Infant Behavior

Abstract

Infant behavior is a form of child behavior: the observable responses and actions of a child from birth through the age of 23 months, the earliest developmental band of the broader category MeSH scopes to childhood as a whole. Cognitive psychology treats it as the measurable surface of a mind under rapid construction, the only evidence a preverbal participant can give. This article sets out what infant behavior is and why its measurement was a methodological turning point; how preferential looking revealed an organized visual world from the first days of life; what violation-of-expectation studies show infants know about objects, and the core-knowledge theory built on them; how infants segment speech by tracking statistical regularities, worked through its arithmetic; the attachment and temperament systems that shape early behavior; and the predictive-processing account now reframing what looking time measures.

Keywords: infant behavior, preferential looking, core knowledge

Infant behavior denotes any observable response or action of a neonate or infant up through 23 months of age, and for most of the history of psychology it was treated as the behavior of a largely blank and incompetent organism. The modern study of the infant overturned that view, and it did so by solving a measurement problem: an infant cannot report, follow an instruction, or give a deliberate response, so the field's central achievement has been to turn the few things an infant does do — look, suck, turn the head, reach, habituate — into rigorous measures of what the infant perceives, remembers, and expects (Fantz, 1963). The study of infant behavior matters to cognitive psychology because it is where the origins of cognition become visible: every adult capacity for perceiving, remembering, and knowing has a starting state, and infant behavior is the window through which that starting state can be observed before language intervenes.

What Infant Behavior Is

Infant behavior is any observable response or action of a neonate or infant through the age of 23 months. MeSH places it as a specific kind within its parent descriptor child behavior, marking off the earliest developmental band — the period before symbolic language and before much deliberate self-regulation is in place — from the wider span of childhood conduct. The boundary is not arbitrary: the infant is the one human research participant who can neither be instructed nor debriefed, so the behavior available for study is restricted to what the infant spontaneously does, and the whole science rests on making those spontaneous acts carry information.

That constraint, long treated as a reason to think infants unknowable, turned out to be the discipline's making. The decisive move was to recognize that an infant who reliably does one thing more than another — looks longer, sucks harder, turns toward one sound rather than another — has thereby discriminated between the alternatives, and discrimination is the raw material of perception and cognition. A demonstration that newborns look longer at patterned than at plain surfaces established that even the youngest infant brings a visually organized response to the world, and, just as importantly, that looking time is a usable dependent measure (Fantz, 1963). From that point the question stopped being whether infant behavior could be studied rigorously and became what each measurable behavior could be made to reveal.

Figure 1

Observable Infant Behaviors as Measures of an Unobservable Mind

Table 1

Landmark Behavioral Methods in the Study of Infant Cognition

How Infants Perceive

The first behavior pressed into service was looking, and the method built on it is preferential looking: present two stimuli side by side and measure how long the infant looks at each. Its founding result was that newborns, far from seeing an undifferentiated blur, look reliably longer at patterned targets — stripes, bullseyes, a schematic face — than at plain fields of the same size and brightness, which means the visual system arrives already biased toward structure and already able to resolve it (Fantz, 1963). The logic is deceptively simple and completely general: if looking time differs between two stimuli, the infant must be able to tell them apart, so any perceptual discrimination can be tested by finding two displays the infant treats differently.

Looking time also supports a second, dynamic technique that reaches beyond perception into memory. When a stimulus is shown repeatedly, looking declines — the infant habituates — and when a new stimulus is then introduced, looking recovers if and only if the infant perceives it as different from the now-familiar one. This habituation–dishabituation logic turns the infant's own loss of interest into an index of recognition, and it is the workhorse behind most of what is known about early perceptual categories. That infants remember at all, and for longer than once assumed, was established with a different behavior entirely: in the mobile-conjugate-reinforcement paradigm, infants learn that kicking a foot tethered to an overhead mobile makes it move, and they retain that learned association over days and weeks, demonstrating a genuine, durable memory well before the first birthday (Rovee-Collier, 1999). Perception and memory, the two oldest problems in the study of mind, both yielded to the strategy of reading them off what the infant spontaneously does.

