Bottle Feeding

Abstract

Bottle feeding is the delivery of milk, infant formula or expressed human milk, to an infant from a bottle rather than from the breast. Although primarily a method of nutrition, it became a sustained subject of cognitive-developmental research, first because the question of whether breastfed children score higher on intelligence tests turned feeding mode into a test case for causal inference under heavy confounding by maternal intelligence. Second, the mechanics of the bottle, and who controls the pace and amount, bear on how an infant learns to regulate intake and form early food preferences. This article traces the breastfeeding-and-cognition debate and the designs built to resolve its confounding, examines the self-regulation of intake, and sets out responsive feeding as the developmentally active ingredient. Three demonstrations make the confounding, the regulation of intake, and the learning of preferences explicit.

Keywords: bottle feeding, infant feeding, self-regulation of intake, responsive feeding, confounding

Bottle feeding is one of the two principal ways an infant is fed, and the one that hands control of the feed from the infant to the caregiver and the vessel. It is, on its face, a nutritional method rather than a cognitive one, yet it has drawn decades of attention from developmental and cognitive psychology because of what feeding mode seems to predict and what that prediction turns out to mean. Children who were breastfed score, on average, slightly higher on later tests of intelligence than children who were bottle-fed — but that difference is entangled with every other way the two groups of families differ, and disentangling it has been a methodological story as much as a nutritional one (Der, Batty, & Deary, 2006). Alongside the cognition debate, bottle feeding raises its own behavioral questions: how an infant fed from a bottle learns, or fails to learn, to stop when full, and how the feeding interaction shapes the preferences a child carries forward (Li, Fein, & Grummer-Strawn, 2010).

What Bottle Feeding Is

Bottle feeding is the feeding of an infant with milk presented in a bottle fitted with a teat, as distinct from feeding directly at the breast. The milk delivered may be commercial infant formula or human milk that has been expressed and stored, so the term names a method of delivery rather than a type of milk: a breastfed infant given expressed milk from a bottle is, by this definition, being bottle-fed for that feed. This distinction matters because the two things the method bundles together — what the infant receives, and how the infant receives it — have separate developmental consequences, and much of the research on bottle feeding is an attempt to pull them apart.

Three features distinguish feeding from a bottle from feeding at the breast, and each has drawn research attention. The first is composition: formula differs from human milk in its nutrients and lacks the bioactive and immune components of breast milk, which is the channel most often proposed for any direct effect on development (Anderson, Johnstone, & Remley, 1999). The second is control: at the breast the infant largely determines the pace and the amount, whereas a bottle makes the volume visible to the caregiver and invites feeding to a target rather than to the infant's cue (Li et al., 2010). The third is flavor: human milk carries the changing flavors of the maternal diet, while formula presents a constant taste, so the two feeding modes offer very different early sensory experience (Ventura & Worobey, 2013). The cognitive-developmental literature on bottle feeding is organized almost entirely around these three: composition drives the intelligence debate, control drives the self-regulation question, and flavor drives the study of food preferences.

The Breastfeeding–Cognition Debate and Confounding

The observation that started the literature is simple and robust: across many samples, children who were breastfed score a few points higher on later tests of cognitive ability than children who were bottle-fed. An early meta-analysis by James Anderson and colleagues put the adjusted difference at roughly three IQ points and concluded that it was real and durable (Anderson et al., 1999). A later and larger meta-analysis by Bernardo Horta and colleagues reached a similar figure of about three-and-a-half points and extended it to adult outcomes (Horta, Loret de Mola, & Victora, 2015). Taken at face value, these numbers imply that feeding mode causes a cognitive difference.

