Learning to like vegetables during breastfeeding: a randomized clinical trial of lactating mothers and infants

Abstract

Background: What lactating mothers eat flavors breast milk and, in turn, modifies their infants’ acceptance of similarly flavored foods.

Objective: We sought to determine the effects of the timing and duration of eating a variety of vegetables during breastfeeding on the liking of vegetables in both members of the dyad.

Design: We conducted a randomized controlled study of 97 mother-infant dyads. Lactating mothers drank vegetable, beet, celery, and carrot juices for 1 mo beginning at 0.5, 1.5, or 2.5 mo postpartum or for 3 mo beginning at 0.5 mo postpartum. The control group drank equal volumes of water and avoided drinking the juices. Mothers rated the tastes of the juices and self-reported dietary intakes at each monthly visit (0.5–4.5 mo). After weaning, when 7.9 mo of age, infants’ acceptance of plain, carrot-flavor (exposed flavor), and broccoli-flavor (nonexposed flavor) cereals was assessed on separate days.

Results: The timing of exposure affected the acceptance of the carrot flavor that did not generalize to the novel broccoli flavor. A relatively brief experience (1 mo) with vegetable flavors in mothers’ milk, starting at 0.5 mo postpartum, was sufficient to shift the hedonic tone, which resulted in a faster rate of eating carrot-flavored cereal than that in infants who were exposed during subsequent months or not at all. One month of exposure had a greater effect than 3 mo of exposure or no exposure. Regardless of when exposure occurred, infants were less likely to display facial expressions of distaste initially when eating the carrot cereal. Over time, mothers liked the tastes of carrot, beet, and celery juices more, but no changes in dietary intake of vegetables were observed.

Conclusions: Early life may be an optimum time for both infants and their mothers to learn to like the taste of healthy foods. More research is needed to facilitate the liking and eating of these foods by mothers, which will, in turn, increase the likelihood of their feeding these foods to their children. This trial was registered at clinicaltrials.gov as NCT01667549.

INTRODUCTION

Diets that are insufficient in vegetables are one of the top dietary risk factors that contribute to both the global burden of disease and mortality (1, 2). Unfortunately, the current reality is that 1 in 4 American toddlers do not consume even one vegetable on a given day (3–5) and, instead, like their mothers and other adults (6), are more likely to consume sweet- and salty-tasting snacks and sweetened beverages (3, 4), all of which have desirable sensory properties that are inborn (7). By 4 y of age, most children have diets that include only a small portion of vegetables (∼5% of calories), whereas >15% of calories come from sweets (8–10). Not surprisingly, Healthy People 2020 includes as one of its goals an increase in the variety and contribution of vegetables in the diets of all Americans ≥2 y of age (11). Similar recommendations will most likely be made for even younger Americans (12).

How do we help children get off to a healthy start by making vegetables a part of their diet? Infants’ chemosensory systems are functioning early in life and are open to learning (13). Research has shown that infants learn about the flavors of foods, including vegetables, before their first taste of solid foods, via the transmission of a variety of dietary volatiles from the mothers’ diet to the infants’ first foods [i.e., amniotic fluid and breast milk (reviewed by Mennella (14)]. Breastfeeding confers a greater acceptance of these foods but only if they are part of the mother’s diet (15–17) particularly during the first 4 mo of life (16). The more varied the mother’s diet is during lactation, the more varied the flavor of her milk will be (14), which could explain why children who were breastfed are less picky and more willing to try new foods as they grow (18).

One challenge in investigating how children learn to like healthy foods such as vegetables is to determine whether there are optimum times when an experience promotes greater liking (19). A previous trial showed a greater liking of carrot-flavored cereal by infants whose lactating mothers were randomly assigned, during the first 3.5 postpartum months, to drink carrot juice, which flavored their breast milk (20), compared with avoiding carrot juice and drinking water (16). Therefore, we aimed to determine whether the timing and duration of the exposure to vegetable flavors in breast milk were important in shifting the hedonic tone of the taste of vegetables. Because babies breastfeed more often during the first months of life (21), we hypothesized that learning may be more effective earlier rather than later. Because research in weaned infants has revealed that repeated exposure to a variety of vegetable flavors increases the infants’ acceptance of the exposed as well as novel vegetables (22–25), we provided mothers with different types of vegetable juices to drink during the exposure period.

A second, related challenge in investigating how children learn to like healthy foods is that the overall diet quality, particularly vegetable intake, in women remains below current recommendations (9, 26, 27). Because mere exposure to the taste of a bitter-sweet beverage can shift the liking of its taste in adults (28), we explored whether manipulating the diets of lactating mothers and repeatedly having them taste a variety of vegetable juices could also modify the liking of the taste of vegetables for both members of the dyad.

