1

Growing a Social Brain 1

2

Shir Atzil* 3

Hebrew University of Jerusalem 4

Mt. Scopus, Jerusalem, Israel 5

shir.atzil@mail.huji.ac.il 6

7

Wei Gao 8

Cedars-Sinai Medical Center 9

Los Angeles, CA, US 10

Wei.Gao@cshs.org 11

12

Isaac Fradkin 13

Hebrew University of Jerusalem 14

Mt. Scopus, Jerusalem, Israel 15

itzik.fradkin@mail.huji.ac.il 16

17

Lisa Feldman Barrett* 18

Northeastern University 19

Massachusetts General Hospital and Harvard Medical School 20

Boston, MA, US 21

l.barrett@neu.edu 22

23 24

25

*Correspondence to: 26

Shir Atzil shir.atzil@mail.huji.ac.il 27

Lisa Feldman Barrett l.barrett@neu.edu 28

29

30

31

32

2

Abstract 33

It has long been assumed that social animals, such as humans, are born with a brain system that has 34

evolved to support social affiliation. However, the evidence does not necessarily support this 35

assumption. Alternatively, social animals can be defined as those who cannot survive alone and rely 36

on members from their group to regulate their on-going physiology (or allostasis). The rather simple 37

evolutionary constraint of social-dependency for survival can be sufficient to make the social 38

environment vitally-salient, and to provide the ultimate driving force for socially-crafted brain 39

development and learning. In this Perspective article, we will propose a framework for sociality and 40

specify a set of hypotheses on the mechanisms of social development and underlying neural systems. 41

The theoretical shift proposed here implies that profound human characteristics, including but not 42

limited to sociality, are acquired at early age, while social interactions provide key wiring 43

instructions that determine brain development. 44

45

46

Introduction 47

Humans are a social species. We cooperate with each other 1. We form long-term pair bonds 48

with selected individuals 2. We bear and care for children over an extended period of time 3. 49

Sociality is a survival strategy, which optimizes securing the resources that are necessary for growth, 50

protection and reproduction 4. Consequently, many human psychological features can be best 51

understood in a social context. Here, we propose an evolutionary theory of social affiliation. We 52

begin by defining a social species as one where animals regulate one another’s fundamental 53

physiological processes (or allostasis), and therefore their survival depends on social bonds. 54

Allostasis is the ongoing adjustment of an individual’s internal milieu that is necessary for survival, 55

growth, and reproduction 5, and social animals gradually learn to regulate their own and others’ 56

allostasis using social communication 6. Therefore, we hypothesize that social affiliation is rooted 57

in allostasis 6. Social dependency for allostasis regulation is a co-evolutionary plan that inherently 58

relies not only on the individual’s physiology for survival, but also on their social environment. This 59

evolutionary plan is adaptive because as a result of a relatively simple evolutionary feature (e.g. 60

inability to control your own allostasis), it maximizes social motivation and developmental 61

flexibility in learning culturally relevant knowledge and behaviours needed to survive in a specific 62

community or social niche. 63

3

All mammals, and most birds are social to some extent, as newborns cannot survive without 64

at least one dedicated caregiver. In newborns, the presence of the caretaker is to actually keep 65

newborns alive, which could have robust implications for social development. Mothers establish and 66

control allostasis in their offspring during embryonic development 7, in eggs or in the womb. This 67

physiological dependency continues once the offspring are born or hatched. Our definition of a 68

social species using the idea of allostasis-dependence, is not meant to restrict or simplify the 69

definition of a social species but to add an important dimension to social neurobiology. The idea that 70

sociality is evolutionarily related to allostasis is supported by comparative evidence linking a central 71

feature of allostasis, energy metabolism, to social strategies. This literature points that across 72

different species, higher allostatic demands are associated with more complex sociality (See Table 73

1). Given mammalian ontogeny, newborns depend on their mothers, and the initial social dyad is 74

designed to physically regulate allostasis in the infant, including energy expenditure, temperature 75

and immune function 8, and mothers implicitly provide bio-behavioural regulation to their offspring, 76

a phenomenon previously referred to by Myron Hofer as maternal Hidden Regulators 9. For 77

example, human mothers feed their infants to regulate their diet, sing and touch their infants to 78

regulate their temperature, heart rate 10 sleep, and arousal 11 (e.g., control many aspects of infants’ 79

autonomous nervous system). Importantly, while a mother regulates her infant’s allostasis she 80

provides nutrition, soothing and comfort. In effect, a caregiver’s allostatic support is rewarding 12, 81

which makes social interactions a strong reinforcement. With repeated care, the infant gradually 82

builds an internal model of the caregiver 9. Since the experience with the caregiver is repeatedly 83

associated with a vigorous reward (i.e., allostasis regulation), we hypothesize that the internal model 84

of the caregiver is acquired as rewarding, which promotes infant attachment and motivation towards 85

social interactions. Altogether, we propose here that parental care is directed towards infant 86

allostasis, and thus provides an optimal incentive for brain development and learning, as via 87

allostasis the social dyad encourages the acquisition of new behaviours and concepts that are 88

necessary for social affiliation. 89

In support of the theory suggested here, we review three lines of evidence. First, the social 90

neuroimaging literature and how the brain processes social information. We propose that neural 91

systems supporting social behaviours overlap with those supporting allostasis 13-15. According to our 92

framework, social affiliation is acquired and develops as a result of an inborn dependency in 93

allostasis. Allostasis is thus considered a domain general process, which is an important ingredient of 94

4

sociality. Correspondingly, neural systems that support allostasis represent a crucial neural 95

ingredient that wires social behaviours 16. Then, a review of brain development literature 96

demonstrates that the neural circuitry needed for social affiliation is not evident in newborns, and 97

develops throughout childhood 17. We propose that this potentiates its susceptibility to environmental 98

input. The third line of evidence is developmental neurobiology literature, which provides 99

compelling evidence that early social experience (particularly maternal care) determines the 100

offspring social biology and behaviour in adulthood. Altogether, these lines of literature points that 101

current theories of social bonding are not supported by empirical evidence, and there is a need to 102

consider an alternative theory. 103

Synthesizing new and longstanding ideas, we propose a developmental trajectory, which is 104

socially-crafted and powered by allostasis. In particular, with early-life experience the brain 105

assembles predictive models, which enables the development of a conceptual system as a 106

prerequisite for social development. We argue that in a next step development of concepts promotes 107

human social development by acquiring social concepts (such as the “mommy”), and social skills 108

(such as synchrony). Integrating empirical findings about the developmental trajectories of (a) neural 109

networks and (b) social competency, we introduce the hypothesis that brain development and social 110

development are two manifestations of the same phenomenon: becoming social experts. Thinking of 111

social affiliation as an acquired skill has scientific, clinical and societal implications. 112

113

Table 1: Comparative association between allostasis and social strategies across species 114

Level of Observation Evidence References

Comparative relationship between social strategy

and metabolism

There is a relationship between social status and standard metabolic rate in juvenile Atlantic Salmon, such that socially dominant fish have higher metabolic rates.

18

Large brains in primates were associated to both increased metabolism and increased social demands.

19, 20

Sloths have extremely low standard metabolic rate compared to other mammals. Accordingly, social interactions among sloths are rare, and sloths are known for their solitary habits.

21

high energy requirements imposed on reproductively active females are an important determinant of the phenomenon of female social dominance in the Sifaka (Propithecus uerreauxi, a Madagascar primate).

22

5

Similar genetic basis for social development and energy metabolism

Changes in gene expression in primates, which were suggested to accompany social evolutionary development include genes controlling energy metabolism (e.g. ASPM, GLUD2 and MCPH1)

23

Social deficits in bees and humans share a neural gene expression signature

Social deficits in non-social bees and Autistic Spectrum Disorder humans are associated with GABA receptor and voltage-gated ion channel genes, which is involved in allostasis 24.

