to bitterness avoid bitter drinks") — this intuitive picture does hold for sensitivity to quinine and PROP. But for genetic sensitivity to caffeine's bitterness alone, the direction reverses, and more sensitive people drink more coffee (+0.146 cups/day per SD). The authors' interpretation is 「カフェインの効き目を学習した結果、苦味が良いものの合図になった」 (in translation: "as a result of learning caffeine's effects, bitterness became a signal of something good"). Through learning, bitterness can turn from an alarm into a forecast.
caffeine — perceived bitterness is positively associated with: Coffee intake: +0.146 [0.103, 0.189] cups/day per 1 SD of genetically predicted caffeine bitterness. OR 1.207 [1.126, 1.294] for becoming a consumer of more than 4 cups/day. Tea intake moves in the opposite direction (-0.045 [-0.061, -0.030])supported
Scope: Mendelian randomization, UK Biobank white British ancestry, n = 438,870. The instrumental variable was estimated in an adolescent twin sample, n = 1757
Verbatim confirmed: 'A standard deviation (SD) higher in genetically predicted bitterness of caffeine was associated with increased coffee intake (0.146 [95%CI: 0.103, 0.189] cups/day)'. The authors propose that 'coffee consumers acquire a taste for (or an ability to detect) caffeine given the learned positive reinforcement (i.e. stimulation) elicited by caffeine'.
Limitations: Limitations noted by the authors themselves: the instrumental-variable discovery sample is small (n = 1757), a single-SNP instrument prevents multi-SNP sensitivity analysis, self-reported beverage data, and no generalization to non-European populations.
Verbatim confirmed: "The caffeine-taste sensitive TAS2R14 alleles were nominally associated with higher liking for coffee and intake of caffeinated coffee, regardless its preparation type" — genetic variants for caffeine-bitterness sensitivity point in the same direction as higher coffee liking and intake, consistent with Ong 2018's +0.146 cups/day. The quinine-sensitivity variant (rs10772420 A) instead shows lower coffee intake (β=-0.02) and higher tea intake (β=0.03), also consistent with Ong 2018's 「紅茶は逆方向」 (in translation: "tea moves in the opposite direction"). The overview states "genetically inferred caffeine sensitivity was more strongly associated with coffee taste preferences than with genetically inferred bitter taste perception", also showing a hierarchy in which physiological caffeine sensitivity dominates over bitter perception.
Limitations: PMC8669025 full text confirmed. The TAS2R14 association is nominal (a level that does not pass multiple-testing correction), and in the US cohorts there is no significant association with coffee intake (only dark chocolate is nominal). Since the cohort is the same UK Biobank as Ong 2018 and the author (Cornelis) overlaps, this is not an independent replication but 「同方向の別解析」 (in translation: "a separate analysis pointing the same direction").
Show datacompound-caffeine —has_perceived_bitterness_positively_associated_with→ "Coffee intake: +0.146 [0.103, 0.189] cups/day per 1 SD of genetically predicted caffeine bitterness. OR 1.207 [1.126, 1.294] for becoming a consumer of more than 4 cups/day. Tea intake moves in the opposite direction (-0.045 [-0.061, -0.030])"
A small-scale experiment has also found a similar link. Bitterness ratings of a caffeine solution were higher among people with higher habitual caffeine intake, and were also related to the level of bitter-receptor gene expression on the tongue. Looking at quinine bitterness ratings in the same subjects, this relationship did not appear — a link specific to caffeine bitterness alone.
caffeine — perceived bitterness is associated with: Habitual caffeine intake and bitter-receptor mRNA abundance in taste papillaesupported
Scope: A small-scale human study reported in the cited paper
Note: The association does not establish the direction of causation, nor can it be generalized to every population.
n=20. gLMS bitterness ratings of 8 mM caffeine were associated with combined TAS2R10/43/46 mRNA abundance (r(18)=0.47) and with habitual intake (t(18)=2.78, p=.012), with the 2 factors together explaining 47% of the variance in ratings. TAS2R43 alone also showed r(18)=0.54, p=.014. No comparable association was seen for quinine.
