Infant-Directed Speech as Biological Communication Signalling During Attachment Formation

1. Infant-Directed Speech as Biological Communication Signalling During Attachment Formation

Across cultures and languages around the world (Hilton et al., 2022), caregivers addressing infants tend to adopt a distinctive vocal register known as the Infant-Directed Register [IDReg] (Ferguson, 1977). Within the IDReg, speech characterized by exaggerated prosody, elevated pitch, slower tempo, and heightened affective expressiveness is known as infant-directed speech [IDS], also described in the literature as "motherese" (Fernald, 1985), "babytalk" (Singh et al., 2002), or "songese" (Longhi, 2009). Decades of research indicate that IDS offers a rich multimodal signal combining acoustic clarity (Kuhl et al., 1997; Liu et al., 2003; Song et al., 2010), affective warmth (Saito et al., 2007; Santesso et al., 2007), and neurocognitive salience (Peter et al., 2016), and far from being merely simplified speech, can be understood as a finely tuned communicative tool that scaffolds multiple aspects of early development, including linguistic, attentional, emotional, and social (Robledo, 2020).

Because IDS typically occurs in the context of caregiving and parenting, a number of authors have suggested connections between these vocal behaviours and attachment-related processes (de L'Etoile, 2006; Smith & Trainor, 2008; Swain et al., 2007; Trehub et al., 1997). Yet such connections have not been posited in detail, nor theoretically scrutinized. A consistent pattern emerges across this literature: the term "attachment" functions as a descriptive placeholder, invoked to explain observed behaviours rather than as a systematically operationalized construct. Without precisely defining and testing the connective tissue between a specific vocal phenomenon and a specific attachment dynamic, the two phenomena remain ultimately isolated from one another. The common risk of these shortcomings is nominalism, treating attachment as a label or a given byproduct rather than an active, operationalized variable.

A critical source of this ambiguity is the failure to distinguish between two independent phenomena that the attachment literature has historically not appraised in a balanced manner: attachment formation and attachment quality. Attachment formation refers to the process by which a specific individual becomes non-interchangeable, the gated selection process through which a bond is built in the first place. Attachment quality refers to what an established bond does, its security, regulatory function, and the internal working models that organize behaviour toward the attachment figure over time. These are not merely different aspects of the same process. They have different determinants, different timescales, and different empirical operationalizations (Freeman, 2025; 2026). Behaviours that promote attention, proximity, positive affect, or social engagement may contribute to affiliative processes without necessarily participating in the formation or maintenance of an actual attachment bond as defined by Bowlby (1969). Consequently, references to IDS as an attachment device remain conceptually ambiguous unless the specific attachment process under consideration is clearly specified.

This distinction has direct implications for understanding IDS. Some of the previously mentioned studies that have positioned IDS as an attachment-related device have focused on attachment quality, examining how vocal interaction contributes to secure attachment, affect regulation, or the development of internal working models (e.g., de L'Etoile, 2006; Swain et al., 2007). In doing so they have missed the more fundamental question: before clear-cut attachment (of whatever quality) is achieved, what role does IDS play in attachment formation itself, in the process by which a specific caregiver becomes selected and irreplaceable to a specific infant? Addressing that question delimits the developmental window of interest to the first year of life and largely the first six months. This developmental window is critical in two ways. It is precisely the period during which IDS is most acoustically exaggerated and perceptually salient (Cooper & Aslin, 1990; Fernald, 1985; Kalashnikova et al., 2018; Soderstrom, 2007), and the window at the end of which an orientation toward the caregiver's signals gives way to the consolidation of a clear-cut, individualized attachment bond. This temporal convergence represents a first central aim of this paper.

Positioning IDS within an attachment formation framework, rather than an attachment quality framework, also requires that we situate it within a broader communicative and evolutionary account of what animal signals do and how they work. Bowlby drew attachment theory mainly from ethology, and the communicative apparatus he assigned to infants, including crying, smiling, and gesturing, was modeled on Lorenz’s releaser and fixed action pattern logic of imprinting theory (Robledo et al., 2022). The psychological community interested in attachment quality has, over time, largely set aside this ethological grounding. Regaining it, however, proves consequential. From an ethological standpoint, the function of signalling is to influence the behaviour of receivers through evolved mechanisms rather than to transmit semantic content (Owren & Rendall, 2001; Smith & Harper, 2003). When this framework is applied to the full repertoire of signals operating in the human parent-infant dyad, including not only the IDReg but also the infant's own signals directed at adult caregivers, a picture emerges that is considerably more complex than the standard account allows. The IDReg and what we hereby designate as the Caregiver-Directed register (CDReg) are not simply two directions of a single bonding mechanism. They are two independent evolved signal systems, each playing a key phase-specific formation role in the development of infant attachment and adult attachment respectively. Specifying what each system does, for whom (parent or infant), and when, is the second central aim of this paper.

The overall aim of the present paper is therefore to provide a systematic examination of IDS as an attachment formation device rather than an attachment quality device. To this end, the paper first introduces the distinction between attachment formation and attachment quality and presents the phase-based sensitive period framework that specifies when formation occurs and what it depends on (Freeman, 2026). It then reviews essential aspects of animal communication in ethological literature, with particular emphasis on influence-based and evolutionary accounts of signalling, equipping the reader with the vocabulary needed to evaluate IDS and infant signals as components of evolved biological communication. Building on both frameworks, the paper examines the specific acoustic, affective, and interactional characteristics of the IDS that allow it to participate in attachment formation dynamics, before developing the argument that the CDReg operates as a complementary system serving a distinct formation process within the same developmental window. A closing section summarizes the main implications, addresses theoretical gaps identified in the introduction, and acknowledges the paper's limitations.

2. Attachment Formation and Its Phases

Bowlby's original account of attachment theory began with an ethological description of bond formation. Drawing directly on Lorenz's concept of imprinting (van der Horst, 2009), Bowlby carried over its core architecture almost wholesale: the signalling equipment he assigned to infants, including crying, gazing, smiling, and gesturing, mirrors imprinting's releaser and fixed action pattern logic, and his substitution of sensitive periods for Lorenz's critical periods echoes imprinting's temporal scaffolding point for point (Robledo et al., 2022). Both theories describe a process in which species-typical signals orient a naive organism toward eligible social targets, and both treat the resulting bond as having lasting consequences for social behaviour across the lifespan.