25

115

116

Social Brain Development: Unexplained Findings 117

According to our framework, social animals are not born with a predetermined “social brain”, 118

but rather biologically adapt to become social as a result of allostasis dependency. In this section, we 119

will describe neural circuits involved in social processing, review evidence demonstrating that they 120

are not innate, and that early social experience effectively determines their functional outcome (as 121

well as behaviour) in adulthood. 122

What is the “social brain” and what is it for? 123

The fully developed adult human brain is organized as anatomically connected and 124

functionally coupled intrinsic networks 13. These networks are held together by thick, long range 125

axonal tracks 26. Of specific importance to both allostasis and social processing are the salience 126

network 13, and the default mode network 27 (also called the mentalizing network 13). The salience 127

and default mode networks together make up an integrated network for implementing allostasis and 128

representing its sensory consequences, called interoception 15. These two networks are considered 129

domain-general core networks in the sense that they are consistently involved in a variety of 130

psychological phenomena, including social functioning 13. These domain-general networks are 131

connected to each other and to other parts of the brain, via cortical nodes called “rich club hubs” 28, 132

which integrate information from across each network, and between the different networks 28. The 133

rich club hubs are also heavily connected to each other and to the sensory and motor networks of the 134

brain 28, and are thought to function as a high-capacity backbone for synchronizing neural activity to 135

integrate information across the entire brain 29. 136

The core intrinsic networks and hubs, specifically in the default mode and salience networks, 137

were repeatedly demonstrated to participate in social brain processing, including maternal bonding 138 30-32, social cognition 13, social network size 33, and to be impaired in patients with social deficits 139

such as in Autistic Spectrum Disorder (ASD) 34, 35. A recent meta-analysis of social neuroimaging 140

6

studies searched across the literature for distinct “neural fingerprint” for social processing. The 141

results from this meta-analysis demonstrated that neural circuits associated with social processing 142

across the neuroimaging literature are similar to domain-general circuits, namely the default mode 143

and the salience networks, which are also involved in allostasis 15, and other mental capacities (Atzil, 144

S. Satpute, A. B., Parrish, M. H., Goel, S., Shablack, H., Brooks, J. A., Kang, J., and Lindquist, K. 145

Manuscript In Preparation). Thus, evidence from human neuroimaging studies suggest an overlap 146

between the neural system that supports social behaviours, and the one that supports allostasis. 147

148

The brain is malleable in early life, and sensitive to social input 149

It was demonstrated in the past that the infant brain is not a miniature version of an adult 150

brain 36. Key aspects of adult brain functional architecture, including long-distance functional 151

synchronization and “rich club hubs” within the core networks (i.e., salience, default), are missing in 152

newborns 37-39 (see Figure 1). Consistently, from structural development perspective, the myelination 153

of the long-distance axon tracts that allow for the fast, efficient information transfer throughout the 154

networks develops for the most part after birth 36, 37, 39-41. Generally, sensory and motor-control 155

networks become synchronized early in life, even during prenatal period and show adult-like spatial 156

topology shortly after birth 41, 42. However, the core networks, mostly residing in association cortices, 157

develop more slowly over time 43. For example, it is reported that major nodes of the default mode 158

network do not become synchronized until 6 months of life 43 and that the network continues to 159

develop well into childhood and young adulthood 44. This is also consistent with the developmental 160

course of dendritic arborization and synaptogenesis which similarly show earlier peaking in primary 161

sensory cortices but prolonged growth in prefrontal and other association cortices 45. Overall, 162

empirical data suggests that humans are born without the neural infrastructure that supports adult 163

sociality. 164

7

165

Figure 1. Growing a Social Brain. This figure shows the development of the (a) default mode 166 network and (b) salience network during the first two years of life derived from a data-driven 167 independent component analysis technique. (a) and (b) demonstrate that key aspects of adult brain 168 functional architecture that support social processing, including long-distance functional 169 synchronization and rich club hubs are missing in newborns. Despite the numerous studies showing 170 that neural development depends on social experience (see supplementary Table 1), the mechanistic 171 role of social care on the developmental trajectory of the default and salience networks was never 172 directly assessed. Given the pliability of neonates’ brains, we hypothesize that the intensive social 173 care during critical years of development have a dominant role in sculpting these networks (c). (a) 174 and (b) were reproduced with permission from 46. 175

176

Human brain development is a protracted process that starts in utero and lasts for up to 25 177

years postnatal 47. During the first perinatal weeks, the brain is characterized by maximized 178

plasticity. There is a massive acceleration of brain growth, myelination, enrichment of neural 179

synapses, cortical volume expansion and cortical folding 36, 48, 49, all are critical to normal cognitive 180

development 48, 50, 51. The extended developmental course in humans, along with massive neural 181

Allostasis deviation Allostasis regulation

C

8

plasticity makes brain development susceptible to environmental input 47, 52, 53, suggesting that the 182

social environment could have a dominant role in determining the outcome- the fully formed adult 183

brain. 184

185

Early-life social care determines social behaviour in adulthood 186

It is well-established that in social animals, social interactions are critical to development 54. 187

Compelling empirical evidence shows that provision of early-life care shapes brain anatomy 55-58, 188

function and development 59-63 (for review see 64, and see supplementary Table 1 for a list of 189

evidence). In addition to brain development, early social care also determines the behavioural 190

phenotype of the offspring. The development of social-behavioural repertoire in adulthood depends 191

on early life social experience 65. Even “typical” social behaviours, like maternal behaviours, sexual 192

behaviour and peer interaction 66, would not develop without maternal care 67. However, not only 193

the complete deprivation of social care shifts the trajectory of social development. Individual 194

differences in maternal care have the power to shape social development in mammal offspring 68. 195

For example, rats demonstrate individual differences in maternal behaviour, as some dams are 196

spontaneously more maternal than others 68. Female rats that provide high levels of maternal 197

behaviours towards their pups received high levels of maternal care 69. This intuitive trajectory of 198

cross-generation transmission of maternal behaviours could rely on a genetic mechanism, i.e., the 199

dam could have genetically inherited a “high social genetic profile” from her parents. Alternatively 200

(or additionally), the mechanism could be acquired, as the dam could have been “programed” to 201

interact with her pups during her own rearing. Cross-fostering studies confirm the role of postnatal 202

experience in mediating this transmission. Females born to dams that provide low levels of maternal 203

behaviour and are fostered by dams that provide high levels of maternal behaviour, will perform 204

high levels of maternal behaviour toward their own pups 70, 71. 205

Human studies have also demonstrated that variation in maternal behaviour impacts children 206

social development 72. Children of mothers with postpartum depression, a condition that impairs 207

maternal behaviour 73, have altered social development (including long-term susceptibility to social 208

problems such as separation anxiety, and social withdrawal 74). On the contrary, children who 209

receive optimal maternal care showed improved social development along with optimal 210

physiological organization across childhood and adolescence 54, 72. A recent longitudinal study in 211

humans show that individuals who experienced sensitive, responsive, and supportive caregiving 212

9

exhibited improved bio-behavioural regulation with their adult romantic partners, 37 years later 75. 213

This study makes two important points: First, even small fluctuations in early care are powerful 214

enough to cause variation in social behaviour in adulthood. Second, the parent-infant dyad is not 215

only responsible for the cross-generation transmission of parental behaviours, but also plays a 216

general role in moulding social behaviours at large. 217

To summarize the first section, allostasis regulation is a rewarding process 12, and as such can 218

potentially motivate learning and development. In social animals, social regulation of allostasis is 219

proposed here to motivate social learning and the maturation of associative neural networks, which 220

are reportedly involved in social functioning in adulthood. We hypothesize that infants will show 221

facilitated network development when their allostatic needs are sensitively regulated. This 222

hypothesis is supported by literature demonstrating that child development is optimized and even 223

accelerated where provision of parental care is sensitively attuned to the infant needs 76. In the next 224

section, we will outline how social care in early life can potentially support social development, via 225

the acquisition of concepts. 226

227

An Alternative Framework for Social Development 228

According to our framework, infants are not born with “core social knowledge” 77, 78, but 229

rather need to learn about social agents and social behaviours. As a result, a prerequisite of social 230

development is the acquisition of rudimentary concepts, which start as multi-modal representations 231

(such as a face) and become more abstract with development. We will describe our proposed course 232

of development from birth onward, starting with the development of a conceptual system, then 233

describing how such conceptual system promotes social development. 234

235

Neural prediction- a potential mechanism for how experience sculpts the developing brain 236

An increasingly popular hypothesis in neuroscience is that the brain runs internal models that 237

function as Bayesian filters for incoming sensory input, driving action and constructing perception 238

and other psychological phenomena 16, 79, 80. This hypothesis is often called predictive coding 80-85. 239