Limitations: A cross-sectional study of n=20. Receptor genotypes other than TAS2R38 were not controlled for. The direction of the association (whether habitual drinkers feel it more or less strongly) is not stated explicitly in the text, and the discussion notes that prior intake-manipulation studies have reported inconsistent results.
Context: In large-scale genetic epidemiology too, people with higher genetic sensitivity to caffeine bitterness drink more coffee (+0.146 cups/day per SD) — consistent with the direction of 「常飲と感受性の関連」 (in translation: "the association between habitual drinking and sensitivity"), and interpreted as learned positive reinforcement.
Limitations: Intake is self-reported. The causal path from sensitivity to intake is not a direct measurement of a learning mechanism.
But this should not be overgeneralized. In a systematic review of adults, bitterness sensitivity itself, as measured by quinine or PROP, showed no coherent association with liking or intake of the major food groups — vegetables, fruit, grains, meat, dairy (within the scope of the abstract). Innate sensitivity strength alone explains very little of what adults choose to eat. What seems to separate whether someone comes to like bitterness is not sensitivity itself, but rather what gets learned further down the line.
bitterness has individual sensitivity minimally associated with Adult food-group liking and intake: pooled associations across vegetables/legumes, fruit, grains, meat, and dairy are small and non-significant. Only carbohydrate intake shows a modest inverse association (Cohen's dz -0.14, 95% CI -0.27 to -0.01, P =.038, I2 = 82%)supported
Scope: A systematic review and meta-analysis of 72 studies (36 of which were meta-analyzed) in adults without chronic disease. Bitterness sensitivity was measured by phenotype (PROP/PTC) or genotype (TAS2R38)
Note: An upper-bound finding across the literature that individual differences in bitterness sensitivity explain little of adult eating behavior. The authors point out that the influence of study design and dietary assessment method may be larger than the phenotype/genotype distinction.
For the major food groups (vegetables/legumes, fruit, grains, meat/substitutes, dairy/substitutes) and most nutrients, pooled associations were small and non-significant. A modest inverse correlation was observed only for carbohydrate intake (18 studies, 53 effects; Cohen's dz = -0.14, 95% CI: -0.27 to -0.01; P =.038; I2 = 82%), and this was sensitive to leave-one-out analysis.
Limitations: Abstract-level: effect sizes by food group and the study list are in the full text, which is not open access. The single significant pooled effect has high heterogeneity (I2 = 82%). The authors note that the effect was larger for some designs and dietary-assessment methods.
Show datasensation-bitter —has_individual_sensitivity_minimally_associated_with→ "Adult food-group liking and intake: pooled associations across vegetables/legumes, fruit, grains, meat, and dairy are small and non-significant. Only carbohydrate intake shows a modest inverse association (Cohen's dz -0.14, 95% CI -0.27 to -0.01, P =.038, I2 = 82%)"
A prior study cited by the review reports that habitual drinkers of canned coffee beverages rate bitterness and retronasal aroma more strongly than infrequent drinkers. The reading offered is that drinking it routinely increases familiarity with coffee, which in turn raises sensitivity to its sensory attributes.
acquired liking — perceived bitterness is associated with: The source reports that frequent consumers of canned coffee beverages rate bitterness and retronasal aroma more strongly than infrequent consumers, and argues this is because daily consumption in everyday life increases familiarity with coffee, which in turn raises sensitivity to sensory attributes.preliminary
Scope: Stratum: literature (a secondhand citation within a review).
Note: A secondhand citation. Content in which Hong 2025 (a review, full text already read) cites Gotow et al. 2015; the original (Gotow et al. 2015) has not been read.
"Gotow et al. (2015) reported that frequent consumers of canned coffee beverages rate bitterness and retronasal aroma more intensely than less frequent consumers. They suggested that frequent consumption in daily life enhances familiarity with coffee, which in turn increases sensitivity..."