The two theories diverge, however, in ways that have not been fully appreciated, and that divergence has consequences for how we understand the role of signalling in bond formation. The first and most commonly noted divergence concerns the level of specificity at which bonding occurs. In Lorenz's imprinting, the organism learns to recognize the features of its own kind at a supraindividual level, acquiring a general orientation toward a class of social objects (its species) rather than a bond to a specific individual. Bowlby's attachment, by contrast, requires proximity-seeking to become preferentially directed at one specifically recognized, non-interchangeable individual, organized into a goal-corrected behavioural system in which the infant monitors feedback and adjusts behaviour to maintain proximity to that specific figure (Bowlby, 1969; Ainsworth et al., 1978; Robledo et al., 2022). The transition from supraindividual orientation to individual-specific consolidation is not merely a theoretical distinction. It maps onto a developmental timetable: broad social permissiveness characterizes the earliest months of life, and individual-level selective attachment consolidates in most human infants between five and eight months of age (Ainsworth, 1967; Freeman, 2025; Schaffer & Emerson, 1964). This temporal structure, as later sections will show, is precisely the architecture within which the signalling systems operating in the parent-infant dyad do their formation work.

The second and more consequential divergence concerns what signalling is understood to be doing in the formation process. In the ethological literature, offspring signals are largely understood as having evolved in the context of parent-offspring conflict (Trivers, 1974), a competitive dynamic where parental investment is limited and not guaranteed, and can range from full provision to abandonment or infanticide (Soltis, 2004). The infant is not signalling to access quality care in a mutually cooperative exchange. The infant is signalling to maximize its own share of parental investment in competition with other claimants, using an evolved repertoire of “honest”, costly signals (Zahavi, 1975) whose intensity and reliability reflect genuine viability as a target for investment (Soltis, 2004). The parent, in turn, is not simply a responsive partner in a mutual bond-building process but a decision-maker managing the allocation of limited resources, assessing the honesty and intensity of signals to calibrate investment accordingly. Simultaneously, the parent directs its own evolved signalling repertoire toward the infant, biasing the infant's emerging social orientation toward itself as a specific target. Each organism has therefore evolved a signalling inventory directed at the other, not to create a cooperative mutual exchange, but to bias the other's attachment formation toward itself. These are parallel co-evolved signal systems serving convergent fitness interests, each independently driving the other organism's formation process in a direction that also benefits the signaller.

In Bowlby's account, and more fully in Ainsworth's elaboration of it, the formation process looks quite different. Infant signals communicate need states to caregivers, who respond with sensitive and contingent care, and the quality of that care determines who the infant eventually selects as a preferential attachment figure (Ainsworth et al., 1978; Bowlby, 1969). The causal chain runs from infant signal to caregiver response to interaction quality to bond. The infant is selecting a caregiver on the basis of what that caregiver provides. To explain the mechanisms through which this selectivity develops and is maintained, Bowlby and Ainsworth drew on object relations theory, psychoanalytic accounts of internalized representations, and control systems theory, generating the constructs of internal working models, attachment security, and goal-corrected partnership that have organized the field ever since (Ainsworth et al., 1978; Bowlby, 1969). These constructs describe how an established bond functions, how it regulates affect, organizes behaviour toward the attachment figure, and shapes expectations of future relationships. They are, in short, mechanisms of attachment quality and post-formation function. They have no direct parallel in the animal imprinting and bonding literature, which does not require internal representational systems to explain bond formation, and which has never treated the quality of what the attachment figure provides as the primary determinant of who is selected (Freeman, 2025).

Freeman (2026) reconceptualized attachment formation within a sensitive period framework, grounding bond formation in the same gated plasticity logic that governs sensitive periods across species, from filial imprinting in birds to pair bonding in rodents to parental bonding in primates, rather than in the accumulation of social experience with a responsive caregiver. A sensitive period is a biologically regulated window during which experience shapes neural organization to a disproportionate degree, not because of the quality of what is experienced but because a plasticity gate is open (Knudsen, 2004; Takesian & Hensch, 2013). From an evolutionary standpoint, sensitive periods exist to reduce uncertainty about the social environment. When an organism faces a critical identification problem, such as which individual to follow, protect, or mate with, a bounded window of heightened learning allows that uncertainty to be resolved rapidly and durably (Frankenhuis & Walasek, 2020). This heightened plasticity is enabled by disinhibition, a temporary relaxation of the inhibitory constraints that normally keep learning systems in check (Hensch, 2005). Signals that generate arousal in a gate-open organism are therefore formation-relevant inputs operating on a system that is temporarily capable of being reorganized by them, a point whose implications for understanding IDS become clear once the full phase architecture is in place.

The framework organizes formation into four phases, retaining Bowlby's phase labels but departing substantially from their classical content. In Phase 1, Pre-Attachment, the gate is open and the organism is broadly socially permissive, with no specific individual yet having acquired motivational priority. In Phase 2, Attachment in the Making, arousal-linked contact with a consistently present individual drives competitive narrowing among candidates, with the system remaining indifferent to the valence of what the attachment figure provides (Freeman, 2025; 2026). In Phase 3, Clear-Cut Attachment, one individual becomes non-interchangeable and the gate closes. In Phase 4, Post-Selection Maintenance, the bond is established and its quality, security, and regulatory function become the relevant variables. This is where the vast majority of the attachment literature resides, and where the Ainsworthian constructs of internal working models and attachment security do their proper explanatory work. The full empirical case for each phase and its signalling implications is developed in Section 4.3.

The framework applies equally to parental bonding, treating it as a gated formation process on equal footing with infant attachment. The peripartum endocrine cascade opens the parental gate acutely, with consolidation to a specific infant occurring within hours to days of birth (Freeman, 2026; Hrdy, 1999; Numan & Insel, 2003), while fathers and adoptive parents arrive at the same exclusive bond through accumulated arousal-linked contact (Abraham et al., 2014; Bick et al., 2013). Parent and infant thus arrive at mutual attachment through two independent unidirectional processes on entirely different timescales, with the parental gate typically consolidating within days to weeks of birth and the infant's gate consolidating between five and eight months of age (Gunnar & Donzella, 2002; Schaffer & Emerson, 1964).

This asymmetry raises the question that the remainder of the paper addresses. If two independent formation processes are running simultaneously in the dyad on different timetables, then the signals operating between parent and infant during the first months of life may be doing different formation work for different organisms at the same time. Specifying what those signals are, how they work, and whose formation they primarily serve requires a framework for understanding animal signals as evolved biological influence. It is to that framework that we now turn.