Prediction signals (also known as ‘top-down’) are embodied, whole brain representations that 240

continuously anticipate (i) populations of upcoming sensory events from inside and outside the body 241

(ii) populations of best action to deal with those events. Unanticipated information is prediction 242

error signal that tracks the difference between the prediction and the actual incoming input from the 243

10

world and the body (also known as ‘bottom-up’ signal). Predictions are generated in agranular 244

association cortices and propagate to primary sensory cortices, always preparing for the next moment 245 86. 246

A key feature of the predictive coding model is the interaction between the forward and 247

backward flow of information: the backward flow delivers predictions while the forward flow 248

computes the residual errors between prediction and sensory inputs 84. In early life as infants’ 249

sensory pathways become intact (infants’ sensation starts in utero and continues after birth), without 250

sufficient sensory experience to form valid predictive models, most sensory input is considered 251

“prediction error”, simply because the brain cannot predict it. This idea was conceptualized by 252

Alison Gopnik as “lantern consciousness”, or how babies take in everything around them 87. With 253

experience, infants start to detect and predict sensory patterns based on co-occurrence probability. 254

This is called ‘statistical learning’ 88-90. Detecting structure within the environment is a critical step 255

in development 91 as from a meaningless stream of unpredicted sensory information, populations of 256

instances are grouped together and mentally represented as concepts 92. The infant’s experience 257

shifts from “lantern consciousness” to “spotlight consciousness”, or intentional selection of 258

perceptual input 87. 259

Of special importance for the development of sociality is neural prediction within the 260

interoceptive system, which is the sensory consequence of allostasis. Importantly, the term allostasis 261

was originally defined in terms of prediction 5. Allostasis, as defined by Sterling & Laughlin 93, is 262

“the core task of all brains to regulate the organism’s internal milieu by anticipating needs and 263

preparing to satisfy them before they arise” 93. The predictive nature of allostasis is one way in 264

which allostasis differs from the concept of homeostasis. Others have marked the importance of 265

predictions in allostasis 94-97, and interoceptive predictions about allostasis were suggested to 266

underlie decision making and motivational behaviour 95. We propose that in newborns, interoceptive 267

information about allostasis is regularly associated with exteroceptive information about caregivers. 268

This conditioning prompts the infant’s brain to regulate the internal milieu by attending to social 269

information. Deviations from physiological balance (for example low glucose levels) elicit distress 270

behaviour (for example cry), which (ideally) elicits social care and dramatically raises the chances of 271

survival 98, 99. Thus, according to our theory, for infants raised in social dyads- interoceptive 272

perception of allostasis is temporally-associated with exteroceptive perception of the caregiver (see 273

Box 1 and Figure 2). With experience, infants learn to predict about allostasis given social 274

11

information. A recent comparative study supports the importance of maternal care predictability by 275

demonstrating that when infants (humans and rats) can predict maternal sensory input, they develop 276

optimally 100. This suggests that efficient Bayesian models about maternal sensory input (computed 277

by infant's brain) promote optimal bonding and development. Accordingly, developmental settings 278

that obstruct the conditioning between the caretaker and allostasis, such as prematurity, post-partum 279

or developmental psychopathology, or orphanhood, are predicted to interfere with social 280

development. We hypothesize that with development, as top-down predictive models gradually 281

govern infants’ experience, infants’ allostasis and allostatic-independence will exponentially 282

increase. Computationally, this process might involve a gradual decrease in the salience of 283

interoceptive prediction errors. Empirical evaluation of the relationship between the predictive 284

coding approach and allostasis during development is warranted. 285

Forming Bayesian models about the social (and non-social) environment depends on temporal 286

and multi-modal contingencies in the infant brain. Multi-modal sensory inputs (exteroceptive and 287

interoceptive) are integrated in agranular association cortices, which rely on those learned 288

contingencies to generate Bayesian models and issue predictions 84. The agranular association 289

cortices, which integrate information from across the brain, become predictive “hubs” 28 (see Figure 290

2). Multi-modal integration is not evident in newborns 101. Moreover, rsfMRI reveals that core 291

networks and predictive hubs are also not traceable in newborns (see Figure 1). This supports our 292

hypothesis that the development of prediction infrastructure relies on postpartum experience that 293

potentially determines the actual multi-modal associations. 294

In the absence of multi-modal association or predictive models, we hypothesize that infants’ 295

experience mostly includes “bottom-up” information, or- prediction errors. Sensory information is 296

mostly processed in primary sensory cortices 101. This idea is supported by empirical data on brain 297

development. Thalamo-cortical connections, which deliver “bottom-up” sensory information to the 298

primary sensory cortex (e.g., prediction error), are the first to develop in the embryo 40, while the 299

cortico-cortical connections that deliver “top-down” predictions, form after birth 40, 102, and in social 300

animals are subjected to social impacts. Intriguingly, thalamo-cortical functional connections in 1-301

year olds, particularly those between the thalamus and salience network regions, show unique 302

predictive values for cognitive development outcomes at 2 years of age 103, 104. Cortico-cortical 303

connections are shaped by experience 105, and organize throughout childhood with significant 304

dendritic growth and synaptogenesis around critical periods of cognitive development 102. Multi-305

12

modal integration is not genetically predetermined, and its emergence and maturation critically 306

depend on cross-modal experiences that shape the neural circuits in such a way that is optimized for 307

the immediate environment in which the animal will function 101. 308

309 Figure 2. The anterior insula is a hub that integrates exteroceptive to interoceptive inputs. The 310 anterior insula is a part of a multi-modal integration network, which overlaps with the salience 311 network 106 (also including the dorsal anterior cingulate cortex and dorsal operculum 107). (a) 312 Interoception: ascending sensory signals from the internal milieu of the body are carried to primary 313 interoceptive cortex in the posterior insula 15, 84. The association cortex in the ventral anterior insula 314 have bidirectional connections to primary interoceptive cortex in the posterior insula, which enable a 315 bidirectional flow of information between the two regions: the ventral anterior insula sends “top-316 down” anticipated interoceptive prediction signals to primary interoceptive cortex in the posterior 317 insula, while “bottom-up” ascending sensory inputs from the body goes from the posterior insula to 318 the ventral anterior insula 84. (b) Exteroception: the anterior insula is also involved in integration of 319 sensory processing of the outside world 108, and in perception of all sensory modalities 109. The 320 involvement of this associative cortical region in both interoception and exteroception 95, 108, 109 can 321 mark its role in integration of allostatic and social information. We hypothesize that with repeated 322 care, infants learn to associate interoceptive input about allostasis and exteroceptive input about the 323 caretaker into one multi-modal experience. One example of social regulation of allostasis is feeding. 324 The feeding of a human-newborn (like all mammals) inherently involves a caretaker. With every 325 feeding, the caretaker regulates the infant’s glucose levels. The newborn’s experience of feeding 326 conditionally comprises interoceptive digestive information (e.g., glucose levels), and exteroceptive 327 information about the caretaker (e.g., caretaker’s face, smell). There is a temporal conditioning 328

(a) Interoception of allostasis- via primary interoceptive cortex in Posterior Insula (PI).(b) Exteroception of mommy- via tactile, visual, auditory, olfactory sensory Cortices.

Anterior Insula (AI)-this neural hub integrates sensory projections from primary exteroceptive and interoceptivecortices into a multi-modal representation of mommy: a neural model that associates social information with allostasis. With experience, the anterior insula is potentiated to predict allostatic changes 84

based on social stimuli.