What Infants Know

If an infant looks longer at the unexpected, looking time can measure not just what the infant perceives but what the infant expects — and expectation is a window onto knowledge. The violation-of-expectation method exploits exactly this. In its founding study, five-month-old infants watched a screen rotate like a drawbridge; when a solid box was placed behind the screen, infants looked longer at an impossible event, in which the screen rotated through the space the box should still have occupied, than at a possible event in which it stopped against the hidden box (Baillargeon et al., 1985). The longer looking at the impossible outcome is evidence that the infants represented the box as continuing to exist when hidden and expected it to obstruct the screen — object permanence, demonstrated far earlier than the classical account had allowed, and demonstrated purely through looking. The classical account is Piaget's: on his stage theory the infant constructs object permanence only gradually across the first year and does not reliably search for a fully hidden object until roughly eight or nine months of age (Piaget, 1954). Violation of expectation pulled that milestone years earlier precisely because it asks whether the infant is surprised rather than whether the infant can act, separating what an infant knows from what an infant can yet do.

Results like these accumulated into a theoretical claim about the starting state of the mind. Core-knowledge theory holds that human cognition is built on a small set of early-developing, domain-specific systems of knowledge — of objects and their mechanics, of number, of space, of agents — that are present in some form in infancy and that scaffold later learning rather than being wholly derived from it (Spelke et al., 1992). On this view the young infant is not a blank slate acquiring everything from sensory association but an organism equipped with foundational expectations: that objects are cohesive and continuous, that they move on connected paths, that they do not pass through one another. The contemporary statement of the theory assembles decades of infant behavioral evidence into the argument that these core systems are the composable building blocks from which more complex human thought is constructed (Spelke, 2022). The debate the theory provoked — how much is given versus how much is rapidly learned — remains central to developmental cognitive science, but the behavioral phenomena it rests on are robust and have been replicated across laboratories and paradigms.

Worked Example

One of the most striking demonstrations of infant learning concerns how a baby finds the words in continuous speech, where nothing in the acoustic signal reliably marks where one word ends and the next begins. The proposed solution is statistical: within a word, the syllables follow one another predictably, whereas across a word boundary the next syllable is far less predictable, so a learner who tracks the transitional probability between adjacent syllables can place boundaries at the dips. Eight-month-old infants were shown to do exactly this after only two minutes of exposure to a continuous, prosodically flat stream of nonsense syllables (Saffran et al., 1996).

Make the arithmetic explicit. The transitional probability from syllable A to syllable B is the conditional probability P(B | A) — the number of times B follows A divided by the total number of times A occurs. Consider a familiarization language of four three-syllable nonsense words — tupiro, golabu, bidaku, padoti — concatenated at random with no pauses, so the stream runs ...tupirogolabubidakupadotibidakutupiro... with no acoustic seam. Inside a word, each syllable is always followed by the same next syllable: tu is invariably followed by pi, so the transitional probability P(pi | tu) = 1.0, and the same holds for every within-word pair. At a word boundary the situation is different: ro, the last syllable of tupiro, can be followed by the first syllable of any of the other three words, each about equally often, so the transitional probability across the boundary is roughly P(next | ro) = 1/3 ≈ 0.33. The contrast the infant exploits is therefore 1.0 within a word against about 0.33 across a boundary — a threefold drop. Tracking adjacent-syllable statistics and segmenting at the low points recovers the four words from an unbroken stream, and after two minutes infants listen longer to the novel non-words that straddle a boundary than to the statistically coherent words, showing they extracted precisely this structure (Saffran et al., 1996). The computation is well within the reach of an eight-month-old, and the finding launched the study of infant statistical learning as a general mechanism of early cognition (Saffran & Kirkham, 2018).

Discussion

Perception, memory, and knowledge do not exhaust infant behavior; its most consequential dimension is social, and the framework that organizes it is attachment theory. Attachment theory holds that the infant is predisposed to form an enduring emotional bond with a primary caregiver, and that the behavioral system serving it — proximity-seeking, distress at separation, the use of the caregiver as a secure base from which to explore — is a product of evolution with clear survival value (Bowlby, 1969). The theory moved infant social behavior from the margins to the center of developmental science by treating the baby's cries, clinging, and following not as immature noise but as an organized, functional system.

Attachment became an experimental science when its quality could be measured behaviorally. The Strange Situation procedure observes an infant's behavior through a scripted sequence of separations from and reunions with the caregiver, and the infant's reunion behavior in particular sorts into recognizable patterns — secure, and several insecure variants — that index the history of the caregiving relationship (Ainsworth & Bell, 1970). It is the direct social-domain analogue of preferential looking: a standardized situation in which spontaneous infant behavior is read as evidence of an underlying state, here the security of a relationship rather than a perceptual discrimination.