3. Animal Signals

As discussed in the previous section, Bowlby largely drew both attachment theory in general, and the theoretical skeleton of attachment formation in particular, from Lorenz’s imprinting theory. However, the psychological community interested in proper, goal-corrected attachment (or Post-Selection Maintenance phase) and its quality has, over time, largely set aside this ethological grounding. Where the initial interdisciplinary community took into account the broader communicative apparatus shared across species, later work tended to narrow its focus toward human-specific, largely verbally-mediated accounts of caregiver-infant interaction.

In order to regain this ethological grounding, the present section provides a brief review of a few essential aspects of animal communication. The aim is to outline a theoretical framework broad enough to encompass all animals, humans included, rather than one built for non-human species with humans treated as an afterthought. This proves especially helpful when considering the developmental window at the heart of this article, namely the first five to eight months of life during which attachment formation unfolds, and during which infant communication bears considerably more resemblance to that of other animals than to the language use that emerges later.

3.1. Signals as Influence

In its use among animals other than humans, a consensus has not been reached as to how exactly do signals trigger behaviour in their recipients. It is not clear either whether such signals generate any sort of ‘meaning’ (Cheney & Seyfarth, 2010; Oller, 2004; Rendall, Owren, & Ryan, 2009) or whether they convey any kind of “information”. A commonly cited example is the vervet monkey alarm call system. Distinct calls are associated with different predators, leading some authors to treat these vocalizations as if they possessed lexical meaning. Critics argue that this conflates Shannon information with semantic information (Rendall & Ryan, 2010; Rendall, Owren & Ryan, 2009). In the canonical Shannon-Weaver sense (1949), an alarm call merely reduces uncertainty about environmental events; describing it as “meaning eagle” introduces a representational claim that information theory was never intended to support. As with a fire alarm, which reduces uncertainty about the presence of fire without literally stating “the building is on fire,” the concern is that animal signals are often interpreted through a linguistic lens that exceeds what the concept of information alone can justify.

The main alternative has been the notion that the function of signalling is to influence the behaviour of perceivers rather than transmitting any language-like information (Guilford & Dawkins, 1991; Owings & Zeifman, 2004). Such an approach highlights links between acoustic structure and vocal function, particularly if senders are assumed to use whatever calling strategies allow them to influence receivers (Owren & Rendall, 2001). Corollaries include emphasizing the role of signal structure in exerting influence and expanding the conception of communication beyond representational exchanges.

Influence-based theories highlight the role of signals' structure such as the motivational-structural rules of animal communication (Morton, 1977), according to which the structural convergence of many animal sounds (mainly birds and mammals) used in ‘hostile and ‘friendly' contexts, suggests a relationship between sound structure and function. Morton’s motivational-structural rules continue to occupy an important place in contemporary communication research both within animal behaviour (Garcia and Favaro, 2017) and human psychology (Tkáčik et al., 2024). To provide a relevant example, it has been posited that IDS's high pitch (a core acoustic feature) is at least partially intended to signal as a non-threatening, affiliative signal (small-body-size correlate) alongside attention-getting and didactic purposes (Trainor et al., 2000). Morton's theory relies on indexical information, meaning the signal indexes something about the source rather than carrying meaning in its own right. This principle is probably more applicable the younger and less enculturated the infant, that is, before attachment has consolidated and language has emerged.

3.2. Signal Manipulation: Honest Signals Versus Machiavellian Deception

A further layer of complexity arises when signallers anticipate how a recipient will react and exploit that reaction. Across and within species, animals are under constant pressure to avoid being deceived by others, and the deceptive abilities different species display have frequently been framed in terms of Machiavellian intelligence (Harms, 2004). Within this framework, the notion of honest signalling is particularly relevant. Honest signalling, in the ethological sense, refers to signals whose reliability is guaranteed by their cost of production (Zahavi, 1975). Signals persist evolutionarily precisely because the investment required to produce them makes them difficult to fake, and therefore trustworthy indicators of the signaller's actual state or quality.

A useful illustration of an honest signal that will recur in later sections is the cry of a helpless human infant. Its acoustic properties reflect substantial vocal effort (the grain of the voice often conveys that the infant is literally hurting their throat because of the vocal effort), and nearby adults register this effort as evidence of genuine investment on the infant's part, making the cry extremely difficult for potential caregivers to disregard. The situation differs markedly for an older child who has discovered that crying can serve Machiavellian ends, since feigning distress can increase their influence over a caregiver's behaviour. Yet faked crying demands far less effort than the real thing (e.g. it will not hurt the child’s throat), and caregivers are generally able to detect this difference. Consequently, a signal lacking this kind of honesty may be nuanced or simply disregarded by the adult. A comparable logic, as later sections will show, underlies the Infant-Directed Register, whose acoustic properties appear to have evolved precisely to function as an honest marker of caregiving investment.

3.3. Signaller and Receiver. Animal Signals as Evolved Influence

A significant number of cases in animal communication can be understood by considering the signaller and the signal's features alone, and receiver responses to vocalizations have accordingly been treated as largely fixed and invariant (Wallman, 1992). However, a growing number of cases prove to be less straightforward, particularly in nonhuman primates and in the parent-infant context. Even when the structural features of a signal seem clear, the possibility of the receiver failing to respond as expected cannot be excluded, nor can the possibility of more sophisticated inference coming into play (Owren & Rendall, 2001). It is therefore crucial to separate a given vocalization and its intended purpose from the influence it actually generates in the receiver, where the purpose of a signal is determined by its evolutionary history rather than by any intention on the part of the signaller.

A useful theoretical lens for this separation is assessment-management theory (Owings & Morton, 1997). Assessment involves self-interested adjustments to circumstances based on signal-based cues extracted from other individuals. Management involves equally self-interested efforts to regulate the behaviour of others by means of signals. The two processes operate simultaneously and independently of communicative intent. A vivid illustration comes from Owings and Zeifman (2004), who describe a mother experiencing weeks of escalating, inconsolable crying from her colicky infant. From an assessment-management perspective, the infant's cry functions on the management side regardless of any intention on the infant's part: its acoustic properties are simply potent enough to compel a behavioural response from the caregiver, mobilizing care even when none of the mother's interventions succeed in stopping it. At the same time, the cry carries assessment value for the listener, who extracts indexical cues from its persistence, intensity, and patterning that eventually contributed to a pediatrician’s diagnosis of colic and a shift in how the mother interpreted and responded to the signal. Owings and Zeifman extend this point with a second example: when a crying infant's voice escapes through a cracked laboratory door, it reliably draws in researchers, colleagues, and secretaries with no caregiving relationship to the infant at all, demonstrating that the cry's management effect operates on listeners largely independent of any communicative intent, relationship, or awareness of who is crying. Together these examples show how a single signal can simultaneously inform and influence a receiver without requiring that the signaller intend either outcome, illustrating the core separation between signal production and its uptake that the theory is built to capture.