METHODS

Participants and random assignment

Pregnant women or newly parturient women who intended to exclusively breastfeed their infants >4 mo were recruited from advertisement in local newspapers, websites, and Philadelphia Women, Infants, and Children offices. Inclusion criteria included a healthy, term, singleton birth, an establishment of lactation with the intention to exclusively breastfeed for >4 mo, and no reported allergies to fruit or vegetables. When infants were ∼2 wk of age, the dyads came to the Human Sensory Testing Facility at the Monell Chemical Senses Center in Philadelphia, Pennsylvania, whereupon they were randomly assigned into 1 of 5 groups who differed in the timing and duration of when mothers drank vegetable juices during their infants’ first 4 mo postpartum. The randomization scheme was generated by a statistician before any subject enrollment, and the random assignment was stratified to ensure a sex balance in the groups.

On the basis of our previous study that focused on the timing and duration of a sensitive period during formula feeding whereby the intake response within each group was normally distributed with an SD of 96.0 (29), we determined that a target sample size of 10–12 subjects/group would provide 95% power to detect a change at the 5% level. Study staff and mothers were not aware of the hypotheses and were blinded to the group assignment or order of testing. All procedures were followed in accordance with the ethical standards of and approved by the Office of Regulatory Affairs at the University of Pennsylvania. Informed consent was obtained from each mother before inclusion. This trial was registered at clinicaltrials.gov as NCT01667549.

Trial design

As shown in Figure 1, when infants were 0.5 mo of age (±1 wk), dyads were randomly assigned into the following groups who differed in the timing and duration of when lactating mothers drank vegetable juices during their infants’ first 4 mo postpartum as follows: exposure from 0.5 to 1.5 mo (1M0.5), exposure from 1.5 to 2.5 mo (1M1.5), exposure from 2.5 to 3.5 mo (1M2.5), exposure from 0.5 to 3.5 mo, or no exposure (control group). Mothers in all groups were given 118-mL cups from which to drink the juice (exposure group) or water (control group). Women in the experimental groups were provided with monthly supplies of the juices gratis at the beginning of the exposure month (i.e., 0.5, 1.5, or 2.5 mo) or months (0.5–2.5 mo). The women were instructed to drink a cupful (118 mL) of the vegetable juices per occasion during their particular month or months of exposure (resulting in 24 total exposures/mo; 8 carrot juice exposures; and 8 vegetable juice exposures, 4 beet juice exposures, and 4 celery juice exposures). The women were given the following instructions. First, they could choose in which order they drank the 4 juices, but once they began drinking one type of juice, they were to finish that bottle before moving on to the next; the number of exposures was based on the size of the juice bottles provided. Second, they were to drink the juice ∼1–2 h before breastfeeding to maximize the likelihood that the infants were exposed to the flavors in breast milk. Previous research established that a variety of volatiles, including those found in carrot juice, transmit to human milk 1–3 h postingestion (14, 20). Third, during the nonexposed months for the experimental groups or during each month for the control group, mothers were instructed to drink an equal volume of water and to refrain from ingesting the exposure juices (or their solid forms). Finally, the mothers recorded what time that they drank the juices (or water) each day and when and how often they breastfed their infants afterward, and telephone contact was made weekly during the first 4 mo of the study to encourage, and later to verify, compliance with study protocols and breastfeeding exclusivity. Every month thereafter, telephone contact was made monthly to determine when mothers began to complement breastfeeding with solid foods.

FIGURE 1.

Trial design. Lactating mothers were randomly assigned to consume a variety of juices [i.e., carrot, celery, beet, and vegetable juices (colored bars)] either from 0.5 to 1.5 postpartum (1M0.5 group), from 1.5 to 2.5 mo postpartum (1M1.5 group), from 2.5 to 3.5 mo postpartum (1M2.5 group), from 0.5 to 3.5 postpartum (3M0.5), or not at all (control group). During the nonexposed months for the experimental groups or each month for the control group, mothers drank an equal volume of water (gray boxes). When infants were introduced to complementary foods (6–8 mo of age), they were videotaped as they consumed, in counterbalanced order and on separate days, cereal prepared with water, carrot juice (exposed flavor), or a broccoli juice (novel flavor). 1M0.5, exposure from 0.5 to 1.5 mo; 1M1.5, exposure from 1.5 to 2.5 mo; 1M2.5, exposure from 2.5 to 3.5 mo; 3M0.5, exposure from 0.5 to 3.5 mo.

At the first visit (0.5 mo), each month thereafter for the next 4 mo (at 1.5, 2.5, 3.5, and 4.5 mo), and for the vegetable acceptance test visit (at 6–8 mo), dyads came to the Monell Center at the same time of day that the mothers drank the juices during the exposure period. At each visit, mothers and infants were weighed and measured, and weight-for-height z scores for the infants were calculated with the use of Anthro software (version 3.2.2; WHO), and infants were observed breastfeeding from their mothers. At the start of the study, mothers were asked about their motivations to promote vegetable consumption (30) to establish levels of motivations in the groups. No instructions were given regarding when to wean the infants or what foods to feed them.

