Reducing relative food reinforcement of infants using a music enrichment program: a randomized, controlled trial

ABSTRACT

Background

Food reinforcement or one's motivation to eat may be established early in life; it might not be the food reinforcement per se that drives weight gain, but rather the imbalance between food and nonfood reinforcement.

Objectives

We implemented a 2-y randomized, controlled trial to assess the effects of a music enrichment program (music, n = 45) compared with an active play date control (control, n = 45) in 9- to 15-mo-old healthy infants who were strongly motivated to eat.

Methods

The 12-mo intensive intervention phase included 4 semesters of Music Together^® or a play date program (Winter, Spring, Summer, and Fall), comprised of once per week group meetings, followed by a 12-mo maintenance phase with monthly meetings. Parents were encouraged to listen to the Music Together^® program CD or play with the play date group's toy with their infants at home, respectively. We performed a modified intention-to-treat analysis using all randomly assigned, non-excluded subjects for the outcome measures [relative reinforcing value of food (RRV_food), food reinforcement, music reinforcement, and weight-for-length z-score (WLZ)].

Results

There were significant differential group changes across time for RRV_food (group × month; P = 0.016; Cohen's f² = 0.034). The music group had significantly greater RRV_food decreases than the control group from baseline to the end of the intensive intervention phase (music change = −0.211; control change = −0.015; P = 0.002; Cohen's D = 0.379). However, these differences were not maintained during months 12–24 (music change = −0.187; control change = −0.143; P = 0.448; D = 0.087). We observed an overall moderation effect by sex for food reinforcement and WLZ. Boys in the music group had a significant attenuation in food reinforcement and WLZ compared with boys in the control group.

Conclusions

This study extends our knowledge in infant eating behavior by providing insight into the role of nonfood alternatives in altering one's motivation to eat. There may be sex differences in altering one's motivation to eat.

This trial was registered at clinicaltrials.gov as NCT02936284.

See corresponding editorial on page 1462.

Introduction

Food is a primary reinforcer, because infants are innately motivated to eat for survival purposes (1). During the first year of life, eating behaviors undergo rapid evolution as infants grow and explore their environments. For the first few months of an infant's life, autonomy over how to spend their time and energy might be limited. However, by the end of the first year, as they have gained more motor control, they have more choices about how to allocate their resources. For example, an older infant may be able to choose between eating or playing with their toys/siblings. One way to study eating in the context of multiple behavioral alternatives is to measure the relative reinforcing value of food (RRV_food), specifically studying how hard individuals will work for food compared with a nonfood alternative. Preschoolers, older children, and adults with obesity have a higher RRV_food than their leaner peers (2–8). A high RRV_food predicts greater weight gain for children, adolescents, and adults (9–11), and is related to energy intake (12, 13).

Our laboratory was the first to adapt the relative reinforcing value task for infants aged 9–15 mo in a developmentally appropriate laboratory setting (14, 15). We have studied food reinforcement compared with that of 3 different nonfood alternatives in 3 cohorts of infants: Baby MacDonald^™ video (DVD; n = 27), playing with bubbles (Bubbles; n = 30), and listening to music (Music; n = 49) (16). We observed a positive relation between RRV_food and infant weight-for-length z-score (WLZ) across the 3 different types of alternatives. A high RRV_food implies that one has a higher motivation to obtain food than a nonfood reinforcer. One of the most intriguing findings of these infant studies is that when we examined food reinforcement itself, there was no significant association between food reinforcement and weight status among infants aged 9–15 mo; however, those with a higher WLZ found the alternatives to food significantly less reinforcing than those with a lower WLZ (15).

These findings suggest that it might not be the food reinforcement per se that drives weight gain, but rather the imbalance between food and nonfood reinforcement. The value of any reinforcer is not fixed but depends on the context, which includes the availability of alternative choices (17). Thus, we further hypothesized that RRV_food is a modifiable, early, proximal behavior of the obesity phenotype. To study this possibility, our laboratory conducted a 6-wk pilot trial, randomly assigning infants, aged 9–15 mo, to a music enrichment program to strengthen alternative reinforcers to food or a play date control (14). The pilot results showed that infants in the music group significantly decreased their RRV_food compared with the control infants.

The present study extended this research in a larger sample of infants studied over a 2-y period to provide a better test of the short- and long-term efficacy of this intervention approach on obesity-related outcomes. We implemented a randomized, controlled trial to assess the effects of a music enrichment program compared with an attention placebo play date control in 9- to 15-mo-old healthy infants (n = 94) who were strongly motivated to eat as opposed to engage in other behaviors. The 12-mo intensive intervention phase included 4 semesters of Music Together^® or a play date program (Winter, Spring, Summer, and Fall), comprised of once per week group meetings, followed by a 12-mo maintenance phase with monthly meetings. We hypothesized that infants in the music enrichment program would have a lower RRV_food and WLZ after the intensive intervention phase and that these effects would be maintained at the follow-up maintenance phase compared with control infants. As an exploratory analysis, we tested infant sex as a moderator of intervention effects in all models, given sex differences in RRV_food (2, 5, 15) and sex differences in physical growth trajectories.