Two observations from the parent-infant literature reinforce the same point for the developmental window this paper is most concerned with. On the production side, symbolic communication, meaning the capacity to express intention or meaning through sound or gesture, emerges only at the end of the first year and depends on cognitive development that is not yet in place during the attachment formation window (Lipschits & Geva, 2024). Before this point infant vocalizations may well lack any intention to inform in the sense discussed for animal callers (Rendall et al., 2009), even as caregivers routinely respond to these signals as though they were intentionally communicative. On the inferential side, even six to twelve month old infants show pragmatic inference from ostensive communication, recognizing communicative actions and recovering intended referents (Tauzin & Gergely, 2018). This is precisely the kind of more sophisticated inference that Owren and Rendall (2001) caution against assuming by default: the receiver's recovery of intended content is a separate cognitive achievement, not something automatically delivered by the signal's acoustic form. Taken together, these observations confirm that during the attachment formation window, infant communication is operating closer to the animal signalling end of the continuum than to the language end, and that influence rather than intention is the appropriate lens through which to understand it.

These arguments make precise what we mean when we refer to an animal signal throughout this paper. A signal is understood not merely as influence but as influence determined by evolutionary history, specifically as “any act or structure which alters the behaviour of other organisms, which evolved because of that effect, and which is effective because the receiver's response has also evolved” (Smith & Harper, 2003, p.3, emphasis added). This definition draws a consequential line between signals and cues. A leopard's sharp teeth may convey ferocity to a smaller animal hiding nearby, prompting escape, but that effect is not what the teeth evolved for. The same teeth deterring competing males, however, is something they did evolve for, and in that context they function as a genuine animal signal. Accordingly, whereas animal signalling may occasionally fail, biological communication is understood here as the successful completion of a signalling act in which an evolved signal triggers the evolved response it was selected for (Scott-Phillips, 2008).

The role of the receiver is therefore a necessary component of both Smith and Harper's conception of animal signals and Scott-Phillips' conception of biological communication. It is also, as the following sections will argue, a necessary component for understanding what IDS and infant signals are doing in the attachment formation context, and for whom.

4. Attachment Vocalization Feedback

With the attachment formation framework and the animal signalling notions now in place, we can turn to the specific signal systems operating in the human parent-infant dyad. In the following sections we will, to a certain extent, argue in favour of Owings and Zeifman’s (2004) suggestion that the attachment vocalization feedback system constitutes a crucial component in the process of human attachment. However, by scrutinising the subject in greater detail, we will specify that it is rather during attachment formation phases before clear-cut attachment where such acoustic feedback can properly be understood as signal serving proper biological communication as previously defined.

4.1. Caregiver-Directed Infant Vocalization

Joseph Soltis’ (2004) extensive review remains a cornerstone analysis of the possible functions of early infant crying as a costly signal (see section 3.3) and serve as this section’s structure. His central claims support the idea that caregiver-directed infant crying is part of an imprinting dynamic, and that the evolved response to such an honest signal of distress is caregiving behaviour.

First, he recognizes early infant crying as an important means through which infants can maintain contact with the mother because active proximity-seeking behaviours such as following have not yet developed. Such a claim echoes Falk’s (2004a) arguments that rely on the inability of the human infant to perform the proximity seeking behaviour of clinging that remains essential for the rest of nonhuman primates.

Bowlby (1969) summarized different sorts of crying, acoustically differentiable given the nature of their origin. For instance, crying from hunger starts gradually and becomes rhythmical, whereas crying from pain starts suddenly and is rather arrhythmical. Soltis’ second claim is that although acoustically-distinct cry types reflecting specific needs progressively arise with development, specialists seem to agree on the notion that adult-directed infant cry’s main function is to maintain proximity to the carer (see also Zeifman, 2001). As Bowlby (1969) points out, infant crying more often than not is effectively terminated by stimuli that, in a natural environment, are almost certain to be of human origin, thus maximizing caregiver proximity.

Thirdly, Soltis’ points out that infant cry stimuli result in brain activity in areas hypothesized to be involved in mammalian parenting behaviour. Further review of the neural correlates of infant crying and parents’ responses to it also portray the highly affective nature of this bonding dynamic; parents experience high levels of distress as a response to that of their offspring, both being simultaneously appeased once contact is resumed. On the side of the infant, physical separation from the mother evokes separation calls in a variety of mammalian infants, the acoustic structure of the human infant cry being similar to that of the separation calls of nonhuman primate infants (Panksepp, 1995).

Fourth, human mothers and infants exhibit features of adaptation typical of mammalian species that carry their infants, as opposed to species that tend to cache them for long periods of time (Blurton Jones, 1972; Zeifman, 2001). In the latter species, infants have independent thermoregulatory mechanisms, and do not vocalize when separated. In contrast, in carrying species such as humans, mothers and infants are in more continuous contact and feedings are more frequent since independent thermoregulation is poorly developed in the infant at birth. In general, mammalian mothers do not respond to isolation calls indiscriminately, but rather a mother’s responsiveness depends on whether the vocalizer belongs to her own offspring (Newman, 2003). The fact that human beings are a notable exception to this trend insofar as caregiving is administered less selectively (Chang, 2013) does not prevent parents from recognizing their own. Sound spectrograms indeed show that 'cry-prints' are as distinctive as finger-prints and thus facilitate identifying new-born babies. In her response to Soltis’ article, Falk reinforces the idea that IDS and caregiver-directed infant cry are complementary behaviours that initially evolved in our hominin ancestors in conjunction with the evolution of bipedalism (Falk, 2004).

It is important to mention that as much as infant crying enhances bonding between caregivers and infants, it can also be involved in conflict between them. In this regard, parent-offspring conflict describes the contrasting fitness interests of parent and infant in terms of parental investment (Trivers, 1974). It predicts that an infant should strive to elicit more investment than a caregiver should optimally provide. Because such an excessive demand would prevent them from investing in existing or future siblings, or attending any other matters, caregivers must assess the costs and benefits of abiding the infant’s request and will sometimes decide not to. As a form of management in an assessment-management dynamic (section 3.4), crying is a self-interested effort to maintain or obtain caregiving by regulating the behaviour of others. Unlike in the case of a neonate, as children grow older and become more intelligent, they eventually come to understand the influential power of their own crying. This would normally occur between the third and fourth phases of attachment (from around 18 moths of age, (see Milligan, Astington, & Dack, 2007), when the infant starts understanding the intentions of others. A case for Machiavellian deception (section 3.3) can thus be made once a child is able to foresee the caregiver’s reaction to their spontaneous crying, and generates a “dishonest signal”— exaggerated or faked crying —as a measure for obtaining whatever they seek. Parents must therefore rely on their ability to appraise the honesty of a signal, as well as modulate their own neural (in general) and emotional (in particular) responses to it. In this matter, sex and parenting experience have been posited as factors mediating response modulation to infant vocalizations (Seifritz et al., 2003).