Prediction signalsPrediction error signals

PI

AI

13

between a caretaker and allostasis, which will be processed by the infant as one multi-modal 329 experience. We hypothesize that based on the temporal conditioning between social and allostatic 330 information, infants imbue social information into allostatic processes, and the association cortex 331 that integrates the multi-modal input learns to predict about allostasis, based on social input. To 332 prepare for the next feeding, the presence of the caretaker will already trigger the infant’s anterior 333 insula to issue predictions about upcoming changes in glucose levels. The predictions will propagate 334 to the posterior insula, and other brain regions that assist to prepare for the upcoming allostatic 335 changes, and achieve stability through physiological (e.g., insulin release) or behavioural (e.g., 336 crying, sucking, salivating) changes. 337

338

Notably, agranular association cortices, which issue most neural predictions 84, such as the 339

anterior insula (AI), anterior cingulate cortex (ACC), and ventro-medial prefrontal cortex (vmPFC), 340 84 are commonly considered major nodes of the “social brain”, and were linked to social 341

competencies like mental inference, empathy and person perception 13, 110-112. The same brain 342

regions, which are cortical “rich club hubs” of the salience and default mode networks, are also 343

suggested to regulate the autonomic nervous system, the immune system and the neuroendocrine 344

system, as part of a predictive allostasis regulation neural system 15, 16, 84, 85, 95, 113. Moreover, the 345

amygdala, nucleus accumbens and hypothalamus, which are also considered key regions in social 346

processing 114, have a key role in allostasis regulation 113 and are thought to compute prediction error 347

and motivate behaviour 115. We hypothesize that these regions’ involvement in social processing 348

reflects an underlying process that occurs simultaneously: preparing the organism for upcoming 349

changes in allostasis. 350

In support of our hypothesis, the association hubs that are involved in social processing, are 351

not exclusive to social processing. While social information is very useful for allostasis prediction, 352

other types of “non-social” information (for example, food) can also be useful to predict allostasis. 353

This would explain the consistent involvement of these association cortices and limbic regions in 354

general affective experiences not necessarily related to social experience (for meta-analysis see 116). 355

According to our framework, consistent provision of social care can impact neural plasticity and 356

promote neural associations between these regions into “core” large-scale networks that implement 357

an acquired system for the purposes of allostasis. Thus, the “social brain” is really the predictive 358

brain, which develops as a function of social experience aimed at allostasis regulation. 359

360

Neural prediction supports the development of a conceptual system 361

14

When a brain is “processing information” to construct perceptions and plan action, it is asking 362

‘what is this new sensory input most similar to?’ relative to situated, past experiences (see 117, 118). 363

The dis/similarity of the current sensory array is computed with reference to the past to estimate the 364

potential energy costs and rewards for the body. That is, a prediction is considered a partially 365

completed category that is used to classify incoming sensory signals to construct sensory perceptions. 366

Similarly, in cognitive science, a category is a population of events or objects in the world that are 367

treated as similar because they all serve a particular function or goal in some context. The mental 368

representations of categories are concepts 119. In effect, then, it has been proposed that when the 369

brain assembles populations of predictions, it is constructing concepts 16, or what L. Barsalou refers 370

to as ‘ad-hoc’ concepts 120, 121 (See Figure 3). It was previously established that prediction errors 371

prompt learning because they modify the future predictions to incorporate the new information 86, 122. 372

A concept, then, represents a group of sensory predictions (See Figure 3), and Concept learning is 373

the encoding of sensory prediction errors 106. In this view, every event of new learning (e.g. the 374

processing of prediction error) is categorized into a concept. 375

376

377

Figure 3. Concepts as predictions. Similar ideas about information processing are described using 378 different terminology in the predictive coding and cognitive science frameworks. At the level of 379 perception, actual incoming sensory input from the world and the body (also known as ‘bottom-up’ 380 signal in cognitive science) is considered in the predictive coding framework as prediction error. At 381 the next level of processing, information is organized into ‘categories’ (cognitive science), or 382 “probabilistic models” (predictive coding). At the higher level of processing, the mental 383 representations of categories are ‘concepts’ 119. Just like concepts, predictions are also whole brain 384

15

representations that organize new incoming sensory input 85. Importantly, in the predictive coding 385 framework, the organization of sensory input is a predominately top-down process that starts with a 386 probabilistic model. The model incorporates new sensory data only when it is not predicted (i.e., 387 prediction error), which in turn updates the model 84. 388

389

Allostasis-driven learning of social concepts 390

According to our framework, the first step in social development is acquisition of 391

rudimentary social concepts. At early infancy, the infant gains experience interacting with the 392

caretaker, and most interactions will be implicitly or explicitly aimed towards allostasis regulation. 393

With consistent on-going care, infants detect the statistical regularities of their social environment 90. 394

For example, by repeated gazing at the mother, infants will gradually recognize the spatial 395

organization of her facial features and form a rudimentary concept of a face. The brain categorizes 396

sensory information to predict about allostasis, and thus we propose that sensory regularities (such as 397

a face) will become concepts more rapidly if they impact allostasis. The association between 398

allostasis and a human agent will result in learning an important social concept: mommy (similarly, 399

this concept can be daddy, or any other caretaker and will be referred to as the mommy concept 400

regardless of the primary caretaker gender/s or familial relation) (See Figure 4 and Box 1). 401

402

Box 1. A predictive coding approach to social development: Infants learn social concepts as 403

generative models for allostatic regulation. 404

According to models of predictive brain function 80-85, the brain constantly constructs multi-modal123 405

hypotheses about the world (i.e. predictions) and then samples evidence from the environment that 406

are measured against these hypotheses (i.e. prediction errors) 16, 79, 80. 407

A common mathematical formulation of these dynamics involves Bayesian inference, where 408

the brain’s hypotheses regarding the environment can be generally described by: 409 = + ( − ) ∙ (1)

410

denotes the brain’s prior prediction, denotes the sensory signal, such that the brackets represent 411

the prediction error, and denotes the brain’s updated hypothesis. denotes the weight of the 412

prediction errors, often called precision, which is a function of the ratio of the uncertainty of the 413

prediction, and the uncertainty and saliency of the prediction error 84, 123. Thus, prediction errors are 414

given a lesser weight when the predictions are relatively certain. 415

16

416

In the allostasis framework, the ultimate goal of the brain’s predictive models is to optimize 417

physiological demands and gain 84. When a caregiver impacts the infant’s allostasis, she/he 418

effectively minimizes interoceptive prediction errors. With experience, the infant learns that 419

exteroceptive cues regarding the caretaker (e.g. smell, voice) predict allostatic regulation. Consistent 420

care thus creates relatively certain predictions regarding allostatic regulation that reduce the 421

precision of interoceptive prediction errors. Then, exteroceptive cues regarding the caretaker are 422

given a higher importance than cues not impacting allostasis. At higher levels, this multimodal 423

information is integrated in the infant’s mind in the form of a concept (See figure 3 and section 2.c), 424

which can be formalized by: 425

(′ ′| , )∝ (′ ) ∙ ( | ) ∙ ( | ′) (2)

426

Corresponding with Bayesian models of the brain, Equation 2 shows that the infant’s concept of 427

mommy is conditional on interoceptive and exteroceptive information. This concept will get more 428

robust and reliable (i.e. with lower uncertainty) as the caretaker’s behaviour affecting the infant’s 429

interoceptive and exteroceptive sensations involves statistical regularity and consistency. Learning 430

occurs as the concept of a mommy given these inputs: 431 [ (′ | , )], becomes the exteroceptive and interoceptive 432

multi-modal prediction regarding the caregiver: [ (′ )]. When caretakers are not successful 433

in social regulation of allostasis (e.g. due to potential parent-related illness such as post-partum 434

depression), statistical regularity between the exteroceptive input about the caregiver and the 435

interoceptive input about allostasis is impaired. According to the framework suggested here, we 436

predict that this will cause a developmental impairment in acquiring social concepts. 437

438

17

439

A schematic Bayesian model of organizing social information in infants. Infants associate 440 exteroceptive and interoceptive information to form social concepts, like mommy. The social concept 441

18

of mommy represents a computational predictive model: Based on previous experience, an 442 association between exteroceptive information about the caregiver and interoceptive information 443 about allostasis is made, and the brain can issue predictions (downward blue arrows) regarding 444 upcoming allostasis changes using social information and vice versa (to make social predictions 445 based on interceptive information). For example, by the age of a few days old, infants already gained 446 repeated experience in feeding, and learned that exteroceptive information about mommy is useful to 447 predict about upcoming changes in their plasma glucose levels (for more detail, please see Figure 2). 448 Upward red arrows represent prediction errors used to update the brain’s model and predictions. For 449 simplicity, we present here only two examples of exteroceptive modalities, and one example of 450 interoceptive input, whereas a full model incorporates a range of interoceptive and exteroceptive 451 modalities. We propose that the brain acquires and uses such concepts (social and others) to optimize 452 information processing, with the on-going ultimate goal of allostasis regulation. 453 454

455

Figure 4. The relationship between allostasis, concepts, and social development. 1. Allostasis 456 is an elementary process needed for survival and is thus the first order of the diagram. 2. 457 Concepts- to prepare for allostatic demands, the brain uses past experience as a model, and 458 organizes incoming sensory input into mentally represented concepts. Organization of sensory 459 input is governed by allostasis, and underlies social development, and is thus the second order of 460 this diagram. 3. Social development- in social species, allostasis is (by definition) regulated in a 461 social environment. Consequently, social animals first and foremost learn social concepts. 462 Allostasis-driven learning is rewarding, and that promotes social bonding. Social development 463 relies on allostasis-driven concepts and is thus the higher order of this diagram. 464

465

466

Social Development: In humans, newborns’ allostasis is always regulated by acaretaker. Therefore, the sensory patterns that are relevant for newborns’ allostasisare social, and infants construct rudimentary social concepts such as ‘face’ and‘mommy’. Social concepts are conditioned with positive outcome of allostasisregulation. Through allostasis, social care becomes rewarding, and the infantattaches to the caretaker, and becomes increasingly driven towards socialinteractions. Gradually, more complex social and non-social information, skill andbehaviours are learned.