Food stimuli

For each month of maternal exposure, mothers consumed 118 mL of one type of the following juices on each of 24 d/mo as follows: carrot juice (946 mL bottle; Lakewood Juices), beet juice (500 mL bottle; Biotta Products), celery root juice (500 mL bottle, Biotta Products), and vegetable juice (eight 163-mL cans; V8 Campbell Soup Co.). These same juices, along with apple juice (Ocean Spray), were used during the psychophysical testing of the mothers. The food stimuli that were used for the vegetable acceptance testing at weaning were plain rice cereal (84 g; Gerber Products Co.) that was made with 16 g H₂O or equal volumes of carrot juice (exposed flavor) or broccoli juice (novel flavor).

Infants: vegetable acceptance tests

Several weeks after the mothers began complementing their infants’ breast-milk diet with solid foods, dyads returned to the Monell Center for 3 d and were videotaped as the mothers fed their infants, in a counterbalanced and randomized order, plain, carrot-flavored, or broccoli-flavored cereal. Testing, which occurred in a private 108-sq ft testing room, was conducted under naturalistic conditions in which infants determined the pacing and duration of feeding (16). During feeding, mothers refrained from talking or making faces to eliminate any potential influence on their infants’ behaviors (31). To minimize possible effects that were due to different levels of satiation, the testing sessions, which lasted 2–3 h each, took place at the same time of day that the mothers drank the juice or water during the exposure periods and that the infants were usually fed (0010 to 1400). The infants, who sat in a high chair (Flair Pedestal Highchair; Boon), were fed until they refused the food on 5 consecutive occasions with the criterion that the infant exhibited such behaviors as turning his or her head away, pushing the spoon away, crying, or becoming playful; the experimenter monitored the behaviors and signaled to the mother when to stop feeding (15, 16). Immediately after each feeding session, the mothers, who were unaware of the hypotheses, rated, on a scale of 1 (extreme dislike) to 9 (extreme like), their infants’ enjoyment of the food.

The infants’ behaviors were monitored via videotape. The number of minutes that elapsed from the start to end of feeding was the duration of the feed, and the amount consumed was determined by weighing the bowl immediately before and after each feed on a Mettler balance (model SG1600; Mettler-Toledo). All food that was spilled onto the tray or bib during the feed was placed in the bowl before weighing. Mothers were never told how much their infants had ingested and were blind to the hypotheses. Each videotape was subjected to a frame-by-frame analysis to determine whether infants made a facial expression of distaste in response to the first 5 spoons offered. The raters focused on brow movements (i.e., brow lowering and inner brow raises), nose wrinkling, upper-lip raising, and squinting because these facial expressions have been identified as prototypical of distaste and are discriminating in gauging infants’ like and dislike of tastes and flavors (15, 32). Intercoder reliability was established by the common scoring of 198 feeds [r(196 df) = 0.99, P < 0.0001], which indicated good reliability.

Mothers: psychophysical testing and diet records

At baseline (0.5 mo) and at the end of each of the first 4 monthly visits, mothers rated their liking of the taste of the juices used during the exposure (e.g., carrot, beet, celery root, and vegetable juice) as well as of a sweet-tasting apple juice. After tasting each juice, which were presented in a randomized order, for 5 s via a swish-and-spit protocol (without swallowing), participants rated its taste with the use of the general labeled magnitude scale (gLMS) to measure taste intensities (e.g., bitterness and sweetness; scores range from not detectable to strongest imaginable) and the hedonic gLMS to measure ab affective experience in terms of the liking (positive) and disliking (negative) on a computer with Compusense 5 Plus software (Compusense). Both measures are validated psychophysical tools for taste research, and participants are trained in their use before testing. In brief, participants score experiences with the use of a vertical axis that is lined with adjectives that are spaced semi-logarithmically on the basis of experimentally determined intervals to produce ratio-quality data that can yield valid group comparisons that cannot be made via other types of rating (e.g., Likert) scales because the labels on the latter are not assumed to be equal (33, 34). For the hedonic gLMS scale, scores range from −95 to 95 with the anchors falling approximately as follows: strongest imaginable like = 95, very strong like = 50, strong like = 33, moderate like = 16, weak like = 6, barely detectable like = 1.4, neutral = 0, barely detectable dislike = −1.4, weak dislike = −6, moderate dislike= −16, strong dislike = −33, very strong dislike = −50, and strongest imaginable dislike = −95. One minute separated the presentation of each stimulus during which subjects rinsed their mouths with water.