Thus, this section thus succinctly stresses the evolutionarily-unprecedented importance of this acoustic signal in a human infant’s survival. Authors suggest infant crying has involved both the context of parent-offspring conflict and attachment, the latter remaining as its primary function in terms of seeking proximity to the carer, thus securing caregiving and ultimately survival.

4.2. Infant-Directed Speech

The IDReg comprises not only IDS but also infant-directed singing, a behaviour proven to be equally universal (Hilton et al., 2022). However, most of the literature informing this article—as well as the broader literature in general—has focused on IDS. Because the characteristics of these two vocal behaviours largely diverge, and because a comprehensive review of both phenomena would require maintaining a distinction between them throughout the article, thereby substantially complicating the argument, we will focus exclusively on IDS.

Evolutionary modifications of the human pelvis associated with the emergence of habitual bipedalism imposed important obstetrical constraints, favouring the birth of increasingly immature, helpless and dependent infants (Falk, 1998). One consequence of this developmental shift was the loss of the infant’s capacity to maintain continuous physical contact through active clinging, a behaviour that remains common among many nonhuman primates. As direct bodily contact became less constant, mother–infant communication increasingly relied on distal forms of interaction. Gestural exchanges expanded in importance (Tomasello & Camaioni, 1997), while affectively charged vocal communication became a pervasive means of maintaining social engagement in the absence of sustained physical proximity. This evolutionary transition is thought to have promoted both the elaboration of infant crying and the emergence of specialized caregiver vocal responses designed to regulate and reassure the infant at a distance.

Infants demand parental investment, an important subcomponent of which is attentional investment. In the case of an infant that cannot yet move and a caregiver that cannot immediately approach it, parents will proceed to vocal engagement. Infants should thus have been under selection to discriminate true signals, to attempt to detect faked ones, and to resist them (Dawkins, Krebs, Maynard, & Holliday, 1979). In other words, infants strive to detect when attention is actually being paid to them. As a response to said selective pressures, the special acoustic features of IDS would have initially been selected to engage, and sooth crying infants that were out of physical grasp. Falk (2004b, p. 462) famously characterized these vocal responses as “disembodied extensions of mothers’ cradling arms.”

As discussed in the previous section, infants demand parental investment, an important subcomponent of which is attentional investment. In the case of an infant that cannot yet move and a caregiver that cannot immediately approach it, parents will proceed to vocal engagement. Infants should thus have been under selection to discriminate true signals, to attempt to detect faked ones, and to resist them (Dawkins, Krebs, Maynard, & Holliday, 1979). In other words, infants strive to detect when attention is actually being paid to them. As mentioned in the introduction, a number of authors have directly or indirectly related IDS to attachment. For instance, in the previous section, a detailed examination was laid, of how infant cry triggers caregiving behaviour. The latter often consists of tactile behaviours like cradling and stroking, as well as vocal ones such as IDS. In turn, these measures will cease the behaviour that prompted them in the first place. By this token, an important extent of parent-infant attachment develops through dynamic and interactive signalling loops of care and cry (Swain, Mayes, & Leckman, 2004). Similarly, the high-pitched voice of IDS also facilitates smiling and other proto-conversational behaviours in the infant. In turn, the infant’s proto-conversational feedback elicits IDS in general, and dynamically affects the height of its pitch (Smith & Trainor, 2008).

As briefly mentioned in Section 2, other important releasers of caregiving behaviour are paedomorphic traits. Protruding cheeks, a large forehead and large eyes below the horizontal midline of the skull have been reported to be prioritized by the human attention system (Brosch, Sander, & Scherer, 2007) and elicit perceptions of cuteness (Glocker et al., 2009). Accordingly, infants’ degree of paedomorphic in their facial traits predicts maternal behaviours and attitudes (Langlois, Ritter, Casey, & Sawin, 1995). Crucially, empirical studies have also shown that paedomorphic traits elicit IDS (see Zebrowitz, Brownlow, & Olson, 1992). In particular, paedomorphic traits seem to release the positive affective tone that characterises IDS (Trehub et al., 1997; de L’etoile, 2006). In this respect, Trainor, Austin and Desjardins concluded that IDS reflects free vocal expression of emotion to infants, as opposed to the more inhibited expression of affect that characterizes most stances of adult-directed speech (Trainor, Austin, & Desjardins, 2000).

A final argument for considering IDS as part of attachment is the progressive acknowledgment of ‘secondary baby talk’ (Ferguson, 1977): speech that has the acoustic features of IDS yet is directed towards addressees other than infants or children. It was previously discussed how imprinting during infancy has an impact in mating later in life, and how attachment describes the incidence of early attachment in partner choice and in old age. Accordingly, IDS directed to friends (Bombar & Littig, 1996) and romantic partners (also called ‘loverese’) has been documented (Bombar & Littig, 1996; Chang & Garcia, 2011). This altered tone of voice is used between roughly two thirds of romantic partners of both genders, across the lifespan (Chang & Garcia, 2011). By representing the emergence of what can be considered an ‘adult-directed variant of IDS’, loverese entails a human instance of early attachment having impact on adult mating behaviour. In other words, loverese seems to be an adult consequence of IDS, the latter being a releaser once delivered by the carer.

It has also been reported that caregivers of institutions for the aged use IDS when addressing elderly people (Caporael & Culbertson, 1986). Content-filtered samples of this form of IDS was identified as speech to children by naïve listeners, regardless of the actual (old) age of the targets. Caporael argues that, like paedomorphic traits, the appearance of elderly people would be designed to convey helplessness and thus elicit caregiving behaviour.

A major focus of research on IDS has stressed its role in language acquisition, often not stressing enough its intimate link with caregiving and attachment. The first scholars to fully articulate and emphasized IDS’ role in processes of intimate psychological connection, not only between mothers and infants but also in other relationships, were American psychologists Meredith Bombar and Lawrence Littig. Having acknowledged that IDS was directed towards adult, literate humans, the authors argued that expressing and facilitating intimate psychological connection is a central— if not the main —function of the IDReg (Bombar & Littig, 1996).