Limitations: The original (Gotow et al. 2015) is unread, a secondhand citation, and details such as the number of frequent/infrequent consumers and statistics cannot be confirmed from the review's description.
Show dataconcept-acquired-liking —has_perceived_bitterness_associated_with→ "The source reports that frequent consumers of canned coffee beverages rate bitterness and retronasal aroma more strongly than infrequent consumers, and argues this is because daily consumption in everyday life increases familiarity with coffee, which in turn raises sensitivity to sensory attributes."
Another prior study cited by the same review records a change running the opposite direction. When Korean consumers, initially unfamiliar with Emmental, Gorgonzola, and Parmigiano Reggiano, evaluated these cheeses, negative terms such as bitterness, off-flavor, and poor balance were strongly selected. But after repeatedly tasting the same cheeses, the frequency of choosing these negative terms had fallen. Repeated experience moves not only how strongly bitterness is felt, but even how it is described.
acquired liking — perceived bitterness is associated with: In a CATA-method cheese survey of Korean consumers, negative sensory characteristics such as bitterness, off-flavors, and poor balance were strongly selected for Emmental, Gorgonzola, and Parmigiano Reggiano, which were initially unfamiliar, but the frequency of selecting these negative terms decreased after repeated consumption.preliminary
Scope: Stratum: literature (a secondhand citation within a review).
Note: A secondhand citation. Content in which Hong 2025 (a review, full text already read) cites Go et al. 2017; the original (Go et al. 2017) has not been read.
"Go et al. (2017) observed that in a CATA test of various natural cheeses, Korean consumers who were initially unfamiliar with cheeses such as Emmental, Gorgonzola, and Parmigiano Reggiano, perceived strong negative sensory characteristics such as bitterness, off-flavors, and a lack of balance. However, after repeated exposure, the frequency of selecting these negative terms decreased."
Limitations: The original (Go et al. 2017) is unread, a secondhand citation, and details such as the number of consumers and the number of repeated exposures cannot be confirmed from the review's description.
Show dataconcept-acquired-liking —has_perceived_bitterness_associated_with→ "In a CATA-method cheese survey of Korean consumers, negative sensory characteristics such as bitterness, off-flavors, and poor balance were strongly selected for Emmental, Gorgonzola, and Parmigiano Reggiano, which were initially unfamiliar, but the frequency of selecting these negative terms decreased after repeated consumption."
The effect of exposure is not permanent — at the longest follow-up point, the effect is no longer detected.
The large-scale genetic-epidemiology figure of 「+0.146杯/日」 (in translation: "+0.146 cups/day") comes from an observational study, not a direct measurement of the learning mechanism itself.
There are also cases where repeated exposure is not a cure-all — in a small pilot trial using an oral capsaicin gel for 14 days, no clear change appeared, either in the ability to distinguish salty, sweet, sour, and bitter tastes, or in their perceived intensity, before and after use.
The claims about habitual coffee drinking and repeated cheese consumption rest on prior studies cited by a review. The claim that the association between bitterness sensitivity and food preference is weak stays within the scope of an abstract.
repeated dietary exposure has limited long-term persistence of Improvement in liking: at the longest follow-up point, no effect remained on liking for novel vegetables (SMD 0.15 [-0.21, 0.50], P = 0.41)supported
Scope: A follow-up analysis within the same meta-analysis. Follow-up periods were often short and inconsistent
Note: The authors' overview: 『Effect sizes, however, are small, and limited evidence suggests long-term benefits』. A counter/limiting claim showing the limits of the durability of acquired liking. Durability may be age-window dependent: in 95 infants, the effect was stable up to 3 months later (Remy 2013, linked as contradicts).
At the longest follow-up point, no effect remained on liking for novel vegetables (SMD 0.15 [-0.21, 0.50], P = 0.41), and when measured at follow-up, effects had generally decreased.
Limitations: Few studies measured long-term outcomes, and the range of follow-up periods is wide (up to several years at most).