Methods

Participants

Families with infants aged 9–15 mo were recruited from October of 2016 to April of 2019 using posted flyers, Craigslist, Facebook, and an existing database. Infants were excluded from participation if they were preterm (<37 weeks of gestation); they had a low birth weight (<2500 g); they had known developmental delays or disabilities according to maternal report; the maternal age was <18 y at the time of pregnancy; any maternal smoking or illicit substance use during pregnancy was reported; maternal alcohol use of >4 alcoholic drinks on a single occasion or an average of >1 alcoholic drink per day during pregnancy was reported; or the pregnancy was a high-risk pregnancy (e.g., occurrence of placenta abruption, pre-eclampsia, gestational diabetes). Given the reinforcement task demands, we also excluded infants who were not yet eating finger foods and families who were unwilling to participate in a 2-y intervention. A total of 183 families enrolled and were screened in our laboratory. Ninety-four families met the criterion of having infants with a high RRV_food, defined by responding relatively more for food than music at baseline {RRV_food = reinforcing value of food (Food P_max)/[Food P_max + reinforcing value of music (Music P_max) ≥ 0.5]}. We excluded 4 families from the final analysis owing to physician-diagnosed autism (n = 3) and epilepsy (n = 1), which yielded n = 45 for the music group and n = 45 for the control group. Figure 1 provides a participant flowchart.

FIGURE 1.

Participant flowchart. COVID, coronavirus disease; FRR, food reinforcing ratio.

Study design

The present study (NCT02936284) is a randomized, controlled trial to assess the effects of a nonfood alternative enrichment program (Music Together^®; music group) compared with an attention placebo play date program (control group) on infant RRV_food and WLZ. The program consisted of a 12-mo intensive intervention phase and a 12-mo follow-up maintenance phase. The intensive phase consisted of 4 semesters of a weekly intervention (Winter, Spring, Summer, and Fall) with breaks in between semesters. The follow-up phase consisted of monthly classes. This study is a mixed design with the group as the between-subject factor, and baseline, 3-, 6-, 12-, and 24-mo assessments as the within-subject factor. Eligible infants were stratified by sex, age, and baseline RRV_food, and randomly assigned to music (n = 47) or control (n = 47) groups, comprising 6 cohorts of families (refer to Figure 1 for the participant flowchart). One caregiver served as the participating parent who attended most of the classes and assessment visits with his/her infant (mother, n = 85; father, n = 6; grandparent, n = 2; other, n = 1).

Sample size calculation

Our sample size was based on effect sizes from pilot data on an enriched music experience compared with play dates for relative food reinforcement (Cohen's D = 0.978), and standardized WLZ (D = 0.712) (14), which could be observed at 80% power and α = 0.05 with 32 participants/group. To account for attrition, we recruited an additional ∼25% of parent–infant dyads at baseline, which increased our participants’ number to 47 dyads/group or 94 dyads in total.

Procedures

The study procedures were approved by the State University of New York at Buffalo Social and Behavioral Sciences Institutional Review Board. Interested parents were screened via telephone interview or an online questionnaire, and eligible infants were scheduled for three 45-min laboratory appointments (preintervention assessment). Visits were scheduled during a time when the parent felt the infant would be awake, alert, and willing to do the food/nonfood reinforcement task. The 3 appointments were scheduled within 1 wk of each other and at the same time of day, if possible, to control for alertness, hunger cues, and mood. Before the laboratory visit, the parent was asked to list 1 of his/her infant's favorite solid foods and rank the favorite food on a 7-point Likert-type scale. Only food that was ranked between 6 and 7 was used as the test food. The parent also indicated during the initial screening if his/her infant had yet to try solid foods. Parents were instructed to avoid feeding their child for 1 h before the visit, although the child should also not be hungry upon arrival. Parents were also instructed to provide the infant's favorite solid food for the food portion of the task.

We considered providing a standardized food for all infants, but there was a possibility that infants might not have been introduced to the standardized food, which could be hazardous if it was consumed for the first time in the laboratory. In addition, there may have been infants who would not eat the standardized food, making the task impossible to do. Even though parents provided a wide range of foods, their infants and toddlers consumed limited categories of food. More than 80% of our infants consumed the following 3 categories of food: cereal/grains (e.g., puffs, crunchies), fruits (e.g., strawberries, blueberries), and crackers/cookies (e.g., Goldfish, Graham crackers). We standardized the size of food provided per trial, so that children who were similarly motivated to eat would consume a similar portion of food. Because infant food preferences can change during development, parents were asked to bring the child's most current favorite food to the follow-up assessments.

Upon arrival to the laboratory, parents were given a brief description of study protocols and completed a consent form for their infant's participation and a daily health condition survey. While parents were completing these forms, researchers interacted with the infants, using toys and reading books to establish a relationship for facilitating completion of the food/nonfood reinforcement computer task (details follow). After the completion of the food/nonfood reinforcement task at their third visit, research staff measured the height and weight of the parent and infant. Infant environmental enrichment measures (i.e., parent–infant interactions and Infant Toddler-Home Observational Measurement of Environment) were collected and are reported elsewhere. Parents were debriefed and received compensation for their participation.

Families qualified for the study if their child had a high RRV_food (as described here) at baseline (preintervention assessment). After baseline data collection, eligible families were randomly assigned to the music or control group. After randomization, participating families attended laboratory assessments for the food/nonfood reinforcement task, allowing staff to collect data on anthropometrics at 3, 6, 12, and 24 mo.

Intervention

The music enrichment and play date programs were scheduled as weekly sessions structured into 4 semesters: Winter (10-wk), Spring (10-wk), Summer (7-wk), and Fall (10-wk), during the first year of the study (intensive phase). During the second year (follow-up phase) families attended monthly sessions. The programs were in a group setting for each cohort of families. Families were given opportunities to attend make-up sessions if they missed any session.