4.3. IDS Through the Gated Sensitive Periods of Attachment Formation

The preceding sections have established IDS as a biologically evolved signal system in the formal ethological sense, whose design features are adapted for the attachment context. What remains underspecified is the temporal architecture within which these signals operate, whose attachment formation they primarily serve, and at what point in development their effects are most consequential. The phase-based sensitive period framework introduced in Section 2 addresses these questions directly, and in doing so reframes the functional significance of the signals reviewed above.

The framework organizes formation into four phases whose names are retained from Bowlby (1969) and Ainsworth and colleagues (1978) but whose content departs substantially from the classical account. What follows develops the empirical case for each phase and specifies what signals do within it.

Phase 1, Pre-Attachment, is the gate-open state. The organism is broadly socially permissive, meaning the disinhibitory state is in place, social uncertainty is at its maximum, and any sufficiently salient individual within the eligible class can in principle become the object of selection. This is not a passive waiting room but an actively regulated biological readiness. In a newly hatched chick, gate opening precedes any specific social contact. In a postpartum ewe, the peripartum endocrine cascade opens the gate acutely before she has encoded her own lamb's odor. In a human infant, a maturational process gradually installs the permissive state across the first months of life. In each case the gate-open organism is primed to receive signals from eligible targets, and the signals received under these conditions carry disproportionate weight. Species-typical social signals, including vocalizations, faces, and movement patterns, orient the broadly permissive organism toward the class of eligible targets and begin concentrating attention on specific individuals within that class. For the human infant, the IDReg is precisely this kind of Phase 1 signal, orienting the broadly permissive infant toward the species-typical class of caregiver stimuli before individual-level selection is possible.

Phase 2, Attachment in the Making, is the period of competitive narrowing within the open gate. Arousal-linked contact with a consistently present individual begins accumulating incentive salience toward that individual, meaning they acquire motivational weight disproportionate to alternatives. Candidates remain substitutable at this stage but the competition is no longer equal. The formation system during this phase is indifferent to the valence of arousal, consistent with the influence-based account of signalling described by Owren and Rendall (2001) and with the assessment-management framework of Owings and Morton (1997). The valence independence of Phase 2 is well documented across species and relationship types. Neonatal rats exposed to aversive odor conditioning during the stress hyporesponsive period, a functional analogue of Phase 1 gate opening, show paradoxical approach and attachment to the conditioned odor rather than avoidance, an effect that disappears once the hyporesponsive period ends and the gate closes (Sullivan et al., 2000). Human infants form selective attachments to abusive caregivers at rates comparable to non-maltreated samples (Cyr et al., 2010), and the same pattern appears across relationship types wherever arousal-linked contact occurs within an open gate (Freeman, 2026). Aversive experiences do not merely fail to prevent bond formation. Under the right conditions they actively accelerate it.

The quality-of-care independence of Phase 2 is equally well supported. Neither Schaffer and Emerson's (1964) nor Ainsworth's (1967) longitudinal studies found any caregiving variable, including responsiveness to crying, warmth, or availability, to predict when attachment formed or toward whom (Freeman, 2025). Harlow's (1958) macaques consolidated around nonresponsive cloth surrogates, and human romantic attachments regularly consolidate around indifferent partners (Baumeister et al., 1993). What evolution shaped was the capacity to form a bond, not the capacity to form a secure one. These findings are not anomalies. They follow directly from the gate logic, since what drives narrowing is arousal-linked presence, and the system is temporarily indifferent to what that presence provides.

Signals matter most in Phase 2, and honest signalling in the ethological sense is particularly consequential here. Because the gate-open receiver has no established history with the signaller and no representational system through which to interpret their motivations, the only reliable basis for response is the signal itself. Costly signals, whose reliability is guaranteed by what they require to produce, work precisely because they cannot easily be faked under these conditions. Any signal that generates sustained arousal in the gate-open organism and is reliably associated with a specific individual contributes to the accumulation of incentive salience toward that individual through the arousal-disinhibition pathway that Phase 2 requires, rather than through information transfer or the interpretation of intent.

Phase 3, Clear-Cut Attachment, marks consolidation and gate closure. One individual becomes non-interchangeable through active inhibitory stabilization of their representation and suppression of competing alternatives (Freeman, 2026; Hensch, 2005). The bond is now exclusive in the precise sense that no alternative target can displace the selected individual regardless of arousal. Signalling continues to matter in Phase 3, but its role shifts. Rather than driving competitive narrowing, signals now function to confirm the selected individual's presence and sustain the consolidating bond during the transition to post-selection maintenance. As the gate closes, the receiver is also beginning to accumulate a relational history with the selected individual, which means signal properties alone are no longer the only basis for response.

Phase 4, Post-Selection Maintenance, is where the vast majority of the attachment literature resides. Security, affect regulation, and individual differences in attachment organization are post-formation phenomena that describe how an established bond functions, not how it was built. Signalling in Phase 4 operates in a fundamentally different context from the preceding phases. The receiver now has an accumulated relational history, internal working models representing the signaller's patterns of availability and responsiveness, and the cognitive capacity to interpret signals in light of what they know about the signaller's motivations and intentions. This is precisely the goal-corrected partnership that Bowlby described, but understood correctly it is a post-formation achievement rather than a formation criterion. A parent whose bond has consolidated can assess whether their toddler's cry reflects genuine distress or a learned strategy for obtaining attention, because they now have the relational history and representational apparatus to interpret the signal rather than simply respond to it. The assumption of honest signalling that characterizes Phase 2 no longer holds by default in Phase 4. Signals are evaluated rather than absorbed, and the quality and contingency of ongoing interaction shape the bond's function in ways that arousal alone cannot.

The same gated formation logic applies to parental bonding, and the parental formation sequence has its own distinct trigger and timetable. Phase 1 gate opening in parental bonding can occur through a discrete biological event or through accumulated contact. In the first route, the peripartum endocrine cascade, including estrogen withdrawal, oxytocin release at delivery, and rising prolactin, opens the gate acutely and consolidation to a specific infant can occur within hours to days of birth (Freeman, 2026; Hrdy, 1999; Numan & Insel, 2003). In rodents the same cascade drives consolidation through olfactory encoding of the specific pup in the first hours after birth (Kendrick et al., 1997; Numan & Insel, 2003). In the second route, accumulated arousal-linked contact gradually opens the gate without a discrete hormonal trigger, as observed in fathers and adoptive parents who arrive at the same exclusive, non-interchangeable bond through sustained exposure (Abraham et al., 2014; Bick et al., 2013).