Contradicts: In 95 infants at the start of weaning (6.4 +/- 0.8 months), 10 repeated exposures to a basic artichoke puree raised intake (+63%) and liking (+21%), and verbatim, 'Learning of artichoke acceptance was stable up to 3 mo postexposure.' An adequately powered infant trial incorporating follow-up at 2 weeks, 3 months, and 6 months records persistence of the exposure effect over several months, conflicting with the meta-analysis finding that no effect remains at the longest follow-up point.
Limitations: The stability described explicitly in the abstract covers up to 3 months. Since the abstract does not report the result at 6 months, persistence beyond 3 months is untested here. Infants at the start of weaning are a different population from the older children who form the core of the meta-analysis, so this may reflect an age window rather than a genuine contradiction.
Show dataprocess-repeated-exposure —has_limited_long_term_persistence_of→ "Improvement in liking: at the longest follow-up point, no effect remained on liking for novel vegetables (SMD 0.15 [-0.21, 0.50], P = 0.41)"
repeated dietary exposure does not significantly influence A test conducted after 14 days to distinguish saltiness, sweetness, sourness, and bitterness, and an intensity evaluationpreliminary
Scope: A small pilot study of 10 healthy women using an oral gel containing capsaicin
Note: This is a result in which no difference was found in a trial with few participants, and is not proof that there is no effect at all.
Before and after using the capsaicin-containing gel in the mouth for 14 days, there was no statistically clear change in the performance of distinguishing saltiness, sweetness, sourness, and bitterness, or in the perceived intensity.
Limitations: This is a small pilot study, and there are also limitations in how the comparison group was set up. It is not proof that there is no effect at all.
Show dataprocess-repeated-exposure —does_not_significantly_influence→ "A test conducted after 14 days to distinguish saltiness, sweetness, sourness, and bitterness, and an intensity evaluation"
Does eating bitter food repeatedly make you come to like it?
An analysis pooling multiple trials in infants and young children confirms that repeated tasting raises both liking for vegetables and intake, if only slightly. Even in adults, there is a record of an experiment in which continuing to drink a commercially available bittersweet beverage once a day at home for 7 days raised ratings of that beverage. The effect is small, however, and in some trials it was no longer detected at the longest follow-up point. What food, how many times, and over how long a period it needs to be tasted to work differs by food, and has not yet been confirmed in a form that can be generalized.
Is coming to like coffee's bitterness thanks to caffeine?
In experiments where a new flavored beverage is repeatedly paired with caffeine, liking for that flavor has been confirmed to rise. But this liking is a conditional result that appears only in a state where caffeine is still being taken. A large-scale genetic-statistics analysis has also found a link running in the direction where people more sensitive to caffeine's bitterness drink more coffee. Sensitivity to other bitter substances, such as quinine, shows no such reversal. Whether bitterness itself came to be liked, or whether it was learned as a signal for reward, has not yet been confirmed.
Around what age does resistance to bitter vegetables disappear?
It has been confirmed that newborns show a stereotyped rejection response to bitter solutions. As for when and how the response to bitterness changes with age, the reviews summarize a shift in sweetness preference, but no primary-study figures giving a similar timeline for bitterness itself are yet in this database. Repeated tasting experience is known to raise liking, but at what age it removes the rejection response itself has not been confirmed.
What is not known
No primary study timing the weakening of the bitterness rejection response by bitterness itself, rather than by proxy, has yet been found in this database.
How many years a liking acquired through repeated exposure lasts is not yet well measured, since follow-up periods in many trials are short.
Whether the same reversal seen with coffee — 「感受性が高いほど好む」 (in translation: "the more sensitive, the more it is liked") — also occurs for bitterness not mediated by caffeine, such as in vegetables or medicinal plants, has not yet been confirmed.
The reports that habitual drinkers of canned coffee and habitual eaters of cheese change how they perceive and describe bitterness are secondhand citations from a review rather than the original studies, and the sample sizes and statistics in the originals cannot be confirmed.