Music enrichment program

Families in the music group attended the Music Together^® program, which was designed in collaboration with a local music studio, Betty's Music Together. One research staff member was present to record attendance and do recordings. This program introduces infants to the pleasure of music-making with their parents rather than them passively hearing music from CDs or TV. This program provides a rich variety of musical and playful activities, which encourage infants and parents to participate at their own level in singing, moving, listening, or exploring musical instruments. Participating parents and infants attended 45-min classes as a group. Besides attending classes, parents were encouraged to listen and sing together with their infants at home during everyday activities such as bath time, mealtime, and bedtime using the CD and instructional song book provided by the program.

Play date control program

Families in the control group attended the 45-min group play dates at the same intervals and time as the music group. One research staff member facilitated these play dates to record attendance and do recordings. Before the arrival of parents and infants, different stations were set up and the play date room was child proofed. We provided a variety of age-appropriate toys (excluding musical toys) for participating parents and infants to play with and enjoy. Parents were encouraged to interact with other parents and infants during the 45-min play time. Research staff were present at the play dates, but they did not interact with parents or infants unless concerns or questions arose. Besides attending play dates, families were given toys to play with at home. Parents were told it was a way for them to spend time bonding with their children. Each family was provided with 2 toys to play with at home each semester during the first year (8 different toys were given, in total), and a total of 2 toys were given during the second year. Parents were encouraged to play with their child at home during everyday activities such as bath time, mealtime, and bedtime using the toy provided by the program as well as other toys the child possessed.

Attendance and home practices

Parents in both groups were advised to attend all classes, with opportunities to attend make-up sessions. In addition, for the first 3 mo, components of an ecological momentary intervention (EMI) were used to foster at-home engagement by prompting the parents to sing songs or use the play date toy with their child daily. Each day, each family received 1 text in the morning to encourage them to sing songs or use the play date toy with their child. In the evening, each family was prompted to fill out the daily home practice survey and was reminded again to either sing songs or use the play date toy with their child. Families who completed the required home practice surveys were entered each week into monthly drawings (during the intensive phase) for a $50 gift card to a local grocery or coffee shop.

Treatment fidelity

All group leaders used structured protocols to implement the treatments, and they completed checklists to ensure they covered all points in the protocol. Music group leaders were trained by an experienced Music Together^® director, and the play date group leaders were trained by program staff experienced in running play dates. Sessions were recorded by research staff for observation to ensure adherence to the protocol. The project coordinator reviewed gathered data weekly for the first month of the intervention period, monthly up to month 6, and bimonthly thereafter for each cohort.

Measures

Food/nonfood reinforcement task

This computerized task was originally developed to assess the reinforcing value of food in children (5, 18) and adults (19), and we made developmentally appropriate modifications for infants (15). The development (15) and validity (20) of the paradigm have been published. All visits were scheduled during a time the parent anticipated that the infant would be awake, alert, and willing to do the food/nonfood reinforcement task. Infants aged 9 mo and older can hit/press a large button with intentionality and understand contingency as seen by our team previously and in other studies (15, 21). Whereas deliberate button presses on a conventional 3-button computer mouse would likely be unfeasible, we used a touch-sensitive single-button mouse that had a large surface area (Microsoft® Touch Mouse) as the response manipulandum. While parents were completing study forms, researchers interacted with the infant using toys and reading books to establish a relationship for facilitating task completion. This orientation period lasted 10–20 min, until the infant had acclimated to their surroundings, as confirmed by the parent. Then, the infant was placed in a highchair next to the parent to avoid separation anxiety and stranger anxiety experienced by infants at this age (22), but the parents were informed not to interact with the child during any of the decision tasks. Research staff then directed the infant's attention to the mouse button. To reduce responding for novelty, infants were exposed to the computer mouse before starting the task. To train infants to respond via the mouse, research staff demonstrated how to press the mouse button to create a “funny noise,” and infants’ button presses released a noise when the mouse was pressed. Once the infant appeared to understand how to control the noise with a button press, they were allowed ∼30 s to play with the mouse button. Training was repeated before studying each type of reward. Throughout the experiment, researchers remained neutral in their instructions to the infant and only used scripted cue phrases. To remove any parental bias, parents were instructed to only interact with the child during the training using scripted phrases or phrases used by the researchers.

After training, the task began. A reward (portion of food or access to music) was presented after the infant met the response criteria. When food was earned during the task, the researcher placed a piece of the infant's favorite solid food (approximately 1 cm × 1 cm × 1 cm) in front of them. Infants were provided the opportunity to consume their food as it was earned. When music was earned, research staff played a song for ∼10 s while simultaneously playing musical instruments. The songs used for all assessments were not used in the music program to avoid familiarity and biases. Infants worked for access to the food or nonfood reinforcer in a counterbalanced order over 2 d (Day 1: training; Day 2: food or nonfood; Day 3: food or nonfood). The task ended when the infant, consistently for 60 s, exhibited 1 of 4 behaviors (i.e., fussing/crying, communication, distraction, avoidance) to show that she/he was done playing. The project coordinator reviewed 30% of the videos to ensure compliance and check if appointments needed to be redone.

We used a developmentally appropriate schedule of reinforcement during the 2-y span of the study. At baseline, 3-, and 6-mo assessments, if the infants were 18 mo of age or younger, the schedule of reinforcement began with 1 button press to earn a reward, and increased linearly every 2 trials, up to a maximum of 30 responses (i.e., 1,1,2,2,3,3,…,15,15); if they were 19 mo old or older during the 6-mo assessment, they followed the schedule of reinforcement that began with 1 button press to earn a reward, and increased every trial, up to a maximum of 30 responses (i.e., 1,2,3,4,5,…,29,30), and this schedule was used in all 12-mo assessments (at ages 21–27 mo). At the 24-mo assessment (at ages 33–39 mo), participants followed a reinforcement schedule that began with 1 button press and increased in counts of 2, up to a maximum of 60 responses (i.e., 1,2,4,6,8,…,58,60). The number of responses, number of rewards earned, and time spent in each session were recorded.