Concepts: To prepare for allostatic demands, the brain organizes environmentalsensory input into concepts. During development, infants detect statisticalregularities in the environment and from a meaningless stream of information,populations of instances are grouped into mentally represented concepts. Sensorypatterns relevant for allostatic will efficiently becomes concepts.

Allostasis: is the elementary ongoing process of optimizing the body’s internalmilieu. The brain is constantly regulating allostasis, while predicting allostaticneeds, and adjusting behaviour and metabolism towards survival, growth, andreproduction.

19

Allostasis-driven learning of social competencies 467

Through social regulation of allostasis, a child experientially acquires not only social 468

concepts (like a face or mommy), but also social competencies. One of the basic social competencies 469

infants gain is synchrony. Bio-behavioural synchrony is an important aspect of mother-infant 470

attachment 54, and was shown to be important for shaping optimal developmental outcomes of 471

physiological regulation, executive functions, and social aptitude 124. 472

Here, we consider synchrony as one efficient strategy for social regulation of allostasis. 473

Starting from gestation, a mother controls her foetus’s allostasis via mother-foetus physiological 474

synchronization 7. After birth, mammalian mothers continue to regulate the infants’ allostasis 125, 126 475

using the same strategy. Mothers regulate their infants’ temperature by holding them close so that 476

their temperatures synchronize 125. Mothers regulate their infants’ immune function by breastfeeding, 477

synchronizing their gut-microbiota and antigen-specific antibodies 127. Mothers regulate infants’ 478

arousal with voice (by singing, or speaking loudly or softly) 128, synchronizing their heart rates 10. 479

Within a healthy dyad, the infant quickly gains a lot of experience in synchrony, and will 480

progressively learn to willingly synchronize. Learning synchrony is one of the first social 481

competencies that infants acquire, as within several months the infant not only learns to intentionally 482

synchronize with others in order to regulate their own allostasis, infants will also start using 483

synchrony to intentionally impact others’ allostasis. 484

Amongst synchrony, another fundamental social competency infants acquire is joint attention 485 129. Infants learn conceptual knowledge by synchronizing their attention with others. Attention is 486

defined as a neural computation that biases certain features out of competing environmental 487

information 130. Attention is a learned cognitive skill that is plastic and shaped by experience 130. 488

Once attention develops, and the infant learns to synchronize, they learn the mental ability of sharing 489

their attention with a caretaker at around six months 130. Using joint attention, caretakers direct the 490

infant’s statistical learning towards cultural and social relevant information. Joint attention is a 491

precursor for additional social competencies, like ‘theory of mind’ 131. However, joint attention is a 492

social competency that is fundamental not only for the development of social cognition, but to many 493

aspects of cognitive development 132, as within the medium of joint attention with caregivers and 494

peers, infants are introduced to all knowledge and proficiency needed to survive in their 495

environment. Mothers explicitly and implicitly teach infants new concepts as they influence the 496

20

spontaneous statistical learning by providing regulatory social input, such as vocalizations 133, gaze 497 134 and touch 135. 498

Learning is an inherent aspect of allostasis, as the term ‘allostasis’ explicitly incorporated 499

learning and predictive responding 5, 84, 136. It was previously demonstrated that attaining 500

physiological stability reinforces learning 12, and that physiological manipulation (e.g., deviation 501

from allostasis, for example hunger) motivates learning 137, 138 and prosocial behaviours in social 502

animals 139, 140. Sensory input with allostatic implications (that is, likely to impact survival, offering 503

reward or threat) will be learned, so as to support allostasis better in the future 15, 141. Accordingly, 504

we hypothesize that stimuli with higher predictive value for allostasis will be learned quicker than 505

stimuli with lower impact on allostasis. During development, infants learn social concepts and skills 506

to prepare for allostatic needs, as caretakers introduce all the culturally relevant concepts, using 507

language (see Figure 4). 508

509

Allostasis-driven development of abstraction 510

Allostasis driven learning marks a special case for humans, comparing to non-human 511

animals. All animals depend on allostasis, and engage in allostasis-driven learning (including social 512

learning). However, humans, on average, have a brain that is three times larger than a chimpanzee 513

brain 142, 143. While humans and chimpanzees have somewhat comparable sensory and motor 514

networks, in humans these networks are connected to an expanded core brain system of association 515

cortices 144, which could imply an advanced capacity for multi-modal integration. These were 516

suggested to evolve in order to sustain the relatively complex demands of human social niche 145. 517

The advanced capacity for integration, could underlie human ability for abstraction. Among non-518

human mammals, allostasis driven learning is limited to sensory concepts with immediate physical 519

impact on allostasis (such as food or pain). Humans can construct high level abstract concepts (for 520

example money, love, pride, god and other cultural concepts) and link them to allostasis (e.g., have 521

bodily representations of abstract ideas). Among “socially entrained” humans, a word or an idea 522

could be sufficient to regulate or disturb allostatic balance. The link between abstract entities such as 523

words for physical regulation of allostasis could underlie language acquisition in humans 146. It has 524

been previously suggested that only as children develop social competencies, they can learn to 525

understand the meaning of abstract concepts 147. We hypothesize that social bonding promotes 526

abstraction because social dyads provide the medium in which abstract concepts become meaningful 527

21

by physically linking them to allostasis. Ad-hoc investigations about allostasis regulation and social 528

learning and abstraction in human infants are warranted. 529

According to our framework, during a critical developmental window, not only spoken 530

language is acquired, but also culture. It was demonstrated in comparative studies that longer 531

developmental courses favour extended social learning 148-151. As a result of human ontogeny, a 532

human-newborn’s brain is highly immature, and thus fellow-humans contribute to infant allostasis 533

and brain development for an extended period during development 145. This allows social input and 534

culture to wire the brain in a more extensive way in humans than in any other animal. Individuals 535

tend to conform to certain cultural schemes and synchronize according to a collective set of 536

concepts, including norms, beliefs and manners 152. The caretakers, and other members of the group, 537

are responsible for cross-generation transmission of such relevant cultural knowledge. As part of the 538

cultural scheme, and reinforced by allostasis, mentally rich human cognition is acquired. In addition 539

to social learning, allostasis-driven learning can also shape other human features such as cognition, 540

emotion and culture 16, 153. Empirical evidence suggests that mental and cognitive capacities (like 541

emotion, sociality and cognition) are not universal 154, 155 and deep cultural differences exist between 542

groups. According to our proposed framework, emotion and social concepts are environmentally 543

constructed in each culture, and transferred between generations in social dyads during early life 544

social “training”. A smooth roll out of such a mechanism can ensure efficient cross-generation 545

transmission of the deepest aspects of human emotion, cognition and sociality, while staying flexible 546

and robust in face of change. 547

548

Summary: Social, Cognitive and Neural Developmental Trajectories are Temporally linked 549