Infants also differ from one another from the first weeks of life in the style of their behavior, and the construct that names those differences is temperament: early-appearing, constitutionally based individual differences in activity, emotional reactivity, and regulation, which the classic longitudinal work grouped into recognizable profiles and tied to later adjustment through the goodness of fit between the child's temperament and the environment's demands (Thomas & Chess, 1977). How far these early differences, and early behavior generally, forecast the later person is a question of continuity. The careful answer is that continuity is real but probabilistic: early behavior constrains later behavior without fixing it, because development is an open system in which later experience can redirect a trajectory (Bornstein et al., 2017). Infant behavior is thus predictive enough to be worth measuring and revisable enough that no early measurement is a verdict.

Cognitive Implications

Infant behavior belongs on a cognitive-psychology site because it is where cognition can be caught in its starting state, before instruction, language, or culture have done their work. Every mechanism the adult mind runs — the selection of information by attention, the brief maintenance of it in working memory, the extraction of regularities by learning, the parsing of a signal into the units of language — has an initial form that infant behavior lets us observe directly. The preferential-looking and habituation methods are, in effect, psychophysics run on a participant who cannot be briefed, and they have shown that the infant's perception is organized and its visual perception acute far earlier than the blank-slate tradition supposed.

The deeper implication is methodological and evidential. Because the infant cannot report, every claim about infant cognition is an inference from behavior, and the discipline's rigor comes from holding those inferences to behavioral evidence — a differential looking time, a recovered memory, a segmented stream. This makes the study of infant behavior a continuous argument about what a given behavior licenses one to conclude: whether longer looking at an impossible event really indexes a represented object, or whether statistical learning of syllables is the same mechanism that later builds a lexicon. Those arguments are the cognitive psychology of development, and they are conducted in the currency of observable behavior precisely because that is the only currency a preverbal mind can offer.

Current Directions

The most active reframing of infant behavior comes from predictive processing, which recasts the infant not as a passive recorder of stimulation but as an active prediction machine that continuously generates expectations about incoming sensory input and learns from the mismatch — the prediction error — when input departs from expectation (Köster et al., 2020). The account unifies phenomena that had been studied separately: habituation becomes the decline of prediction error as a stimulus is predicted ever better, and the violation-of-expectation response becomes a large prediction error provoked by an outcome the infant's model did not forecast. On this view the longer looking at an impossible event is not merely a tell that the infant knew something but a signature of a brain updating its generative model, which ties the infant behavioral literature to the same computational framework now applied across adult perception and learning.

Statistical learning, too, has grown from a single demonstration into a broad research program. The mechanism first shown for speech segmentation turns out to operate over visual sequences, tone sequences, and action streams, and across species, raising the still-open question of how domain-general the capacity really is and how it interacts with the domain-specific core systems that core-knowledge theory posits (Saffran & Kirkham, 2018). The contemporary synthesis of core knowledge sharpens the complementary question from the other side: which expectations are genuinely foundational and composable, and which are themselves products of early statistical learning (Spelke, 2022). The relationship between the infant's early-given structure and its powerful early learning — between what is built in and what is rapidly built up — is the central unsettled problem the next generation of infant behavioral work is organized around.

Common Misconceptions

Newborns see only a meaningless blur.

Preferential-looking studies show that even newborns look longer at patterned than at plain surfaces, which means the visual system arrives already organized and able to resolve structure, not confronting an undifferentiated field (Fantz, 1963).

Infants have no concept of a hidden object until they can search for it.

Violation-of-expectation studies show that five-month-olds already look longer at events that violate object permanence, indicating they represent a hidden object even before they have the motor skill to reach for it (Baillargeon et al., 1985).

A baby must be explicitly taught where words begin and end.

Eight-month-olds segment a continuous speech stream after two minutes of exposure by tracking the transitional probabilities between syllables, an unsupervised statistical computation that needs no feedback or instruction (Saffran et al., 1996).

Glossary

Attachment.

The enduring emotional bond an infant forms with a primary caregiver, served by an organized behavioral system of proximity-seeking, separation distress, and secure-base exploration.

Attention.

The cognitive selection of some stimuli for processing over others; its early form is probed in infants by what and how long they choose to look at.

Child Behavior.

The observable activity of a child; the parent MeSH category of which infant behavior is the earliest developmental band.

Core Knowledge.

A set of early-developing, domain-specific systems — of objects, number, space, and agents — proposed to be present in infancy and to scaffold later learning.

Habituation.

The decline in looking or responding to a repeated stimulus; its recovery to a new stimulus (dishabituation) indexes that the infant perceives the new one as different.

Infant Behavior.

Any observable response or action of a neonate or infant through the age of 23 months, studied as the measurable evidence of a developing mind.

Learning.

A change in behavior or knowledge produced by experience; in infancy it is demonstrated by durable conditioned associations and by the extraction of statistical structure.

Mobile Conjugate Reinforcement.