The dependent measure was a standard measure of reinforcing value, operationalized by P_max, or the highest schedule the infant was required to complete in order to earn a food or nonfood alternative. To examine P_max across each time point, a percentage of maximum schedule was calculated [(P_MAX/max schedule) × 100%]. The food and nonfood reinforcers were offered to the infants in a sequential fashion in each cohort, and the food reinforcing ratio (FRR) was calculated as effort to obtain food out of total effort; however, for ease of interpretation and comparison with other populations, the term RRV_food is used instead of FRR throughout this article. The RRV_food calculated the proportion of responding for food [P_max food/(P_max food + P_max nonfood alternative)].

Anthropometrics

Research staff measured the infant's or toddler's (<24 mo old) weight using a Seca 374 Digital Baby Scale and length (supine position) using a Seca 416 Infantometer. When the participants turned 24 mo old or older, weight was measured using a digital weight scale (Tanita) and standing height was measured using a stadiometer (23). Research staff were trained to obtain infants’ and toddlers’ lengths by following the WHO's Multicentre Growth Reference Study protocols (24). All length and height measurements were taken and recorded independently by 2 research staff, who then compared their values. If the measurement difference was greater than the maximum allowable difference (7.0 mm) between the 2 staff, the infant/toddler was remeasured (24). The mean value of the last 2 measurements was recorded as the infant length. Infants’ and toddlers’ WLZ and weight-for-age and length-for-age z-scores were calculated using the WHO infant growth chart (for children 0–5 y of age) (25).

Energy intake

Parents were contacted by phone on 3 separate occasions (2 weekdays and 1 weekend day) to obtain three 24-h dietary recalls for their infant. Trained researchers followed modified procedures from the Feeding Infants and Toddlers Study (FITS) 2008 (26). All phone calls took place within 10 d of obtaining child recumbent length and weight and occurred at a suitable time for the participant. Parents were asked about foods and beverages given by them and other caregivers (e.g., childcare providers). Families with a child in childcare were asked to report foods and beverages based on reports from the provider. At the start of each call, researchers asked if the previous 24 h were a typical eating day for the infant. If not, the recall was rescheduled. Trained researchers collected infant dietary recalls from caregivers (as proxies) using the USDA Automated Multiple-Pass Method (AMPM), a computer-assisted dietary interview that includes standardized probes to collect dietary information on all foods and beverages consumed during the previous 24 h (27).

Trained researchers entered dietary recalls into the Nutrition Data System for Research (NDSR; version 2019; Nutrition Coordinating Center, University of Minnesota) dietary analysis program (28). When reported brand-name foods were missing from the program, a generic food with a comparable nutrition profile was selected by the researcher. For parents who were breastfeeding, procedures from the FITS were used to estimate energy and nutrients from breast milk (29). Researchers entered 600 mL breast milk/d for exclusively breastfed infants 9–12 mo of age and 29.6 mL (1 fluid ounce) per 5 min of breastfeeding for toddlers 12–15 mo of age. For infants who consumed breast milk and formula, the volume of formula reported was subtracted from 600 mL and the remainder entered as breast milk.

Demographic and socioeconomic questionnaire

Demographics and socioeconomic statuses (SESs) were assessed using the MacArthur Questionnaire developed by the MacArthur Foundation (27–29). This questionnaire assesses objective measures of SES, including years of education, and subjective measures of the family's social standing in the community. Measurements also include both individual and household income levels, along with household size. Years of education were coded separately for each parent, and each family was given a highest education score.

Pregnancy history and feeding practices questionnaire

In order to obtain information regarding breastfeeding duration and timing of solid food introduction, parents were asked to complete a shortened version of the feeding questionnaire from the Infant Feeding Practices Study II (30). This questionnaire consists of questions that assess the initiation and duration of breastfeeding, timing of solid food introduction, infant's/toddler's age at breastfeeding cessation, and gestational information of the mother. This questionnaire was administrated online via SurveyMonkey.

Impact of coronavirus disease 2019 on intervention implementation and data collection

As the coronavirus disease 2019 (COVID-19) pandemic interrupted normal procedures, several adaptations were made so that we could continue our intervention implementation and data collection in a remote and safe environment. Both music and play classes that were intended to occur in-person were transitioned to virtual Zoom video conferences. Classes met at the same day and time as their in-person classes but were shortened to 30 min because of the difficulty in keeping young children in front of a computer screen for a long period of time. Parents of both classes were instructed to set up their cameras in a place where they could interact with their child. The music program utilized the same instructor from Betty's Music Together^® to conduct the classes. The play program utilized 1 research staff member to oversee the class. Parents in the play classes were instructed to use the time to play with toys they had available in their home. The virtual classes commenced in March of 2020 and ended in March of 2021. Families in study cohorts 4 and 5 were in their monthly follow-up maintenance phases and families in cohort 6 were ending their last semester of the intensive intervention phase when we started all virtual classes. Families in cohort 6 completed their entire monthly follow-up maintenance phase virtually.