The difficulty is that feeding mode is not assigned at random. Mothers who breastfeed differ systematically from mothers who bottle-feed: on average they have more education, higher incomes, and — critically — higher measured intelligence, and they provide different home environments. Any of these can raise a child's test scores on its own, so the raw gap between breastfed and bottle-fed children is a mixture of whatever milk does and everything that comes bundled with the choice to breastfeed. This is the textbook problem of confounding: a third variable, here maternal intelligence and family circumstances, drives both the predictor and the outcome and manufactures an association between them. Figure 1 lays out the structure of the problem.

Figure 1

The confounding structure of the breastfeeding-and-cognition association: maternal intelligence and family circumstances influence both the choice of feeding mode and the child's later test scores, generating an association between feeding mode and cognition that is not wholly causal.

Two research designs were built to cut the confounding paths in Figure 1. The first is the sibling-pairs design, which compares siblings within the same family who were fed differently, holding constant everything the family shares. Geoff Der, David Batty, and Ian Deary applied this design and a meta-analysis to a large United States sample and found that once maternal intelligence in particular was accounted for, the apparent effect of breastfeeding on children's intelligence largely disappeared: the raw advantage was mostly a marker of the mothers who breastfed, not of the milk (Der et al., 2006). The second design is the only randomized one available. Because it is unethical to assign infants to be breastfed or not, Michael Kramer and colleagues instead randomized a breastfeeding-promotion intervention across maternity hospitals in Belarus — the PROBIT trial — so that the intervention, not the feeding itself, was under experimental control. Among the children whose mothers received the promotion, and who were therefore breastfed longer and more exclusively, measured intelligence at age six was higher, which is the strongest evidence that the association is not entirely confounded (Kramer et al., 2008). The first demonstration lets these adjustments be applied one at a time to a raw feeding-mode gap, making visible how much of it each confounder absorbs.

The weight of the evidence, pulled together across observational, sibling, and quasi-experimental designs, is that a small cognitive advantage associated with breastfeeding survives adjustment but is far smaller than the raw gap suggests, and that most of the headline difference between breastfed and bottle-fed children reflects who their mothers are rather than how they were fed (Belfort et al., 2013; Victora et al., 2016). For cognitive psychology the episode is a standing lesson in confounded inference: an association that is real, replicable, and dose-dependent can still be mostly an artifact of a third variable, and only designs that break the confounding path can say how much of it is causal. Table 1 sets the three design families side by side.

Table 1. Three families of design used to estimate the effect of feeding mode on cognition, what each holds constant, and what it concluded.

The Self-Regulation of Intake

The second strand of research concerns not what is in the bottle but how it is used. Infants are born able to regulate their energy intake — to take what they need and stop when sated — and feeding is one of the earliest arenas in which this self-regulation is either supported or overridden. The bottle introduces a specific hazard: because its contents are visible and measurable, caregivers are drawn to feed to a target, encouraging the infant to finish the bottle rather than to stop at the infant's own signal of fullness. Ruowei Li, Sara Fein, and Laurence Grummer-Strawn showed that this early experience leaves a trace: infants who had been fed from a bottle in early infancy were more likely, late in the first year, to empty the cup or bottle entirely, consistent with weaker self-regulation of intake than directly breastfed infants (Li et al., 2010).

The mechanism is a matter of who controls the feed. At the breast the infant sets the pace and ends the feed; with a bottle the caregiver can continue past the infant's cue, and repeated feeding past satiety is thought to blunt the infant's reliance on internal hunger and fullness signals in favor of external cues such as the amount remaining. The consequence most studied is weight: a systematic review by Kristen DiSantis and colleagues found that less responsive feeding — feeding that does not track the infant's own signals — is associated with more rapid weight gain and greater risk of overweight in infancy and toddlerhood (DiSantis, Hodges, Johnson, & Fisher, 2011). A later review by Jessica Appleton and colleagues traced the same pattern specifically to formula-feeding practices, including feeding to schedule and encouraging bottle emptying, which were associated with the rapid weight gain that predicts later obesity (Appleton et al., 2018). The second demonstration contrasts a feed governed by the infant's fullness cue with one governed by emptying the bottle, and accumulates the difference across a day of feeds.