A paradigm in which an infant learns that kicking moves a tethered overhead mobile, used to show that infants form durable memories retained over days and weeks.

Object Permanence.

The understanding that an object continues to exist when it cannot be seen; infants show it through longer looking at events in which a hidden object fails to obstruct a moving screen.

Perception.

The organization and interpretation of sensory input; in infants it is inferred from differential looking, sucking, and head-turning responses.

Predictive Processing.

The framework that casts the brain as continuously predicting sensory input and learning from prediction error; it reinterprets habituation and violation-of-expectation as model updating.

Preferential Looking.

A method presenting two stimuli side by side and measuring relative looking time; longer looking at one implies the infant can discriminate it from the other.

Statistical Learning.

The unsupervised extraction of regularities from input, such as segmenting speech by tracking the transitional probabilities between adjacent syllables.

Strange Situation.

A scripted sequence of caregiver separations and reunions whose observed infant behavior classifies the security of the attachment relationship.

Temperament.

Early-appearing, constitutionally based individual differences in activity, emotional reactivity, and regulation, grouped into profiles and tied to later adjustment through goodness of fit.

Transitional Probability.

The conditional probability that one element follows another; high within a word and low across a word boundary, it is the statistic infants use to segment speech.

Violation of Expectation.

A method inferring infant knowledge from longer looking at an impossible or unexpected event than at a matched possible one.

Key Researchers

Mary D. Salter Ainsworth

(1913-1999). Developmental psychologist at the University of Virginia; she created the Strange Situation procedure that made the security of infant attachment measurable through observed reunion behavior.

Wikidata

Renee Baillargeon

(b. 1954). Alumni Distinguished Professor Emerita of Psychology at the University of Illinois at Urbana-Champaign; her violation-of-expectation studies demonstrated object permanence in five-month-old infants.

ORCID

John Bowlby

(1907-1990). Psychiatrist and psychoanalyst at the Tavistock Clinic in London; he originated attachment theory, recasting the infant's bond to the caregiver as an evolved behavioral system with survival value.

Wikidata

Robert L. Fantz

(1925-1981). Psychologist at Western Reserve University; he devised the preferential-looking method and showed that newborns look longer at patterned than at plain surfaces, founding the experimental study of infant perception.

Wikipedia

Jenny R. Saffran

. Vilas Distinguished Achievement Professor of Psychology at the University of Wisconsin–Madison; she discovered infant statistical learning, showing that eight-month-olds segment speech by tracking transitional probabilities.

Google Scholar

Elizabeth S. Spelke

(b. 1949). Marshall L. Berkman Professor of Psychology at Harvard University; she developed core-knowledge theory, the account of infant cognition as built on early, domain-specific systems of knowledge.

ORCID

Frequently Asked Questions

What is infant behavior?

Infant behavior is any observable response or action of a neonate or infant through the age of 23 months, which MeSH places as the earliest developmental band of child behavior and which cognitive psychology studies as the measurable evidence of a developing mind (Fantz, 1963).

How do researchers study the mind of a baby who cannot talk?

By turning the few things an infant spontaneously does (look, suck, turn the head, reach, habituate) into measures: an infant who reliably looks longer at one of two stimuli has discriminated between them, so looking time becomes a report (Fantz, 1963).

What is the preferential-looking method?

It presents two stimuli side by side and measures how long the infant looks at each; longer looking at one implies the infant can tell it from the other, which first showed that newborns prefer patterned over plain surfaces (Fantz, 1963).

Do infants understand that hidden objects still exist?

Yes, earlier than once thought: five-month-olds look longer at an impossible event in which a screen rotates through where a hidden box should be, indicating they represent the box as continuing to exist (Baillargeon et al., 1985).

What is core-knowledge theory?

It is the view that human cognition is built on a small set of early-developing, domain-specific systems (of objects, number, space, and agents) present in infancy and scaffolding later learning rather than wholly derived from it (Spelke et al., 1992).

How do babies find the words in continuous speech?

They track the transitional probability between adjacent syllables, which is high within a word and low across a word boundary, and segment at the dips; eight-month-olds do this after only two minutes of exposure (Saffran et al., 1996).

What is attachment, and how is it measured?

Attachment is the infant's enduring emotional bond to a caregiver, an evolved behavioral system; its quality is measured by the Strange Situation, which classifies the infant's behavior across scripted separations and reunions (Ainsworth & Bell, 1970).

Does early infant behavior predict the later child?

At the group level it does, but the continuity is probabilistic: early behavior constrains later behavior without determining it, because later experience can redirect a developmental trajectory (Bornstein et al., 2017).

References