Despite evidence that social experience affects both neural and cognitive development, the 550

structure-function developmental mechanism remains poorly understood. Characterizing the neural 551

mechanisms of cognitive development is a pending empirical task. However, several neural 552

developmental milestones temporally match the development of new cognitive skills. Of specific 553

interest to social cognition is the temporal contingency between the developmental trajectories of the 554

default mode network and of cognitive abilities like conceptualization 13. The default mode network 555

is believed to construct mental representations of concepts 156, including complex representations 556

about other people’s minds (e.g. Theory of Mind) 157. The adult default mode network cannot be 557

functionally traced in newborns 38, 42. The default mode network starts from an isolated posterior 558

22

cingulate region that can be traced in neonates, and evolves to becoming a synchronized network 559

later in life 38. More precisely, most of the core nodes of default mode network become synchronized 560

by 6 months of age, making default mode network among the first domain general networks to 561

achieve qualitatively adult-like spatial topology 158. It has been demonstrated that the connectivity 562

and volume of the major default mode network nodes (the posterior cingulate cortex and medial 563

prefrontal cortex) are immature at birth 159. The grey matter volumes as well as functional and 564

structural connectivity in the default mode network continue to develop during childhood, reaching 565

full maturity in late adolescence 42, 160, right when social cognition abilities and mentalizing mature 566 160, 161. Intuitively, it has been postulated before that changes in brain connectivity might be linked to 567

cognitive development 103, 162, 163, and that the absence of a default mode network in early infancy 568

indicates the absence of cognitive skills like conceptualization and Theory of Mind 42. Accordingly, 569

default mode network maturation is potentially a crucial developmental step, which serves as the 570

precursor for social and cognitive development. Importantly, network plasticity seen during 571

development is not sufficient to conclude the lack of an inborn social system. A circuit or network 572

can be innate and emerge only with experience. Alternatively, it is possible that a capacity, such as 573

sociality, can be a by-product of something else that is innate, like allostasis. The degree of plasticity 574

in brain network connectivity during the early years of life, combined with the importance of 575

allostasis for reinforcement learning, together suggest that researchers should reconsider the common 576

assumption that a network or system for sociality is inborn. Empirical investigation of mechanisms 577

facilitating the impact of dyadic care on brain and social development is warranted. 578

The literature on social development in rodents provides robust support for a mechanistic role 579

social care has on brain development, and specify hard mechanistic evidence for how parental care 580

physically controls brain development in the offspring (including receptor expression, plasticity and 581

cortical folding, see supplementary Table 1). By integrating cognitive and neural developmental 582

literatures, we hypothesize that social care controls social and cognitive development via maturation 583

of whole-brain neural networks. Specifically, we propose that parental care, which is consistently 584

reinforced by allostasis, is necessary for the infant to build and refine a multi-sensory mental 585

representation of concepts 92, 147. Starting from first simple multi-sensory concepts, such as a face, 586

and then mommy, infants will gradually learn to represent more abstract concepts, including words 587

and ideas by linking them to allostasis via dyadic interaction. The developmental neural shift from 588

primary motor-sensory circuitry seen in newborns, to association networks in adults, suggests a 589

23

potential dramatic shift in the development of human experience: from undefined raw sensory 590

experience in early infancy, to constructed cognition in adulthood. 591

592

Discussion: Growing a social brain is highly adaptive and promotes affiliation 593

In this perspective paper, we present a theoretical framework for social affiliation. 594

Specifically, social affiliation is learned and allostasis is the incentive. Social attachment could be 595

the result of one evolutionary feature: helplessness in achieving and maintaining physiological 596

stability. Such social co-dependency creates an ultimate driving force for attachment and learning. 597

The mother-infant dyad is a social “boot camp” for learning social affiliation, which is extremely 598

efficient since infants’ lives depend on their caregiver. While caretakers teach newborns the 599

behavioural repertoire of sociality (including bonding, but also more complex social behaviours as 600

cooperation, competition, aggression, etc.), they also program their biology, helping them to “grow a 601

social brain”. Infants will learn to identify humans as important, and to synchronize with them. They 602

continue to learn through synchronizing conceptual knowledge and mental abilities with their 603

caregivers. The caregiver explicitly and implicitly teaches the newborn a saliency road map of the 604

world. Gradually, infants are trained not only to regulate their own allostasis, but will become 605

experts in decoding and attending to other people’s allostasis, as they become socially functioning 606

adults. However, not only social development relies on social regulation of allostasis, but also other 607

cognitive and emotional developmental trajectories. During childhood and within social dyads and 608

other members of their social group, such as family members and friends, infants are trained to 609

acquire all skill and knowledge needed as adults. Social care is therefore not merely responsible for 610

shaping a “social brain”. We propose that social care is needed to grow A brain. Consequently, the 611

entire brain, which is potentially sculpted by ongoing social transactions, can be thought of as a 612

social brain. Human brains are transactive and cannot be considered outside the context of other 613

human brains. Transactive brains form a collective and flexible system that sustains many of our 614

human features, including knowledge 164, skill and biology. 615

Early life is a critical window for social learning due to the acute nature of the social 616

dependency. As long as children lives depend on social communication, their brains attend to learn 617

the complex social information in order to survive 165, 166. As such the social dyad catalyses social-618

related statistical learning. With maturation, the “life-or-death” motivation for social learning 619

gradually subsides, and the window of immense learning narrows. With that, social regulation of 620

24

allostasis is not limited to early life and there is evidence from rodents and apes (including humans) 621

for a Social Buffering effect (a regulatory social impact on an individual’s physiology and behaviour) 622

on different types of allostatic processes, including HPA activity, oxytocin levels, affect and immune 623

function (for review see 167). In highly social species such as humans, adults regulate each other’s 624

allostasis as well 168, 169. As such, socially-mediated learning is a life-long process 170, 171. 625

The potentially crucial role of social experience during infancy, in shaping brain and social 626

development, suggests that social animals do not necessarily rely on a predetermined specialized 627

brain system to support affiliation 172, 173. Instead, according to our framework, domain general 628

neural systems implement a conceptual system to regulate allostasis, and that underlies social 629

behaviour. Moreover, since cultural changes evolve much faster than natural selection does, 630

evolution is more likely to select for flexible biological systems that are robust to unexpected 631

environmental change 174. A transactive human brain growing in the context of other human brains 632

and genetically predisposed to wire itself to the environment 175, is just such a system. Thinking 633

about bonding and social development as temporal conditioning between social information and 634

allostasis is a hypothesis. This hypothesis is based on the integration of literature on allostasis and 635

bonding, but also on learning. For example, a recent review about social learning in non-human 636

animal models examined experimental work in animal models of learning, and concluded that 637

domain-general associative mechanisms that track the predictive relationships and contingency 638

between two salient events, mediate learning about both social and asocial cues. Moreover, the 639

authors conclude that while associative learning mechanisms are genetically inherited, at least in 640

non-human animals there are no mechanisms dedicated to social v. asocial learning 176. This supports 641

our idea that any stimulus that conditionally impacts allostasis (e.g., salient) will be learned, and that 642

in social species, these stimuli are prominently social. 643

This perspective paper introduces a theoretical framework about how the developing brain 644

respects the physical and social surroundings as wiring instructions, along with a series of 645

hypotheses about the early allostatic experiences with caregivers as one source of these wiring 646

instructions. Our framework sets the ground for new research and ad-hoc empirical evaluation 647

targeted at allostasis, and its mechanistic role in the development of social affiliation. It also provides 648

a different interpretation to existing findings and not everyone will agree. Particularly, disentangling 649

environmental from genetic effects is challenging, and further empirical investigations on the role of 650

social care on child brain and behaviour development are warranted. While we predict that social 651

25

experience is a major determinant in social development, genetic predisposition is not over-looked. 652

We distinguish between genetic predisposition of a socially pre-wired brain (limited to domain-653

general processes, like allostasis dependency), vs. genetic determinism of a hard-wired brain (coding 654

for detailed social behaviours or neural circuits). For example, the cross-fostering studies from the 655

Meaney group 59, 70, 71, 177, 178 as well as the Lorenz imprinting studies 179, 180 suggest that social 656

behaviour does not depend on hard wired social “programs”, and that particular phenotypes are 657

flexible. Unlike some studies, which suggested that social knowledge is hardwired in our genes and 658

that newborns have innate social preference and behaviour 181, 182, others have resisted this idea and 659

provided evidence that there is no social preference in newborns 183-189. It is possible that newborns’ 660

behaviours that are typically interpreted as ‘social’ are in fact ‘allostatic’ at first and only become 661

social once infants learn to associate social information with allostasis. Accordingly, innate 662

individual differences that are seen in infants relate to allostasis 190, 191, and the developmental 663

literature applies the term temperament 192 to describe traits like ‘irritability’, ‘positive affect’ and 664