The lesson of this strand is that the vessel is not neutral. A bottle does not force overfeeding, but it makes overfeeding easy and self-regulation harder, and the behavior it invites in the caregiver — feed to the mark — is precisely the one that teaches the infant to eat to an external cue rather than an internal one.

Responsive Feeding and the Learning of Preferences

If the vessel is not the whole story, what is? The convergent answer from the recent literature is the feeding interaction — the moment-to-moment exchange in which a caregiver reads the infant's hunger and satiety signals and responds to them. Maureen Black and Frances Aboud placed this within a broader theory of responsive parenting: feeding, like other caregiving, works best when it is prompt, contingent, and matched to the child's signals, and it is this responsiveness, not the method of delivery, that is the developmentally active ingredient (Black & Aboud, 2011). On this account a bottle-fed infant whose caregiver watches for and respects fullness cues is in a very different developmental situation from one fed to empty the bottle, even though both are, in the crude binary, bottle-fed. Rafael Pérez-Escamilla and colleagues built this insight into practical feeding guidelines organized around responsive parenting rather than around feeding mode (Pérez-Escamilla, Segura-Pérez, & Lott, 2017).

The feeding interaction also shapes what the infant comes to like. Early flavor experience is a form of learning: repeated exposure to a flavor increases its acceptance, a simple associative process sometimes called mere-exposure learning. That flavors actually transfer to the infant through the feed is an experimental finding, not an assumption: Julie Mennella and colleagues showed that distinctive flavors from the maternal diet pass into both amniotic fluid and breast milk, and that infants exposed to a flavor prenatally or through breast milk later accept it more readily, establishing flavor learning as a genuine pre- and postnatal process (Mennella, Jagnow, & Beauchamp, 2001). Here the composition difference between the feeding modes becomes a difference in sensory experience. Alison Ventura and John Worobey reviewed how the earliest influences on food preferences operate through just this exposure, with human milk conveying the varying flavors of the maternal diet and formula presenting a constant one, so that breastfed and formula-fed infants arrive at the introduction of solid foods with different flavor histories (Ventura & Worobey, 2013). Ventura later drew out the link to later outcomes, arguing that these early feeding experiences help set the preferences and intake patterns that bear on obesity risk (Ventura, 2017). The third demonstration reproduces the exposure-learning curve, comparing how acceptance of a novel food grows under a varied versus a constant early flavor history. The acceptance of such a food is itself partly a matter of taste perception and its early tuning.

Taken together, the responsive-feeding literature reframes the whole subject. The developmentally consequential variable is not breast-versus-bottle but responsive-versus-nonresponsive, and the bottle matters chiefly because of the feeding behavior it tends to invite.

Worked Example

Because the breastfeeding-and-cognition association is confounded, its size depends entirely on what is adjusted for, and the arithmetic of that adjustment is worth making explicit; the first demonstration reproduces it. Begin with a raw difference of 8.0 IQ points between breastfed and bottle-fed children — a deliberately large starting gap in the range reported by unadjusted studies. Suppose that, in this illustrative sample, three confounders each account for a fixed fraction of that original gap: maternal intelligence explains 45 percent of it, socioeconomic status a further 20 percent, and the quality of the home environment a further 15 percent.

Adjusting for a confounder removes its share of the original gap. Adjusting for maternal intelligence removes 45 percent of 8.0, or 3.6 points, leaving 4.4. Adjusting additionally for socioeconomic status removes 20 percent of 8.0, or 1.6 points, leaving 2.8. Adjusting finally for the home environment removes 15 percent of 8.0, or 1.2 points, leaving a residual difference of 1.6 points. The three confounders together explain 80 percent of the raw gap, and the fully adjusted estimate — the part that might reflect feeding itself — is one-fifth of what the unadjusted comparison showed.