In terms of data collection, changes were made to laboratory assessments based on a family's availability and comfort level. Owing to mandates by the state and our university, our laboratory ceased to conduct in-person visits from mid-March to mid-June of 2020. When in-person laboratory procedures were allowed to resume, staff and parents were required to wear masks at all times while they were in the building and to pass a COVID-19 screening, which included a temperature check and completion of a questionnaire. The only families who were affected by COVID-19 for the 12-mo assessment, after the intensive intervention phase, were families of cohort 6 (n = 15). The original assessment dates for families of cohort 6 were in April of 2020, but their assessments were delayed to June–July of 2020. Families who were affected by COVID-19 for the 24-mo assessments, after the follow-up maintenance phase, included cohorts 3, 4, 5, and 6. The 24-mo assessment for families in study cohort 3 was in March of 2020. A total of 7 families were able to complete all assessments in-person; 4 families were able to complete contactless anthropometric measurements; 5 families provided their child's anthropometrics from their most recent well-child visit. For contactless anthropometric measurements, participants were asked if they would be comfortable having a scale and infantometer dropped off at their house. Parents who were comfortable with this procedure received written instructions along with live video assistance from a research staff member to measure their child's height and weight at home. Families in study cohorts 4, 5, and 6 were able to attend in-person 24-mo laboratory assessments and precautionary actions were taken while they were in the laboratory. For the analyses, COVID impact was captured and coded in a 3-level COVID class variable: 1) completed all assessments before COVID onset (n = 29); 2) 12- and 24-mo assessments occurred during COVID (n = 15); and 3) only 24-mo assessments occurred during COVID (n = 46).

Statistical analysis

Group baseline characteristics were compared using ANOVA or Pearson's chi-square test for frequency data. We performed a modified intention-to-treat (ITT) analysis using all randomly assigned, non-excluded subjects. The primary outcomes (RRV_food, food reinforcement, music reinforcement, WLZ, and energy intake) were analyzed by using a mixed-model ANCOVA, which handles missing data using maximum likelihood estimation and retains all randomly assigned subjects in the analysis. In an effort to choose a reasonably fitted model satisfying the model assumptions, the residual plots and the residual-based fit statistics such as Akaike's Information Criterion for various covariance structures were examined. The models included group, months, and the group × month interaction as class variables using an autogressive (AR1) covariance structure and accounted for the clustering of multiple observations within participants as well as the nesting of infants within their respective cohorts. Models included children's sex, birth weight, breastfeeding duration, and percentage class attendance as covariates as well as a 3-level COVID categoric covariate that indicated families who were influenced by the onset of the COVID pandemic during their follow-up assessments and when they were affected. In addition to the omnibus group × month effect, planned comparisons tested for group differences from baseline to each follow-up assessment (3, 6, 12, and 24 mo) based on the fitted mixed models. Infant's sex and COVID were tested in separate models as moderators of group effects on each outcome change by including sex or COVID as well as its interactions with group, month, and group × month effects. These were also followed by planned comparisons to estimate group × month effects separately for boys and girls at each follow-up assessment (3, 6, 12, and 24 mo). All models included the aforementioned covariates. Effect sizes were reported as Cohen's f² for mixed models and Cohen's D for contrasts. All mixed models were performed using PROC MIXED in SAS version 9.4 (^©2020, SAS Institute). Effect sizes provided are based on these models and contrasts.

Results

Participant demographics, attendance, and attrition

Ninety-four infant–parent dyads were randomly assigned. The final sample consisted of 90 parent–infant dyads for the modified ITT analysis. Of the 90 dyads, 69 completed the 12-mo assessment after the intensive intervention phase (76.7%) and 71 completed the 24-mo assessment after the follow-up maintenance phase (78.8%). Attrition at 24 mo did not differ between the groups (music = 20.0%, control = 22.2%; chi-square = 0.067, P = 0.796). Table 1 presents demographic and anthropometric data on the parent–infant dyads for both groups at baseline. Our sample consisted of primarily highly educated (college graduate or higher = 71.1%), mid-upper income (≥$50,000 = 81.5%) white families (mothers who self-identified as white = 86.7%). There was a significant group difference in class attendance (music = 75.9%, control = 50.1%; P = 0.001), but there were no child biological sex (P = 0.397) or group × sex (P = 0.414) differences.

TABLE 1.

Child, maternal, and household characteristics of participants¹

Variable	Music group (n = 45)	Control group (n = 45)
Child
Sex, male	21 (46.7)	22 (48.9)
Age, mo	12.10 ± 1.69	11.93 ± 2.05
Race
African American	0 (0.0)	2 (4.4)
American Indian	0 (0.0)	1 (2.2)
Asian	1 (2.2)	0 (0.0)
White	35 (77.7)	37 (82.2)
Multiracial	9 (20.0)	4 (8.9)
No response	0 (0.0)	1 (2.2)
Ethnicity
Hispanic (single race)	0 (0.0)	0 (0.0)
Hispanic (multiracial)	4 (8.9)	2 (4.4)
Birth weight, kg	3.45 ± 0.54	3.51 ± 0.48
Weight-for-length z-score2	−0.96 ± 1.63	−0.55 ± 1.72
Weight-for-age z-score2	0.28 ± 1.08	0.42 ± 0.93
Length-for-age z-score2	−0.57 ± 1.38	−0.25 ± 1.26
Breastfeeding duration, mo
<6	18 (40.0)	12 (26.7)
≥6	27 (60.0)	33 (73.3)
First introduction to solids, mo
<4	2 (4.4)	1 (2.2)
4–5	23 (51.1)	17 (37.8)
≥6	20 (44.4)	27 (60.0)
Baseline RRVfood	0.66 ± 0.14	0.65 ± 0.16
Baseline Food Pmax	6.00 ± 3.69	5.11 ± 2.80
Baseline Music Pmax	3.29 ± 2.63	2.98 ± 1.99
Mother
Age, y	31.6 ± 3.9	33.2 ± 4.1
Race
African American	1 (2.2)	2 (4.4)
American Indian	0 (0.0)	2 (4.4)
Asian	1 (2.2)	1 (2.2)
White	39 (86.7)	39 (86.7)
Multiracial	4 (8.9)	0 (0.0)
No response	0 (0.0)	1 (2.2)
Ethnicity
Hispanic (single race)	0 (0.0)	0 (0.0)
Hispanic (multiracial)	3 (6.7)	0 (0.0)
Education level
Some college or less	13 (28.9)	13 (28.9)
College graduate or more	32 (71.1)	32 (71.1)
Parity,3 primiparous	24 (53.3)	24 (53.3)
Current BMI, kg/m2	30.3 ± 7.8	30.1 ± 7.6
Normal	12 (26.7)	14 (31.3)
Overweight	18 (40.0)	15 (33.3)
Obese	15 (33.3)	16 (35.6)
Household
Total income, $
<50,000	7 (15.5)	9 (20.0)
≥50,000	35 (77.8)	35 (77.8)
No response	3 (6.7)	1 (2.2)