Self-reported dietary intake data were recorded with the use of the Automated Self-Administered 24-Hour Recall System that was developed by the National Cancer Institute (35). During each study visit, mothers reported a 24-h dietary recall to a trained researcher. After a subject reported a specific food or beverage, the Automated Self-Administered 24-Hour Recall System provided a visual depiction of the item in an appropriate dish, which allowed subjects to accurately estimate portion sizes. Response sets with implausible energy intake on the basis of estimated energy needs were eliminated before analyses (36). From the data collected, we focused on daily caloric intake (kilocalories per day) and cup equivalents of daily fruit and vegetable intakes, and we determined whether the mother was meeting daily recommendations for fruit (2 cups/d) and vegetables (2.5 cups/d) (37).

Statistics

Primary outcome measures for the infants were total intake (grams), the rate of feeding (grams per minute), and maternal perceptions of the infants’ enjoyment after eating each of the different flavored test foods (plain cereal, carrot-flavored cereal, and broccoli-flavored cereal). To determine whether there were main effects, we conducted a repeated-measures ANOVA with test foods as the within-subject factor and the experimental group as the grouping factor. Planned comparisons were carried out to test our timing and duration hypotheses. First, we conducted separate ANOVAs with a priori contrasts that were specified to determine whether the 1-mo exposure groups (1M0.5, 1M1.5, and 1M2.5) differed from the no-exposure control group to test the hypothesis that the timing of exposure would affect acceptance (timing hypothesis). Second, we conducted separate ANOVAs with a priori contrasts that were specified to determine whether 1 mo of exposure differed from 3 mo of exposure (1M0.5 group and the group with exposure from 0.5 to 3.5 mo) and whether these groups differed from the no-exposure control group (duration hypothesis). Because 3 comparisons were made for each outcome measure, a Bonferroni adjustment was made such that only a planned comparison with P < 0.02 was considered significant.

Primary outcome measures for the mothers were gLMS ratings of the juices and dietary intake over time, which were analyzed with the use of repeated-measures ANOVAs with the type of juice as the within-subject factor and the experimental group as the grouping factor. Pearson correlations were conducted to determine associations between continuous variables, and Fisher’s exact probability tests were used to determine the effect of the group on nonparametric data (e.g., demographics and parenting attitudes toward vegetables). Analyses were conducted with the use of Statistica software (version 12; StatSoft). All summary statistics are expressed as means or least-squares means ± SEMs.

RESULTS

Study participants

Of the 97 dyads who were randomly assigned, 77% of the dyads completed the study. As shown in Supplemental Figure 1, the remaining women dropped out of the study because they no longer wanted to participate (n = 8), did not comply with the drinking of the juices (n = 3), moved out of the area (n = 3), or were lost to follow-up (n = 8). There were no differences in the groups regarding the reasons why they did not complete the study.

Table 1 summarizes the characteristics of 75 dyads of mothers and infants who completed the study. All but 3 women exclusively breastfed for at least the first 3.5 mo of life (the remaining mothers exclusively breastfed for 2.5 mo and predominantly breastfed the subsequent month). The race/ethnicity of the dyads and the mother’s highest educational level reflected the racial and socioeconomic diversity of the Philadelphia area (38). There were no significant group differences in the weight-for-length z score, race-ethnicity, sex ratio of the infants, or age of the infants at the time of the vegetable acceptance tests, and there were no group differences in maternal parity, education, income levels, or quality of the maternal diet related to vegetable and fruit intake at study entry.

TABLE 1.

Subject characteristics¹

	Group
Characteristic	1M0.5 (n = 17)	1M1.5 (n = 15)	1M2.5 (n = 14)	3M0.5 (n = 14)	Control (n = 15)	P
Infants
Sex, F, %	59	47	43	57	47	0.87
Race, %	—	—	—	—	—	0.68
African American	18	40	29	21	27
Caucasian	59	47	43	64	60
Asian	0	7	0	0	7
More than one	24	7	29	14	7
Vegetable acceptance test
Age, mo	7.8 ± 0.22	7.8 ± 0.2	7.5 ± 0.2	8.1 ± 0.2	8.1 ± 0.2	0.29
Weight-for-length z score	0.55 ± 0.25	0.56 ± 0.27	0.05 ± 0.28	0.77 ± 0.28	0.58 ± 0.27	0.45
Mothers
Age, y	33.0 ± 1.3	30.7 ± 1.3	29.7 ± 1.4	31.5 ± 1.4	31.9 ± 1.3	0.47
Parity, multiparous, %	47	80	79	64	80	0.18
Education, college, %	76	67	50	71	67	0.52
Household income, $/y, %	—	—	—	—	—	0.94
<10,000	6	13	7	14	13
10,000–49,999	29	40	43	43	27
>50,000	65	47	50	43	60
Dietary intake at study entry (0.5 mo)
Vegetables, cups/d	1.6 ± 0.3	1.6 ± 0.3	1.5 ± 0.3	1.3 ± 0.3	1.5 ± 0.3	0.98
Fruit, cups/d	1.9 ± 0.4	1.9 ± 0.4	1.4 ± 0.4	2.0 ± 0.4	1.0 ± 0.4	0.29
Total, kcal	1967.9 ± 145.4	1815.7 ± 154.7	2209.2 ± 160.2	2044.3 ± 160.2	1876.3 ± 154.7	0.44