Two things follow from the arithmetic. First, the order of adjustment does not change the final residual but it does change the story a partial analysis tells: a study that adjusts only for socioeconomic status would report a 6.4-point effect and conclude feeding matters greatly, while one that adjusts for maternal intelligence first would report 4.4 points and already doubt it. This is why the single most important confounder — maternal intelligence — is the one the sibling and quasi-experimental designs were built to control (Der et al., 2006). Second, the residual is not zero: in this illustration 1.6 points survive every adjustment, which is the quantitative form of the conclusion that a small genuine effect may remain even after the confounding is stripped away (Kramer et al., 2008). The demonstration lets the raw gap and the three fractions vary, making visible how a conclusion about feeding depends less on the data than on the model of confounding brought to it. The model is a deliberate simplification — real confounders overlap rather than partition cleanly — but it isolates the central point: an adjusted effect is a claim about what has been controlled, not a fact read directly from the children.

Discussion

Bottle feeding occupies an unusual place in cognitive psychology: it is not a cognitive construct at all, but a nutritional method that became a vehicle for three distinct cognitive-developmental questions. The first — whether feeding mode affects intelligence — turned out to be, as much as anything, a lesson in causal inference, and the careful answer is that the large raw advantage of breastfeeding is mostly confounded, while a small effect may survive the best designs built to test it (Der et al., 2006; Kramer et al., 2008). The second — how an infant learns to regulate intake — identified the bottle as a device that makes overriding the infant's fullness cues easy, with measurable consequences for weight gain (Li et al., 2010; DiSantis et al., 2011). The third — how early preferences form — located bottle feeding within the exposure learning that tunes a child's acceptance of flavors (Ventura & Worobey, 2013).

What unifies the three is a shift of the explanatory weight from the vessel to the interaction. The composition of formula may contribute a small amount to the cognition difference, but most of that difference is about mothers, not milk; the control a bottle affords matters because of the feeding behavior it invites, not because of the plastic; and the flavor history a feeding mode delivers is one input to a general learning process. In each case the modern synthesis replaces a crude contrast of breast and bottle with a finer variable — responsive versus nonresponsive feeding — that cuts across the method (Black & Aboud, 2011; Pérez-Escamilla et al., 2017). That reframing is the field's most useful product: it redirects attention from a choice many caregivers cannot freely make to a behavior any caregiver can adopt.

Cognitive Implications

The study of bottle feeding touches several core topics in cognitive and developmental psychology. The self-regulation of intake is an early, concrete instance of the broader capacity for self-regulation, and the finding that bottle feeding can shift an infant from internal to external control of eating is a developmental-origins account of how appetite comes to be governed by cues other than hunger and satiation. The formation of food preferences through repeated exposure is an application of associative learning to the sensory world of infancy, and it connects the feeding literature to the psychology of taste perception. Responsive feeding, in turn, is a special case of the contingent, signal-matched caregiving that developmental psychology treats as the engine of healthy development, and it shares its logic with the responsive interaction that supports secure object attachment. Finally, the breastfeeding-and-cognition debate is a case study in the measurement and interpretation of intelligence under confounding — a reminder that a difference in test scores is only as meaningful as the design that produced it.

Current Directions

The most active recent work has moved decisively from the question of what is in the bottle to the question of how the bottle is used. The responsive-feeding framework, consolidated in practical guidelines, has reframed the policy conversation around the caregiver-infant interaction rather than the choice of feeding mode, and the open empirical question is whether interventions that teach responsive bottle feeding measurably improve self-regulation and slow excess weight gain (Pérez-Escamilla et al., 2017). A second front extends the self-regulation findings into the mechanisms of later obesity risk, asking how early feeding experience sets intake patterns that persist (Ventura, 2017; Appleton et al., 2018). A third keeps refining the causal estimate in the cognition debate with ever-larger linked datasets and quasi-experimental methods, narrowing the band within which any true effect of feeding mode on cognition must lie (Victora et al., 2016). Across all three the through-line is methodological maturity: the field increasingly treats bottle feeding not as a crude binary but as a bundle of separable variables, each demanding its own design.