¹Values are mean ± SD or n (%). Food P_max, reinforcing value of food; Music P_max, reinforcing value of music; RRV_food, relative food reinforcement.

²Calculated using the WHO infant growth chart (at birth).

³Nulliparas = first-time pregnant; paras ≥1 = with ≥1 pregnancy.

Relative food reinforcement, weight, and energy intake change

Figure 2 shows changes in RRV_food. There were significant differential group changes across time for RRV_food (group × month; P = 0.016; f² = 0.034). The music group had significantly greater RRV_food decreases than the control group from baseline to the end of the intensive intervention phase (music change = −0.211; control change = −0.015; P = 0.002; D = 0.379). However, these differences were not maintained between months 12 and 24 (music change = −0.187; control change = −0.143; P = 0.448; D = 0.087).

FIGURE 2.

Changes in RRV_food at all assessment time points (baseline, 3, 6, 12, and 24 mo) during the 12-mo intensive intervention and 12-mo follow-up maintenance phase. Ninety 9- to 15-mo-old infants who were highly motivated to eat were randomly assigned to either the music enrichment program (music) or play date control (control) group. A model adjusted for covariates is presented and data are expressed as mean ± SEM. There were significant differences across time between the groups for RRV_food (group × month, P = 0.016; f² = 0.034). The music group had significantly greater RRV_food decreases than the control group from baseline to the end of the intensive intervention phase (music change = −0.211; control change = −0.015; P = 0.002; D = 0.379). However, these differences were not maintained between months 12 and 24 (music change = −0.187; control change = −0.143; P = 0.448; D = 0.087). RRV_food, relative reinforcing value of food.

When we examined the food reinforcement and music reinforcement individually, there were no differential group changes across time. For food reinforcement, infants in the music group had a continuous decrease in food reinforcement from baseline to the end of the intensive intervention phase (music change = −17.052; P < 0.001) with a slight decrease from 12 to 24 mo (music change = −0.556; P = 0.901); infants in the control group showed a slight decrease from 6 to 24 mo (control change = −0.869; P = 0.081). For music reinforcement, infants in the music group had a continuous increase in music reinforcement from baseline to month 6 (music change = 13.424; P < 0.001) before it started to decrease from 6 to 24 mo (music change = −11.191; P = 0.002); infants in the control group showed a decrease in music reinforcement from 0 to 12 mo (control change = −3.294; P = 0.394), with an increase from 12 to 24 mo (control change = 4.629; P = 0.233).

Figure 3 shows changes in infant WLZ. No significant differences between groups in WLZ changes were observed over time. From baseline to the intensive intervention phase, the change in WLZ was +0.073 (P = 0.540; D = 0.105) for infants in the music group compared with +0.180 (P = 0.177; D = 0.272) for infants in the control group. None of the covariates (children's sex, birth weight, breastfeeding duration, class attendance, and COVID) were significant in any of the models.

FIGURE 3.

Changes in WLZ at all assessment time points (baseline, 3, 6, 12, and 24 mo) during the 12-mo intensive intervention and 12-mo follow-up maintenance phase. Ninety 9- to 15-mo-old infants who were highly motivated to eat were randomly assigned to either the music enrichment program (music) or play date control (control) group. The model adjusted for covariates is presented and data are expressed as mean ± SEM. No significant differences in WLZ changes over time were observed between the groups. From baseline to the intensive intervention phase, the change in WLZ was +0.073 (P = 0.540; D = 0.105) for infants in the music group compared with +0.180 (P = 0.177; D = 0.272) for infants in the control group. WLZ, weight-for-length z-score.

On average, infants consumed 897.1 ± 24.86 kcal at baseline, 1135.6 ± 28.04 kcal at month 12, and 1232.4 ± 27.24 kcal at month 24 assessments. No significant differences between groups in energy intake changes were observed over time (P = 0.681). There was no significant difference in energy changes from baseline to month 12 (268.9 kcal for music, 208.0 kcal for control) or baseline to month 24 (346.1 kcal for music, 324.5 kcal for control) between groups.

Sex differences in relative food reinforcement, weight, and energy intake change

We observed an overall moderation effect of sex for food reinforcement (group × month × sex, P = 0.042; f² = 0.022), with sex moderating between-group differences from 0 to 3 mo (P = 0.004; D = 0.628) and at 24 mo (P = 0.032; D = 0.462) (see Figure 4). Food reinforcement changes from baseline to month 24 were significantly different by group; boys in the music group had a significant decrease from baseline to 24 mo (β = −24.29 ± 9.42; P = 0.010; D = 0.278) as opposed to girls, in whom the groups did not differ over 24 mo (β = +2.61 ± 8.84; P = 0.768; D = 0.032).