¹P values are for the main effect of the group and were obtained from Pearson chi-square analyses or 1-factor ANOVAs with the group as the between-subject factor. 1M0.5, exposure from 0.5 to 1.5 mo; 1M1.5, exposure from 1.5 to 2.5 mo; 1M2.5, exposure from 2.5 to 3.5 mo; 3M0.5, exposure from 0.5 to 3.5 mo.

²Mean ± SEM (all such values).

During the month of exposure for the experimental groups, groups significantly differed in the number of times that mothers breastfed their infants per day. For example, mothers in the 1M0.5 and 1M1.5 groups breastfed more often during the month that they drank the juices (9.6 and 9.3 times/d, respectively) than did those in the 1M2.5 group (7.6 times/d; F_[2,43] = 5.65, P = 0.007). Most infants (83%) were breastfed for >10 mo and, thus, were breastfeeding at the time of the vegetable-flavor acceptance tests; only 8% of the infants were weaned from the breast between the ages of 3.5 and 7 mo. The groups did not differ in terms of when mothers started complementing breastfeeding with solid foods (P = 0.93); all infants were introduced to solid foods before the age of 8 mo with most infants (67%) being introduced to solid foods between 4 and 8 mo. Most lactating mothers (69%) prepared their infants’ cereal with liquids other than water (e.g., fruit juices, mother’s milk, and commercial baby foods). When mothers introduced solid foods, they fed their infants pureed table foods along with commercial baby foods; 20% of the mothers never fed their infants commercial baby foods.

Infants: vegetable acceptance at weaning

The groups did not differ in terms of the timing of testing relative to the number of hours since infants last fed (P = 0.22); the infants were fed 2.6 ± 0.1 h before the test sessions. Also, there was not an effect of the order of presentation on intake of the cereals (P = 0.81). Six infants refused to eat any food on all 3 testing days; 1 infant refused to eat on 2 test days, and 2 infants refused to eat on 1 test day (Supplemental Figure 1, Table 2).

TABLE 2.

Acceptance of flavored cereals by group¹

	Group					P-planned comparisons
Cereal type	1M0.5 (n = 17)	1M1.5 (n = 15)	1M2.5 (n = 14)	3M0.5 (n = 14)	Control (n = 15)	Timing hypothesis2	Duration hypothesis3
Plain cereal
Infants refused to eat, n	1	1	4	1	0	–—	–—
Infants completed test, n	16	14	10	13	15	–—	–—
Intake,4 g	56.8 ± 13.25	42.4 ± 12.1	29.0 ± 8.5	40.7 ± 8.3	36.5 ± 6.2	0.34	0.31
Rate of feeding,4 g/min	6.1 ± 0.9	3.9 ± 1.1	4.0 ± 0.8	4.3 ± 0.4	4.9 ± 0.6	0.27	0.21
Maternal enjoyment rating (1–9)4	5.4 ± 0.6	4.6 ± 0.8	4.1 ± 0.9	5.2 ± 0.6	3.9 ± 0.5	0.39	0.12
Broccoli cereal
Infants refused to eat, n	1	1	3	1	0	–—	–—
Infants completed test, n	16	14	11	13	15	–—	–—
Intake,4 g	41.8 ± 12.7	22.5 ± 8.8	30.9 ± 10.1	34.4 ± 12.4	30.4 ± 6.5	0.56	0.74
Rate of feeding,4 g/min	4.8 ± 0.8	2.4 ± 0.9	4.5 ± 1.0	3.8 ± 0.9	4.1 ± 0.8	0.19	0.70
Maternal enjoyment rating (1–9)4	4.2 ± 0.7	2.9 ± 0.6	4.7 ± 0.7	4.3 ± 0.6	4.1 ± 0.7	0.32	0.98
Carrot cereal
Infants refused to eat, n	2	2	4	1	0	–—	–—
Infants completed test, n	15	13	10	13	15	–—	–—
Intake,4 g	78.2 ± 14.8	36.6 ± 10.0	57.7 ± 19.9	52.2 ± 9.8	41.0 ± 7.6	0.03	0.02
Rate of feeding,4 g/min	7.2 ± 1.1	3.6 ± 0.8	4.8 ± 1.3	3.9 ± 0.5	3.9 ± 0.6	0.01	0.005
Maternal enjoyment rating (1–9)4	6.0 ± 0.7	4.7 ± 0.7	7.1 ± 0.4	6.1 ± 0.5	6.1 ± 0.7	0.16	0.99

¹Control group had no exposure. 1M0.5, exposure from 0.5 to 1.5 mo; 1M1.5, exposure from 1.5 to 2.5 mo; 1M2.5, exposure from 2.5 to 3.5 mo; 3M0.5, exposure from 0.5 to 3.5 mo.