Common Misconceptions

Bottle feeding means formula feeding.

Not necessarily. Bottle feeding names the method of delivery; the bottle may contain commercial formula or expressed human milk, so an infant fed expressed breast milk from a bottle is being bottle-fed (Li et al., 2010).

Breastfeeding has been proven to make children more intelligent.

The raw association is real but largely confounded by maternal intelligence and family circumstances. Sibling and quasi-experimental designs shrink the apparent effect substantially, leaving at most a small residual (Der et al., 2006; Kramer et al., 2008).

A bottle-fed baby is bound to overeat.

The bottle makes overriding fullness cues easy, but it does not compel it. Responsive bottle feeding that respects the infant's satiety signals preserves self-regulation; the risk lies in the practice of feeding to empty the bottle, not in the bottle itself (Black & Aboud, 2011).

The feeding method is what determines a child's development.

The modern synthesis points instead to the feeding interaction. Responsive versus nonresponsive feeding — reading and respecting the infant's signals — cuts across breast and bottle and is the developmentally active variable (Pérez-Escamilla et al., 2017).

Glossary

Bottle feeding.

The feeding of an infant with milk — formula or expressed human milk — presented in a bottle fitted with a teat, as distinct from feeding directly at the breast.

Confounding.

The distortion of an association between two variables by a third that influences both; here, maternal intelligence and family circumstances inflating the apparent effect of feeding mode on cognition.

Exclusive breastfeeding.

Feeding an infant only human milk directly at the breast, with no formula or other food, the comparison condition against which bottle feeding is usually studied.

Expressed human milk.

Human milk removed from the breast by hand or pump and stored for later feeding, which may itself be given by bottle, decoupling the type of milk from the method of delivery.

External cue.

A signal outside the body — such as the amount of milk left in a bottle — used to decide when to stop eating, in contrast to an internal signal of fullness.

Feeding mode.

Whether an infant is fed at the breast or from a bottle; the predictor variable in the cognition and self-regulation literatures.

Infant formula.

A manufactured substitute for human milk designed to meet an infant's nutritional needs, presenting a constant composition and flavor rather than the varying flavors of human milk.

Mere-exposure learning.

The increase in acceptance of a flavor or food that follows repeated exposure to it, the associative process underlying the early formation of food preferences.

PROBIT trial.

The Promotion of Breastfeeding Intervention Trial, a cluster-randomized trial in Belarus that randomized a breastfeeding-promotion intervention across hospitals, providing the strongest quasi-experimental test of feeding mode and cognition.

Rapid weight gain.

Faster-than-typical gain in weight during infancy, associated with less responsive feeding and formula-feeding practices and predictive of later overweight.

Responsive feeding.

Feeding that is prompt, contingent, and matched to the infant's hunger and satiety signals; the developmentally active ingredient that cuts across breast and bottle.

Self-regulation of intake.

The infant's capacity to take the amount of milk it needs and stop when sated, relying on internal hunger and fullness signals rather than external cues.

Sibling-pairs design.

A study design comparing siblings within a family who were fed differently, holding constant the family characteristics they share to reduce confounding.

Socioeconomic status.

A family's standing in income, education, and occupation, a confounder of the feeding-and-cognition association because it predicts both the choice to breastfeed and a child's test scores.