FIGURE 4.

Changes in food reinforcement (food %max schedule) at all assessment time points (baseline, 3, 6, 12, and 24 mo) for boys and girls during the 12-mo intensive intervention and 12-mo follow-up maintenance phase. Ninety 9- to 15-mo-old infants who were highly motivated to eat were randomly assigned to either the music enrichment program (music: boys, n = 21; girls, n = 24) or play date control (control: boys, n = 22; girls, n = 23) group. The model adjusted for covariates is presented and data are expressed as mean ± SEM. An overall moderation effect of sex for food reinforcement (group × month × sex, P = 0.042; f² = 0.022) was observed with sex moderating between-group differences from 0 to 3 mo (P = 0.004; D = 0.628), and at 24 mo (P = 0.032; D = 0.462).

Sex also moderated the group × month effect for WLZ (group × month × sex, P = 0.033, f² = 0.023), with sex moderating group differences from 0 to 3 mo (P = 0.022; D = 0.493), 12 mo (P = 0.004; D = 0.624), and 24 mo (P = 0.031; D = 0.465) (see Figure 5). Boys in the music group showed significant WLZ attenuation compared with those in the control group from baseline to the end of the intensive intervention phase (β = −0.59 ± 0.26; P = 0.021; D = 0.247), whereas no group differences were observed among girls. Infant sex was not a significant moderator for RRV_food or music reinforcement.

FIGURE 5.

Changes in WLZ at all assessment time points (baseline, 3, 6, 12, and 24 mo) for boys and girls during the 12-mo intensive intervention and 12-mo follow-up maintenance phase. Ninety 9- to 15-mo-old infants who were highly motivated to eat were randomly assigned to either the music enrichment program (music: boys, n = 21; girls, n = 24) or play date control (control: boys, n = 22; girls, n = 23) group. The model adjusted for covariates is presented and data are expressed as mean ± SEM. An overall moderation effect of sex for WLZ (group × month × sex, P = 0.033; f² = 0.023) was observed with sex moderating group differences from 0 to 3 mo (P = 0.022; D = 0.493), 12 mo (P = 0.004; D = 0.624), and 24 mo (P = 0.031; D = 0.465). WLZ, weight-for-length z-score.

We did not observe an overall moderation effect of sex for energy intake (group × month × sex, P = 0.260). Change in food reinforcement was significantly correlated with the change in WLZ among boys from baseline to the end of the intensive intervention phase (r = 0.455, P = 0.017) and baseline to month 24 (r = 0.362, P = 0.050). On the other hand, change in food reinforcement was not significantly correlated with the change in WLZ among girls from baseline to the end of the intensive intervention phase (r = 0.108, P = 0.557) and baseline to month 24 (r = 0.035, P = 0.840) postintervention. Change in energy intake was significantly correlated with the change in WLZ (r = 0.268, P = 0.032) and food reinforcement (r = 0.381, P = 0.041) among boys from baseline to month 24, but not girls.

Impacts of COVID-19 on relative food reinforcement and weight change

We did not observe an overall moderation effect of COVID for RRV_food (group × month × COVID, P = 0.151), food reinforcement (P = 0.995), music reinforcement (P = 0.053), WLZ (P = 0.221), or energy intake (P = 0.541) between groups.

Discussion

Our findings provide evidence that alternatives to food may alter the reinforcing value of food in young children who are strongly motivated to eat. Infants significantly reduced their RRV_food after a 12-mo music enrichment program. We also observed a moderating effect by sex in that the music enrichment program may be more beneficial for boys than for girls. Boys in the music group had a significantly lower food reinforcement and less weight gain than boys in the play date control group; however, interestingly, girls seem to have benefited from the active control (the play date). The change in food reinforcement was significantly correlated with change in weight status among boys, but not girls.

We previously demonstrated that food reinforcement could be measured in 9- to 18-mo-old infants, and that overweight infants did not work as hard for a nonfood reinforcer as did normal-weight infants (15). Our current intervention demonstrated that it is possible to alter the RRV_food of infants by cultivating alternatives to food in their environment. One hypothesis is that the development of obesity may be due to a lack of pleasurable alternatives in one's environment, thereby increasing the reinforcing value of eating (31). These findings may provide evidence of the role of alternative reinforcers as a protective factor against overeating among young children. Despite the success in reducing the RRV_food in children who were in the music enrichment program after the intensive intervention phase, we did not observe a sustained effect at the 24-mo follow-up phase. It is common to observe challenges in long-term behavior change (32), with the deterioration in treatment effects occurring when treatment intensity was reduced from weekly to monthly. The shift in meeting frequency may have been associated with a reduction in parent/child interactions in regard to music, which could affect treatment effects. However, the biggest change was an improvement in RRV_food for the control group, who showed an increase in music reinforcement.

In addition to the change in frequency of intervention meetings as a cause for the deterioration in treatment effects, almost all children in our study had entered preschool age by the end of the follow-up maintenance phase at month 24; therefore, it is possible that children in the control group were exposed to music at preschool that might have increased their reinforcing value for music. Perhaps children in the music group were being exposed to music in preschool as well, but they might have reached the ceiling effect for children of this age group. Second, our proposed intervention only tested 1 alternative reinforcer: music engagement. One of our future goals is to identify other alternative reinforcers that could be incorporated into the program, such as infant reading/story time, dance/gymnastic classes, etc., to increase the variety of nonfood reinforcers. Third, there is limited research on the developmental trajectory of food and nonfood reinforcement. To our knowledge, no researchers have examined RRV_food trajectory in young children. The 2-y RRV_food trajectory from infancy to preschool age observed by our present study is the first to be reported. During the first 3 y of life, eating behaviors undergo rapid evolution as infants grow and explore their environment. It is possible that the environment of children becomes more enriching as they enter preschool age. There might be individual differences in their nonfood environment that may have influenced their RRV_food.