The priming state of the receiver matters considerably here. A gate-open parent in the immediate postpartum period, whose hormonal milieu is actively reorganizing social salience, is a fundamentally different receiver than a sexually mature but biochemically unprimed adult. Signals directed at a primed receiver are operating on a system already tuned toward selective social investment, which is precisely the condition under which Phase 2 arousal-target coupling proceeds most rapidly toward consolidation. The parallel with honest signalling is direct. Just as Phase 2 in infant attachment depends on a gate-open receiver whose response is driven by signal properties rather than relational interpretation, parental Phase 2 depends on a hormonally primed receiver whose incentive salience system is maximally responsive to infant signals. Recognizing parental bonding as a gated attachment process on equal footing with infant attachment therefore reframes not only the functional significance of infant signals but also the question of whose formation sequence those signals are primarily serving, a question that Section 4.4 addresses directly.

4.4. The Signaller as Attachment Figure

The standard account inherited from Bowlby treats infant signals as mechanisms of infant attachment. Crying, smiling, vocalizing, and displaying paedomorphic features (rounded face, large eyes, and protruding forehead characteristic of young infants) are classified as attachment behaviours that serve the infant's goal of maintaining proximity to a caregiver (Bowlby, 1969; Ainsworth et al., 1978; Carlson & Sroufe, 1995; Cassidy, 1999; Zeifman, 2001). In such an account, the functional outcome presumed to matter is the infant's own eventual bond. This framing is at least partially inconsistent with the animal signalling literature, and an ethological approach leads to a fundamentally different conclusion: the signaller is the candidate for selection, and signals function to disinhibit the receiver so that the signaller is more likely to be perceived as an attachment figure. This is as true of infant signals directed at parents as it is of courtship displays in pair bonding species, or of parental vocalizations directed at infants.

In parent-offspring systems across birds and mammals, offspring signals are understood as working on the parent's investment and bonding systems rather than on the offspring's own. Crucially, these signals carry real costs. Nestling begging calls attract predators to active nests, with empirical studies demonstrating significantly higher predation rates at nests broadcasting begging calls relative to silent controls (Leech & Leonard, 1997; McDonald et al., 2009). Infant crying in human ancestral environments carried the same risk, as cries of helpless infants are easily detected by predators (Soltis, 2004). Signals that place the signaller at risk of predation can only be maintained by natural selection when the benefit of being selected as the object of parental investment outweighs the cost of exposure. The signal is working on the parent's motivational system, and its costliness is precisely what makes it an honest indicator of the infant's viability as a target for that investment.

Lorenz's observation that paedomorphic infant features function as innate releasing mechanisms activating caregiving behaviour in adults has been extensively supported by behavioural and neuroimaging evidence showing rapid reward circuit activation in response to infant faces across parents and non-parents alike (Brosch et al., 2007; Glocker et al., 2009; Zebrowitz et al., 1992). In pair bonding contexts the same logic holds. Elaborate courtship displays carry predation costs equivalent to those of begging, and they exist because the signaller benefits from being selected as a mate, not from doing the selecting (Zahavi, 1975). Across relationship types the signaller is the candidate and the receiver is the selector. In evolutionary terms, the signal is saying “pick me”.

Freeman’s (2026) sensitive period framework makes explicit what is implicit in this ethological evidence. What infant signals accomplish during the parent's open gate is precisely what disinhibition and arousal-linked exposure accomplish in any Phase 2 system. They reduce avoidance and concentrate incentive salience in the parent toward a specific infant. Signals that generate high arousal in a gate-open parent, such as intense crying resolved by contact or a compelling social smile, drive precisely the arousal-target coupling that Phase 2 requires for competitive narrowing. We designate this system the “cariver-directed register”, or CDReg, to distinguish it from the IDReg operating in the complementary direction. The IDReg does Phase 1 work on the infant's side, orienting the broadly permissive infant toward the species-typical class of caregiver stimuli before individual-level selection becomes possible (Robledo et al., 2022; Freeman, 2026). The CDReg does Phase 2 work on the parent's side, driving the arousal-target coupling that accumulates incentive salience toward a specific infant and ultimately consolidates an exclusive bond in which that infant’s non-interchangeability is amplified or consolidated.

The temporal profiles of the two systems are offset in a way that reflects their different functions. CDReg signals are most intense earliest. Crying peaks in the first three months before declining (Soltis, 2004), the social smile emerges around six weeks (Bowlby, 1969), and paedomorphic features are present from birth as a fixed morphological stimulus to the parental reward system (Brosch et al., 2007; Glocker et al., 2009; Langlois et al., 1995). This front-loading corresponds to the period when the parental gate is most recently opened and Phase 2 coupling is most consequential. The IDReg is most acoustically exaggerated across the broader zero-to-six-month window and remains functionally central throughout the first year (Cooper & Aslin, 1990; Fernald, 1985; Kalashnikova et al., 2018; Soderstrom, 2007), spanning parental Phase 4 maintenance on one side and infant Phase 1 through Phase 3 on the other.

The infant does not signal in order to attach. The infant signals in order to be attached to. Once parental consolidation is achieved, the infant's own formation sequence proceeds within the protective context that parental attachment provides, with the parent as a consistently present and arousal-generating target across the infant's own Phase 2 narrowing. Situating the IDReg within this formation architecture clarifies both its timing and its function. As a Phase 1 instrument operating on the infant's broadly permissive social system, it is not in competition with the CDReg but complementary to it, each signal system doing phase-specific work on a different organism's formation sequence within the same developmental window. Together they constitute a coordinated biological communication system whose full significance only becomes visible once the formation processes they serve are distinguished from each other and placed on their proper developmental timetables.

5. Conclusion and Future Directions

This paper has advanced two interconnected arguments about the relationship between infant-directed speech and attachment. The first is that IDS is best understood as a formation-relevant signal, operating within a biologically regulated window of heightened plasticity (Knudsen, 2004; Takesian & Hensch, 2013). The second is that the signals operating in the parent-infant dyad during the attachment formation window constitute two independent evolved registers serving different organisms on different timetables (Freeman, 2026). The CDReg drives parental attachment to the infant. The IDReg drives infant attachment to the parent. Both involve co-evolved signal-response pairs in the Smith and Harper sense (Smith & Harper, 2003), meaning both signal production and receiver response have been shaped by natural selection. The direction of bond formation, however, is independent of this co-evolution. The standard account from attachment theory treated the mutual interactive exchange between parent and infant as a single shared bonding mechanism (Bowlby, 1969; Ainsworth et al., 1978). The present framework treats it as two independent unidirectional processes that produce mutual attachment through shared presence during overlapping sensitive windows rather than through a common formation mechanism (Freeman, 2026).