²Values were obtained from ANOVAs with a priori contrasts to determine whether the timing of the 1-mo exposure groups (1M0.5, 1M1.5, and 1M2.5) differed from that of the control group.

³Values were obtained from ANOVAs with a priori contrasts specified to determine whether the duration of the exposure of the 1-mo exposure groups (1M0.5 and 3M0.5) differed from that of the control group.

⁴Data are from infants who completed tests.

⁵Mean ± SEM (all such values).

The repeated-measures ANOVA revealed a main effect of the flavor of the food that the infants were eating. As a group, the infants preferred some flavors more than others as was evidenced by how much they ate (F_[2,122] = 6.96, P = 0.001) and their mothers’ perception of their enjoyment of the food (F_[2,120] = 14.47, P < 0.0000). Infants ate significantly more of the carrot-flavored cereal (53.1 ± 5.6 g) than either of the broccoli-flavored cereal (32.8 ± 4.9 g; P = 0.0003) or plain cereal (42.0 ± 4.8 g; P = 0.05). Likewise, mothers gave significantly higher scores, although in the neutral range, when rating how much they thought that their infants enjoyed eating the cereal that was flavored with carrot juice (6.1 ± 0.3) than either the broccoli-flavored cereal (4.1 ± 0.3; P = 0.0000) or plain cereal (4.6 ± 0.3; P = 0.0003).

Table 2 shows the results of the planned comparisons that followed from the timing and duration hypothesis. As shown in Figure 2, the timing of when mothers drank the vegetable juices affected their infants’ acceptance of the carrot-flavored cereal but not of the plain or broccoli-flavored cereals (Table 2). Infants who had 1 mo of exposure to the vegetable flavors beginning when they were 2 wk of age were more accepting of the carrot-flavored cereal as determined by how fast they ate the food (P = 0.008) and how much they ate (P = 0.03). There were no differences between the groups whose exposure began either at 1.5 or 2.5 mo and the control group. Timing was more important than duration. The infants’ intake and rate of feeding the carrot-flavored cereal was greater in those who had 1 mo of exposure, but not 3 mo of exposure, than in the group with no exposure (P < 0.02). One month of exposure was no different from 3 mo of exposure. Regardless of the timing and how long mothers drank the juices, infants in the 4 experimental groups tended to be less likely to display a facial expression of distaste to their first spoonful of the carrot-flavored cereal (29%) than were the control infants (53%; P = 0.08).

FIGURE 2.

Least-square means ± SEMs of infants’ acceptance of carrot-flavored cereal. (A) Effect of the timing of maternal drinking of vegetable juices on infants’ intake. Infants whose mothers drank a variety of juices (carrot, celery, vegetable, or beet) from 0.5 to 1.5 mo postpartum (1M0.5 group) ate more carrot-flavored cereal and ate it at a faster rate than did those whose mothers did not drink the juices and avoided vegetables (control group). There were no differences between the groups who drank the juices either from 1.5 to 2.5 mo (1M1.5 group) or from 2.5 to 3.5 mo (1M2.5 group) and the control group. (B) Effect of the duration of maternal drinking of the vegetable juices on infants’ intake. Infants whose mothers drank the juices for 1 mo (1M0.5) ate more carrot-flavored cereal than did infants in the control group but were no different from those with 3 mo of exposure. *Significant planned comparison compared with the control group. 1M0.5, exposure from 0.5 to 1.5 mo; 1M1.5, exposure from 1.5 to 2.5 mo; 1M2.5, exposure from 2.5 to 3.5 mo; 3M0.5, exposure from 0.5 to 3.5 mo.

Mothers: psychophysical ratings of vegetable flavors, dietary intake, and attitudes about vegetables

Of the 75 women, 2 women did not participate in the gLMS testing on 1 of the test days, and therefore, they were excluded (Supplemental Figure 1). There was an interaction between the type of juice and the time postpartum on how much mothers liked or disliked the tastes of the juices (P = 0.02). At the start of the study, mothers liked the taste of apple juice the best and the taste of celery the least; vegetable juice was liked more than carrot juice, which was similar in liking to beet juice (P = 0.0001). Regardless of the experimental group, maternal hedonic ratings for carrot (P = 0.002), beet (P = 0.001), and celery (P = 0.006) juices increased over time (Figure 3). For carrot juice, the change in liking first occurred on the second visit; ratings increased from barely detectable dislike at baseline to weak like at 4.5 mo. For celery and beets, changes did not occur until the third visit; hedonic ratings for celery went from weak-moderate dislike to barely detectable-weak dislike and, for beets, went from barely detectable like to weak-moderate like. Paralleling the increases in liking were decreases in how bitter the mothers rated each of these juices (carrots: P = 0.02; beets: P < 0.0001; celery: P < 0.0001); there were no such time-related differences in perceived sweetness (all P > 0.14). Hedonic ratings for the more-preferred juices (i.e., apple and vegetable) did not differ between the groups or over time.