‘activity level’ 193. This supports the idea that infant’s genetic predisposition is limited to domain 665

general processes of allostasis, which interact with social experience to determine social 666

development. Synthesizing human and non-human evidence about allostasis, bonding, learning and 667

brain function, establishes considerable doubt in the common view (attributing a dominant genetic 668

account to social phenotype) and encourages research to test the alternative hypothesis suggested 669

here (by which the social phenotype is shaped by social experience, driven by the genetic 670

predisposition of allostasis dependency). 671

The framework suggested here could have implications for research in the field of 672

neuroscience because it assumes that neural circuits that support sociality are not inborn. For 673

example, across evolution, social affiliation depends on cortico-striatal (or anatomically 674

homologous) pathways 165. Accordingly, recent human fMRI studies have demonstrated that social 675

affiliation depends on the neural association between domain-general circuits of the mesolimbic 676

system and the default mode network 32. The dopaminergic mesolimbic system has been shown to 677

underlie allostatic processes 194-196, while the default mode network is important for 678

conceptualization 197. Thus, social affiliation relies on the connectivity within a multi-function neural 679

system supporting allostasis and conceptualization. 680

Considering that the “social brain” is not an isolated module, and that the entire brain is 681

wired with respect to the social environment, could have implications for psychopathology. For 682

26

example, current studies of ASD focus on boosting the “social brain” with “social drugs” 198, yet to 683

date there are no successful pharmacological interventions aimed at the core social deficits of ASD 684 199. Instead, it was previously suggested that social impairments seen in ASD are attributed to 685

domain general processes, such as deficits in organization of sensory input into efficient predictive 686

models 200. Patients with ASD show deficits in categorizing information into concepts, manifested as 687

atypical language acquisition or naming, which are accompanied by weaker connectivity of 688

association networks 201. Similarly, learning disorders, such as dyslexia, were suggested to be 689

computationally understood as a deficit in integrating prior information (e.g., forming efficient 690

prediction models) with noisy observations (e.g., prediction errors) 202. In dyslexia patients, an 691

atypical ratio between the a-prior model and the noisy input, may accounts for patients’ perceptual 692

deficits 203. A brain with limited ability to use past knowledge for prediction, is also limited in 693

conceptualization, and thus social bonding. Empirical assessments for this potential mechanism are 694

called for. Moreover, in cases of postpartum depression, where mothers are avoidant, or cases of 695

postpartum anxiety, where mothers are intrusive 30, infants’ allostasis is constantly dis-regulated. 696

Distorted allostasis regulation can shift infants’ developmental trajectories, including neural, social 697

and cognitive development 204. A clinical evaluation of infants’ allostasis regulation within mother-698

infant dyads could support diagnosis and treatment for both mothers and infants, and help determine 699

the developmental prognosis in maladaptive dyads. 700

Finally, acknowledging the environmental impact on social development has implications for 701

societal issues such as race, family structure, and religion. For example, treating mommy as a learned 702

concept, and not a biological imperative, makes a baby born to a family of two fathers not deprived 703

of any natural necessity 205. As a society, we construct many abstract concepts, which are powerful 704

because they impact allostasis. Completely abstract ideas like god, race, money or love, materialize 705

to become concrete (i.e. have immediate allostatic implications), powerfully motivating human 706

behaviour. This is because via social interactions humans learn to link those abstract concepts to 707

their allostasis to survive and prosper in their culture. This can potentially explain how beyond the 708

immediate dyadic bond with the caregiver, extended social effects, including social class or 709

economic status may carry powerful effects on child development 16, 206, even brain development 207. 710

Future research on the role of allostasis as part of a very complex control system of social behaviour, 711

and on familial and extra-familial impacts on biology and behaviour of child development is 712

27

warranted. Moreover, realizing the powerful potency of programing our children’s brains and 713

concepts can impact education, and raise the issue of social and cultural responsibility. 714

“Growing a social brain” is at the basis of every human’s well-being. Early infancy is a 715

critical time for establishing the biology of a healthy mind. However, the brain is plastic throughout 716

life and so are many aspects of behaviour and cognition 208. The human “social brain” is not a pre-717

determined organ, but rather a plastic product of ongoing acculturation. As humans, taking 718

responsibility on shaping our (social) brains can potentially impact science, societies, our children’s 719

education. 720

721

Author contribution: S.A., W.G. I.F. and L.F.B. contributed to writing the manuscript. 722 723 Acknowledgements: We thank Kfir Toledano for his contribution to the illustrations. 724 725 726

727

Supplementary Table 1: Evidence showing that post-partum social experience shapes brain 728

development. 729

Level of Observation

Evidence References

Receptors Expression

The adult female offspring of increased licking/grooming rat mothers show increased Oestrogen Receptor alpha (ERα) expression in the medial preoptic area (MPOA) and are themselves high licking/grooming mothers. Moreover, the biological offspring of low licking/grooming mothers fostered at birth by high licking/grooming dams show increased ERα expression in the MPOA.

59

The adult offspring of increased licking/grooming rat mothers have increased glutamate receptor expression in the hippocampus

209

Offspring of high licking grooming rat mothers have reduced plasma adrenocorticotropic hormone (ACTH) and reduced corticosterone responses to acute stress. In the brain, they have increased glucocorticoid receptor expression in the hippocampus, and decreased corticotrophin releasing hormone expression in the hypothalamus.

210

Maternal care permanently alters pups' expression of gamma-Aminobutyric acid (GABA) benzodiazepine receptor in the medial and lateral amygdala, medial prefrontal cortex and hippocampus. Offspring of high licking/grooming rat mothers have increased expression of GABA benzodiazepine receptor and reduced stress reactivity.

211

The adult male offspring of increased licking/grooming rat mothers show increased amygdala vasopressin 1a receptor binding.

212

28

The adult female offspring of increased licking/grooming rat mothers show increased oxytocin receptor binding in the central nucleus of the amygdala and bed nucleus of the stria terminalis.

212

Epigenetics

Maternal behaviour increases glucocorticoid receptor expression in the offspring by epigenetic modulations of increased histone acetylase transferase activity, histone acetylation and DNA demethylation.

213

Maternal behaviour in early life controls pup’s long-lasting gene expression and by that “programs” her offspring function of certain receptors, hormones and neurotransmitters.

69, 178, 214-217

Post mortem analysis of human suicide victims who had been abused as children show they had more methyl groups on the glucocorticoid receptor gene, and decreased levels of glucocorticoid receptor mRNA in the hippocampus than did suicide victims who were not abused.

60

Offspring of high licking grooming rat mothers have elevated dopamine receptor mRNA levels within the nucleus accumbens and increased conditioned place preference for a high-fat diet.

61

The adult female offspring of increased licking/grooming rat mothers show increased oestrogen receptor alpha and beta mRNA levels within the MPOA of the hypothalamus.

62

The adult offspring of increased licking/grooming rat mothers show increased oxytocin receptor mRNA levels within the MPOA

62

Neural Plasticity

Brain-derived neurotrophic factor (BDNF) and Nerve growth factor (NGF) are neurotrophines and key players in brain development and plasticity. Rodents and non-human primates’ studies show that neurotrophines are sensitive to stress of maternal separation.

218

Maladaptive mothering alters neurotrophine levels, which impacts stress reactivity in adulthood.

219

In rodents, rich social experience in infancy, with both maternal and peer interactions, increases neural plasticity, levels of BDNF and NGF in the hippocampus and hypothalamus, and the amount of social behaviours through adulthood.

220

Postnatal sensory experience controls cortical development through activity-driven BDNF expression

Typical, expected, postnatal experience is necessary for the emergence of normal patterns of neocortical organization. Animals reared in complex environments show enhancement in density of cortical synapses, increases in the number of brain support cells, and even augmentation of the complexity of the brain vascular system.

221

Cortical Growth and Folding in Newborns

Accelerated cortical folding in the perinatal weeks depends on social auditory stimuli, and human language.