We observed differential effects of the intervention on boys and girls, which may have been due to sex differences in enjoyment of the various nonfood activities. Research has shown that girls mature early and are more social than boys (33, 34); therefore, our play date control might have unintentionally provided social interactions that were more reinforcing to girls. Social reinforcers can be powerful reinforcers because children value positive interactions with adults, especially with parents during infancy because they are their primary source of social interaction and support (35, 36). Pleasant parental interactions, often during playtime, could be alternative reinforcers that are equally powerful to food. We have recently shown that high-quality parent–infant interactions during playtime were associated with a lower weight status cross-sectionally (37) and normal weight gain trajectories longitudinally (38). In the future, perhaps a different type of educational program can serve as the active control group in our study design. Because of the possibility of individual differences in the reinforcing value of the nonfood reinforcer, it might be crucial to test multiple nonfood reinforcers and optimize the intervention per individual preferences. It is important to note that in general, boys found food more reinforcing than did girls in our previous studies (2, 5, 15). Thus, our results might have suggested that music engagement may be a powerful reinforcer for boys because they were able to reduce their food reinforcement and weight gain.

Our research is an initial step toward developing new and innovative approaches to preventing obesity during infancy that are different from, but complementary to, the traditional approach of focusing on energy intake and expenditure. The intervention proposed is easy to implement, enjoyable for many families, and has been associated with long-term retention. The effect sizes of our interventions (i.e., f² = 0.034 and D = 0.379) are between small and medium, which is slightly higher than the small effect size observed in the landmark INSIGHT (Intervention Nurses Start Infants Growing on Health Trajectories) study, on responsive parenting for obesity prevention from infancy to 3 y of age [i.e., BMI z-score (zBMI) change for intervention group = −0.09; zBMI change for control group = 0.11; D = 0.218 or η² = 0.0118] (39). However, it is important to note that our intervention successfully reduced young children's motivation to eat, which is a proximal obesity-related behavior, and not their weight gain. If activities can be strengthened that are substitutes for eating, this approach is easily scalable and can be expanded to larger socioecological contexts, such as schools or community centers. This takes advantage of the fact that experience of a behavior in a social context can amplify the value of that behavior (40). As evidenced by our work in adults with obesity (41) and Leventhal et al.’s (42) work with adolescent substance abuse, the identification and building of alternative reinforcers may be uniquely beneficial for families with low parental education. In addition, contemporary behavioral interventions, particularly those containing behavioral parenting components, could be incorporated into these programs.

Children grow up in home environments that differ greatly in terms of their access to a variety of cognitively stimulating activities (e.g., reading, musical engagement). Theoretically, early enriched home environments that promote comfort and pleasurable behaviors which are also alternatives to eating may mitigate young children's food-seeking behavior and, thus, alter the trajectory of RRV_food. Research on enriched environments was originally used to study the effects of the environment on brain plasticity in animal models. Rodents raised in enriched environments improved their performance on learning and memory tasks, as well as showed an increase in brain size (43, 44). Later research showed that environmental enrichment plays a significant role in attenuating substance abuse. These results suggest that enriched environments can reduce sensitivity to rewards (45, 46) and maladaptive choices (47). An enriched environment, characterized by access to a variety of nonfood cognitively stimulating activities, may play a role in the development of a brain reward system that favors nonfood reinforcers over food reinforcers. Thus, the next step for our team is to analyze environmental enrichment data collected during the 2-y period of the trial.

There are a few limitations to our present study. We studied a group of participants who were from highly educated, middle-class households, and recruited infants were of mothers who did not drink, smoke, or use drugs during pregnancy. Future research should extend the study to infants of families with low-SES backgrounds because obesity development could be due to economic circumstances experienced by low-SES individuals, such as a lack of alternatives to eating and chronic deprivation. These adverse circumstances could augment food reinforcement by promoting maladaptive eating behaviors starting early in life. In addition, we only studied 1 alternative reinforcer: music. Nonetheless, this is the first evidence showing long-term (12-mo) change in the differential patterns of responding to food compared with alternative reinforcers in infants. Future work should consider studying a wider sample of alternative reinforcers to maximize the chance of identifying alternative reinforcers that can be maintained by families, especially as the infants grow.

In conclusion, results from this study extend our knowledge in infant eating behavior by providing insight into the role of nonfood alternatives in altering one's motivation to eat. Eating is often an option among many other available activities for infants and young children. Having parents foster an enriched home environment at an early age may be important in altering a child's choice of food and, perhaps, future weight trajectories. However, there may be sex differences in altering one's motivation to eat.

Notes

Author disclosures: The authors report no conflicts of interest.

Supported by National Institute of Child Health and Human Development grant R01 HD087082 (to KLK and LHE). The study sponsor had no role in study design; the collection, analysis, and interpretation of data; the writing of the report; or the decision to submit the manuscript for publication.

Abbreviations used: COVID-19, coronavirus disease 2019; FITS, Feeding Infants and Toddlers Study; FRR, food reinforcing ratio; ITT, intention-to-treat; RRV_food, relative reinforcing value of food; SES, socioeconomic status; WLZ, weight-for-length z-score; zBMI, BMI z-score.

Data Availability

Data described in this article will be made available upon request pending application and approval by KLK.