FIGURE 3.

Least-square means ± SEMs of mothers’ hedonic gLMS ratings of the taste of vegetable juices and apple juice over time. There were significant main effects for the type of juice (P < 0.00001) and time postpartum (P = 0.0005); data are collapsed across all groups. *Taste rating was significantly different from the 0.5-mo value for that juice. BD, barely detectable; CAR, carrot; CEL, celery; gLMS, general labeled magnitude scale; VEG, vegetable.

Total daily caloric intake reported by the mothers did not change during the exposure period (P = 0.67) and did not differ by group (P = 0.38). Fruit intake significantly decreased over time in all groups (P = 0.01) from 1.6 ± 0.2 cups/d at 0.5 mo postpartum to 1.1 ± 0.1 cups/d at 4.5 mo. Vegetable intake (P = 0.91) did not change. Like that observed at study entry, most mothers did not meet the daily recommendations for either fruit intake (81% of mothers ate <2 cups/d) or vegetable intake (87% of mothers ate <2.5 cups/d) at 4.5 mo postpartum.

Regardless of the group assignment, all but one (99%) of the mothers reported at the start of the study that it was either very important or important to them that their children’s future diet include vegetables. Mothers also reported that it was very important or important to them that their children’s diets also include fruit (99%) and, to a lesser extent, meats (76%).

DISCUSSION

The current study builds on the evidence base that mammalian infants learn about the food environment in which they live and their mothers’ diet during breastfeeding (13, 14, 39) and provides further evidence that there is a period during early life when the hedonic response to food flavors experienced in mother’s milk can be enhanced. Infants whose mothers drank a variety of vegetable juices, including carrot juice, were more accepting of cereal that was flavored with carrot juice (as determined by intake and rate of feeding) than were infants in the control group whose mothers drank water and avoided intake of the same vegetables. One month of exposure was more effective when it began at 0.5 mo postpartum than when it began at either 1.5 or 2.5 mo postpartum, which indicated a timing effect. In infants whose lactating mothers began drinking the juices shortly after birth (2 wk postpartum), those whose mothers drank the juices for 1 mo were more accepting of carrot-flavored cereal than were infants whose mothers drank the juices for 3 mo, thereby indicating that the duration was less important than the timing. But regardless of when and how long the flavor exposure were, infants in each of the exposure groups displayed fewer faces of distaste during the early parts of feeding than control infants did, which indicated that there was a shift in the hedonic tone of the carrot-flavored cereal.

Why was there an effect of the timing of exposure and not of the duration on how much babies ate and how fast they ate the carrot cereal? One hypothesis is that the timing was confounded by compliance with the protocol. Mothers may have been more motivated to drink the juices during the early postpartum period perhaps because of the rewarding properties of lactating on the mothers (40) or because drinking the juices for 3 mo was too burdensome. However, there were no differences in the groups in compliance or willingness to participate as was assessed by dropout rates or the evaluation of the logs. Neither were there group differences in maternal motivations because the vast majority of mothers reported that it was important that their children’s future diets include vegetables. Another hypothesis is that the timing of exposure is related to the amount and frequency that infants experience the flavors. Infants breastfeed more often during the first few months of life, and thus, they may have more opportunity to experience the flavors of vegetables in their mothers’ milk. However, mothers who drank the vegetable juices from 1.5 to 2.5 mo postpartum nursed as often as those who drank the juices from 0.5 to 1.5 mo postpartum, and thus, changes in the composition of milk (41) and the quantity of milk ingested by the infants (42) may have contributed to a varying potency of flavor experiences during the first 2.5 mo of life.

The timing of the plasticity of the neural substrates that underlie the observed developmental transition in human flavor learning remains unknown (43). Knowledge about the sensitive periods for learning about flavors that are not associated with the maternal diet (19, 29) and for other senses, particularly vision (44), may shed light on aspects of flavor learning that have not been explored. Most likely, there are multiple sensitive periods especially because learning about flavors and foods does not end with breastfeeding but continues throughout complementary feedings (13). Changes in intake after repeated exposure to fruit and vegetables during infancy contrast with slower changes that have been shown in toddlers (45), thereby suggesting that plasticity is more pronounced early (46). Such functional plasticity, which one of the main characteristics of the brain, highlights the ability to change behavior on the basis of previous experience (47).