Key Researchers

Maureen M. Black

. University of Maryland School of Medicine (Department of Pediatrics, Professor Emeritus) and RTI International; embedded responsive feeding within a broader theory of responsive parenting, clarifying that the caregiver-infant feeding interaction, not the vessel alone, is the developmentally active ingredient. ORCID - Faculty Page - Google Scholar - Wikidata

Ian J. Deary

. University of Edinburgh (School of Philosophy, Psychology and Language Sciences; Professor of Differential Psychology); brought the methods of differential psychology and the sibling-pairs design to the breastfeeding-and-intelligence question, showing how much of the raw feeding-mode advantage is explained by maternal and family confounding. ORCID - Faculty Page - Google Scholar - Wikipedia

Michael S. Kramer

. McGill University (Department of Epidemiology, Biostatistics and Occupational Health; Department of Pediatrics); led the PROBIT cluster-randomized breastfeeding-promotion trial, the strongest causal design in the infant-feeding literature, isolating the effect of feeding mode on child cognition from the confounds that cripple observational studies. Faculty Page - Google Scholar

Alison K. Ventura

. California Polytechnic State University (Cal Poly), San Luis Obispo (Department of Kinesiology and Public Health; Director, Healthy Kids Lab); established the experimental study of bottle-feeding behavior itself — how bottle properties, milk type, and caregiver responsiveness shape infant intake regulation and the early formation of food preferences. ORCID - Faculty Page - Google Scholar

Cesar G. Victora

. Federal University of Pelotas, Brazil (Postgraduate Program in Epidemiology; International Center for Equity in Health; Emeritus Professor); led the 2016 Lancet Breastfeeding Series and the meta-analytic synthesis of breastfeeding's long-term effects, setting the global evidentiary baseline against which bottle-feeding outcomes are weighed. ORCID - Faculty Page - Google Scholar - Wikipedia

Frequently Asked Questions

What is bottle feeding?

Bottle feeding is the delivery of milk to an infant from a bottle fitted with a teat, rather than directly from the breast. The milk may be commercial infant formula or expressed human milk, so the term describes a method of feeding rather than a type of milk (Li, Fein, & Grummer-Strawn, 2010).

Does bottle feeding lower a child's intelligence?

The evidence does not support a large effect. Breastfed children score a few points higher on average, but most of that gap is explained by maternal intelligence and family circumstances; designs built to remove that confounding shrink the apparent effect to a small residual at most (Der, Batty, & Deary, 2006; Kramer et al., 2008).

Why is the breastfeeding-and-cognition link so hard to establish?

Because feeding mode is not assigned at random. Mothers who breastfeed differ systematically from those who bottle-feed, notably in measured intelligence, so the raw association is confounded, and only sibling or quasi-experimental designs can separate the effect of feeding from the effect of the families (Der et al., 2006).

Does bottle feeding affect how much a baby eats?

It can. Because a bottle's contents are visible and measurable, caregivers tend to feed to empty the bottle rather than to the infant's fullness cue, and infants fed this way show weaker self-regulation of intake, which is associated with more rapid weight gain (Li et al., 2010; DiSantis et al., 2011).

What is responsive feeding?

Responsive feeding is feeding that is prompt, contingent, and matched to the infant's hunger and satiety signals. The recent literature identifies it, rather than the choice of breast or bottle, as the developmentally active ingredient in infant feeding (Black & Aboud, 2011; Pérez-Escamilla, Segura-Pérez, & Lott, 2017).

How does early feeding shape later food preferences?

Through exposure learning. Repeated exposure to a flavor increases its acceptance, and because human milk carries the varying flavors of the maternal diet while formula is constant, breastfed and formula-fed infants reach the introduction of solid foods with different flavor histories (Ventura & Worobey, 2013).

Is bottle feeding a cognitive process?

No; it is a nutritional method. It belongs in cognitive psychology because it became a vehicle for studying causal inference under confounding, the self-regulation of intake, and the early learning of food preferences, each of which is a genuine cognitive-developmental question (Ventura, 2017).

Can a bottle-fed baby still self-regulate intake?

Yes. Self-regulation is preserved when feeding respects the infant's satiety cues; it is undermined by the practice of feeding to empty the bottle. Responsive bottle feeding is designed precisely to keep the infant's own signals in control of the feed (Appleton et al., 2018; Pérez-Escamilla et al., 2017).

References