This account is also a better fit with evolutionary theory. Selection operates on individual organisms rather than on dyadic relationships (Trivers, 1974), and the evidence supports each organism adapting on its own terms. Mutual attachment emerges as a convergent outcome of two independent formation processes rather than as the product of a mechanism that evolved to serve the bond itself.

A further contribution of the paper concerns the phase-specific timing of infant and parental signalling across the attachment formation window. To our knowledge, no prior account has specified when each signal system matters most, for whom, and why the temporal profile of each is calibrated to the formation process it serves. The IDReg is most acoustically exaggerated and perceptually salient across the zero-to-six-month window, spanning parental Phase 4 maintenance on one side and infant Phase 1 through Phase 3 on the other. The CDReg is most intense earliest, front-loading its salience during the period when the parental gate has most recently opened and Phase 2 coupling is most consequential. The two systems are offset rather than concurrent, and that offset is precisely what the independent timetable account predicts. Honest signalling, in the ethological sense, is also most consequential during Phase 2 when the gate-open receiver has no relational history with the signaller and the only reliable basis for response is the signal itself. As the bond consolidates in Phase 3 and the gate closes, signal evaluation gradually replaces signal absorption, and the relationship enters the goal-corrected interpretive mode that characterizes Phase 4.

The phase-based account also makes sense of a phenomenon that might otherwise seem peripheral to attachment formation. The use of IDS in adult romantic relationships, documented as loverese (Bombar & Littig, 1996; Chang & Garcia, 2011), suggests that the IDReg is reactivated whenever a gate-open receiver is present. If the IDReg evolved to orient a Phase 1 permissive organism toward an eligible attachment figure, its reappearance at the onset of adult pair bonding, when a new formation window opens, is precisely what the framework predicts rather than a curiosity requiring separate explanation.

The question of how human attachment fits within the broader phenomenon of animal bonding remains comparatively understudied within mainstream attachment research, and the present paper is an attempt to narrow that gap. Much productive work is already underway in ethology and comparative psychology, yet it is not always followed by researchers whose primary focus is human attachment quality, perhaps owing to differences in disciplinary tradition and vocabulary. Engagement with this literature need not require wholesale adoption of ethological frameworks. Even partial engagement stands to enrich existing accounts of attachment formation considerably.

Across studies bringing together IDS and attachment, none define attachment from first principles or address early attachment formation in ethological terms. Existing work treats the construct either as a nominal placeholder for emotional warmth (de l'Etoile, 2006; Smith & Trainor, 2008; Swain et al., 2007; Trehub et al., 1997) or as a procedural classification for predicting relationship quality (Beebe et al., 2010; Granat et al., 2017; Shai & Belsky, 2017). Smith and Trainor (2008) and Trehub and colleagues (1997) examine the functional mechanics of IDS as an interactive register, and Beebe and colleagues (2010) capture data at four months, a point within the pre-clear-cut attachment window. Both studies, however, interpret this period as either a socio-cognitive learning loop or a predictor of future security rather than as a biological formation mechanism. The present paper stands largely alone in identifying this early period as an asymmetric, open-gated evolutionary intersection governed by distinct biological signalling systems.

Grounding early interaction data within a model of asymmetric phase-specific sensitive periods reframes several prominent empirical findings. Beebe and colleagues (2010) micro-analyse four-month communication to predict twelve month Strange Situation classifications, treating early interaction as the origin of relationship quality. The present framework reinterprets this window differently. At four months the parental gate has already closed and parental attachment is in Phase 4 maintenance, while the infant's gate remains in its peak period of open plasticity. What is described as co-created relationship quality can therefore be understood as an active influence system in motion, where IDReg signals are driving the infant's ongoing biological selection process rather than laying down a relational template. Granat and colleagues (2017) evaluate social synchrony at nine months, concluding that divergent synchrony levels in depressed and anxious mothers disrupt the developing attachment template. By nine months both parental and infant gates have consolidated, meaning clear-cut attachment has already formed. Applying the honest signalling framework to these data suggests that maternal depression may reflect difficulty emitting the high-cost signals required to sustain a Phase 4 bond, while anxiety may produce hyper-synchronous signalling that encroaches on the infant's emerging autonomy.

Swain and colleagues (2007) map parental brain responses to infant cries at two to four weeks postpartum, framing early neural activations as foundations for future relationship expectations. The present framework locates this activation at the rapid closing of the parental consolidation gate, where intense early motivation functions not to establish a cognitive template but to sustain the months of asymmetric IDReg signalling that the infant's still-open selection gate requires. The importance of attachment quality and its profound consequences for individual wellbeing are not in question here. The aim is to show that the mechanisms of formation and the mechanisms of quality are distinct, and that conflating them has obscured both.

Several limitations of the present paper are worth noting. We deliberately excluded an analysis of infant-directed singing as an evolved signal in attachment formation. The acoustic features of infant-directed singing have been argued to conform to Krebs and Dawkins' criteria for effective ritualized signals, including redundancy, rhythmic repetition, and the use of supernormal stimuli, leading Mehr and Krasnow (2017) to propose that it constitutes an even more honest signal of parental attentional investment than IDS. That argument was developed within a parent-offspring conflict framework without explicitly addressing implications for attachment formation. Incorporating it within the present framework is a natural next step and one we hope future work will take up.

The debates surveyed here do not exhaust the relevant literature in animal communication. Multimodal signalling, signal ontogeny, and the evolution of signalling systems under varying social structures were left unaddressed and may bear further on the questions raised here. The present paper is best understood as a first pass at a problem that will benefit from continued cross-disciplinary engagement rather than as a definitive treatment.

The present paper has focused on establishing the case for IDS as a formation-relevant signal within an ethological framework. This was a deliberate theoretical choice rather than an exhaustive review, and we recognize that a complete account will require engagement with evidence and arguments that push back against the positions advanced here. We welcome that engagement. The framework is most useful not as a settled account but as a set of testable claims about when signals matter, for whom, and through what mechanisms, claims that can be evaluated against existing and future empirical work across developmental psychology, ethology, and comparative neuroscience.