Although the timing of exposure had significant effects on infants’ acceptance of the carrot-flavor foods, as a group, only 8% (6 of 75) of the breastfed infants rejected all of the foods. Although there was personal variation in feeding behaviors at this age, most infants ate sufficient amounts, preferring the flavor of carrot to broccoli and preferring the carrot-flavored cereal to the unflavored cereal. In addition, infants in the experimental groups liked the taste of the carrot-flavored cereal more than the control infants did as was evidenced by the facial expressions that were made during the early part of the feeding. Such hedonic shifts that are based on an experience with flavor in milk may underlie the finding that breastfed infants are more accepting of flavored foods after repeated exposure to a vegetable than are formula-fed infants (17). Mothers often used liquids other than water to prepare the first complementary foods and fed their infants commercial baby foods less often, which may have been related to their infants’ greater acceptance of foods at weaning (17) and the decreased likelihood that they will be picky eaters as they grow (18). However, we showed no effect of mothers drinking a variety of vegetable juices during lactation on their infants’ acceptance of the novel broccoli-flavored cereal.

Two explanations, which are not mutually exclusive, are suggested. First, although we established that the sensory properties of human milk change when mothers drink carrot juice (20), to our knowledge, such research has not been conducted with the other juices (e.g., beet and celery). Thus, infants may not have been exposed to a variety of flavors. However, the evaluation of the taste of mothers’ milk by sensory panelists in the 1980s revealed a wide range of flavors that acted in an additive manner (48). Second, the taste of green vegetables is initially disliked by both infants and children (4). Because the rejection of a bitter taste is inborn, a repeated experience with the actual bitter taste of broccoli rather than an experience with the flavors of a variety of vegetables may be required to shift the hedonic tone (13, 15).

Consistent with previous research (30), mothers were motivated for and desired their children to have a healthy diet, which included vegetables, fruit, and meats. Such beliefs and parenting practices have been shown to be effective modifiers of vegetable and fruit intake in children (49). In the present study, we focused on another important barrier that needs to be addressed: how to get mothers to like how vegetables taste. How much mothers liked the vegetable juices increased over time, and how bitter the vegetable juices tasted decreased over time; this pattern was particularly true for the vegetable flavors that they initially did not like. Because the mothers in the control group tasted each of these drinks during the 5 psychophysical test sessions at the Monell Center, mere exposure to the taste, separated by as much as 1 mo (during nonexposure periods), resulted in mothers liking the taste of the juices more over time, which is a finding that is consistent with previous results that relatively few repeated tastings of a bitter-sweet beverage can significantly increase the hedonic evaluation of the taste of the beverage in adults (28). However, although liking modestly increased, vegetable and fruit intakes in these new mothers, as has been reported previously (27, 50), was well below current recommendations, and fruit intake, which is typically higher during pregnancy perhaps because of cravings for fruit (51, 52), began to decrease, which highlighted the problem at hand. The monitoring of changes in taste hedonics during future interventions may shed light on how much of a hedonic shift and how much exposure are needed to have a positive effect on vegetable intake. Pregnant women and new mothers are often motivated to improve their diets (53), and intervention studies have shown that they can change food habits, but this change will require more than just tasting vegetables a few times (54–57). In addition to taste, core barriers for healthy eating that have been reported by caregivers include a lack of meal preparation skills, difficulty in changing eating habits, and costs and availability (58).

Getting children off to a healthy start requires more than having their mothers eat vegetables during breastfeeding. Although we have identified a period during early life during which infants may be more receptive to learning about dietary volatiles of vegetables in mothers’ milk, learning to like foods has to continue throughout childhood. Although vegetables, in particular, have been most cited as being a food children do not like (59), children learn through imitation and observation and the contingencies between their action and its consequence; what infants observe influences what they do, and what they do changes behaviors (60). Children are quicker to try a new food, and they like and eat more healthy foods such as fruit and vegetables when the eating of these foods is modeled by an important other, such as a parent (61–64). The more fruit and vegetables that are eaten by the mother, the more likely it is that her preschooler will follow her lead (65). Also, more fruit and vegetables in the home result in greater consumption of fruit and vegetables by the child at school (66). Because learning to like healthy foods such as vegetables requires repeated and continued exposures to learn to like them, if vegetables are part of the mothers’ own diets, the mothers will not consider buying vegetables to be a waste of money even if their children reject the vegetables initially (67).

In conclusion, the challenge is to start early and to understand how infants develop healthy eating patterns that are sustainable, particularly in today’s food environment, which is rich in added sugars. Also, starting early cannot only focus on children but must also include their mothers and other family members. The time is ripe to focus on the diets of women during motherhood and, in turn, the types of foods the women proffer (and model) to their children to learn how to better transition children to eating healthy foods such as vegetables.