222

Development of the auditory cortex in preterm infants will resemble their full-term infant counterparts if their incubators are equipped with recordings of their mothers’ voices.

63

Early experience with maternal voice has enduring effects on the developing auditory system and autonomic nervous system, with later social and emotional developmental implications.

223

29

Human Neuroanatomy

Adults who experienced maltreatment as children showed significant gray matter volume reductions in the medial frontal gyrus, dorsolateral prefrontal cortex and anterior cingulate gyrus compared to their healthy counterparts, superior frontal gyrus and orbitofrontal cortex.

55

Childhood maltreatment and abuse impact hippocampus volume and development in adulthood.

57

Maternal support in early childhood predicts larger hippocampal volumes at school age.

58

Positive parenting predicts structural development of adolescent amygdala (reduced growth) and accelerated thinning of the orbitofrontal cortex and anterior cingulate cortex.

56

Large-Scale

Brain

Networks

Function

Intrinsic networks like the default mode network, salience or control networks are not traceable at birth. They develop over years, while children grow within a social environment.

38, 41, 42, 44,

159

Children with social-developmental deficits, such as in Autistic Spectrum Disorder, have aberrant default mode network connectivity is linked with severity of social impairments.

224

Network analysis of resting-state fMRI and DTI tractography in ASD show reduced local and long-range functional connectivity. Moreover, the developmental trajectory of structural networks in children with ASD is altered.

225

730

References 731

732

1. Rand, D.G. & Nowak, M.A. Human cooperation. Trends in cognitive sciences 17, 413-425 733 (2013). 734 2. Johnson, Z.V. & Young, L.J. Neurobiological mechanisms of social attachment and pair 735 bonding. Current opinion in behavioral sciences 3, 38-44 (2015). 736 3. Hawkes, K. Grandmothers and the evolution of human longevity. American Journal of 737 Human Biology 15, 380-400 (2003). 738 4. Dunbar, R.I. & Shultz, S. Evolution in the social brain. Science 317, 1344-1347 (2007). 739 5. Sterling, P. Allostasis: A model of predictive regulation. Physiology & Behavior 106, 5-15 740 (2012). 741 6. Atzil, S. & Barrett, L.F. Social Regulation of Allostasis: Commentary on "Mentalizing 742 Homeostasis: The Social Origins of Interoceptive Inference" by Fotopoulou & Tsakiris. 743 Neuropsychoanalysis, 1-24 (2017). 744 7. Rao, P.N.S., Shashidhar, A. & Ashok, C. In utero fuel homeostasis: Lessons for a clinician. 745 Indian Journal of Endocrinology and Metabolism 17, 60-68 (2013). 746 8. Winberg, J. Mother and newborn baby: mutual regulation of physiology and behavior—a 747 selective review. Developmental psychobiology 47, 217-229 (2005). 748 9. Hofer, M.A. Hidden regulators in attachment, separation, and loss. Monographs of the 749 Society for Research in Child Development 59, 192-207 (1994). 750

30

10. Feldman, R., Magori-Cohen, R., Galili, G., Singer, M. & Louzoun, Y. Mother and infant 751 coordinate heart rhythms through episodes of interaction synchrony. Infant Behav Dev 34, 569-577 752 (2011). 753 11. Feldman, R., Eidelman, A.I., Sirota, L. & Weller, A. Comparison of skin-to-skin (kangaroo) and 754 traditional care: parenting outcomes and preterm infant development. Pediatrics 110, 16-26 755 (2002). 756 12. Keramati, M. & Gutkin, B. Homeostatic reinforcement learning for integrating reward 757 collection and physiological stability. eLife 3, e04811 (2014). 758 13. Barrett, L.F. & Satpute, A.B. Large-scale brain networks in affective and social neuroscience: 759 towards an integrative functional architecture of the brain. Current opinion in neurobiology 23, 361-760 372 (2013). 761 14. Atzil, S., et al. A social Dimension to the Affective Circumplex. 762 15. Kleckner, I., et al. Evidence for a Large-Scale Brain System Supporting Allostasis and 763 Interoception in Humans. bioRxiv (2017). 764 16. Barrett, L.F. How emotions are made (Houghton Mifflin Harcourt, 2017). 765 17. Gao, W., Lin, W., Grewen, K. & Gilmore, J.H. Functional Connectivity of the Infant Human 766 Brain Plastic and Modifiable. The Neuroscientist, 1073858416635986 (2016). 767 18. Metcalfe, N.B., Taylor, A.C. & Thorpe, J.E. Metabolic rate, social status and life-history 768 strategies in Atlantic salmon. Animal behaviour 49, 431-436 (1995). 769 19. Leonard, W.R. & Robertson, M.L. Evolutionary perspectives on human nutrition: The 770 influence of brain and body size on diet and metabolism. American Journal of Human Biology 6, 77-771 88 (1994). 772 20. Dunbar, R.I. The social brain hypothesis and its implications for social evolution. Ann Hum 773 Biol 36, 562-572 (2009). 774 21. SOARES, C.A. & CARNEIRO, R.S. Social behavior between mothers ´ young of sloths Bradypus 775 variegatus SCHINZ, 1825 (Xenarthra: Bradypodidae). Brazilian Journal of Biology 62, 249-252 776 (2002). 777 22. Richard, A.F. & Nicoll, M.E. Female social dominance and basal metabolism in a Malagasy 778 primate, Propithecus verreauxi. American Journal of Primatology 12, 309-314 (1987). 779 23. Curley, J.P. & Keverne, E.B. Genes, brains and mammalian social bonds. Trends in Ecology & 780 Evolution 20, 561-567 (2005). 781 24. Schulkin, J. Allostasis, homeostasis, and the costs of physiological adaptation (Cambridge 782 University Press, 2004). 783 25. Shpigler, H.Y., et al. Deep evolutionary conservation of autism-related genes. Proceedings of 784 the National Academy of Sciences (2017). 785 26. Bullmore, E. & Sporns, O. The economy of brain network organization. Nat Rev Neurosci 13, 786 336-349 (2012). 787 27. Yeo, B.T., et al. The organization of the human cerebral cortex estimated by intrinsic 788 functional connectivity. J Neurophysiol 106, 1125-1165 (2011). 789 28. van den Heuvel, M.P. & Sporns, O. Rich-club organization of the human connectome. J 790 Neurosci 31, 15775-15786 (2011). 791 29. van den Heuvel, M.P., et al. Abnormal rich club organization and functional brain dynamics 792 in schizophrenia. JAMA psychiatry 70, 783-792 (2013). 793

31

30. Atzil, S., Hendler, T. & Feldman, R. Specifying the Neurobiological Basis of Human 794 Attachment: Brain, Hormones, and Behavior in Synchronous and Intrusive Mothers. 795 Neuropsychopharmacology 36, 2603-2615 (2011). 796 31. Atzil, S., Hendler, T. & Feldman, R. The brain basis of social synchrony. Social cognitive and 797 affective neuroscience (2013). 798 32. Atzil, S., et al. Dopamine in the medial amygdala network mediates human bonding. Proc 799 Natl Acad Sci U S A 114, 2361-2366 (2017). 800 33. Bickart, K.C., Hollenbeck, M.C., Barrett, L.F. & Dickerson, B.C. Intrinsic amygdala-cortical 801 functional connectivity predicts social network size in humans. J Neurosci 32, 14729-14741 (2012). 802 34. Uddin, L.Q., et al. Salience network–based classification and prediction of symptom severity 803 in children with autism. JAMA psychiatry 70, 869-879 (2013). 804 35. Di Martino, A., et al. The autism brain imaging data exchange: towards a large-scale 805 evaluation of the intrinsic brain architecture in autism. Molecular psychiatry 19, 659-667 (2014). 806 36. Dubois, J., et al. The early development of brain white matter: a review of imaging studies in 807 fetuses, newborns and infants. Neuroscience 276, 48-71 (2014). 808 37. Gao, W., et al. Temporal and spatial evolution of brain network topology during the first two 809 years of life. PloS one 6, e25278 (2011). 810 38. Gao, W., et al. The Synchronization within and Interaction between the Default and Dorsal 811 Attention Networks in Early Infancy. Cerebral cortex 23, 594-603 (2013). 812 39. Gao, W., et al. Temporal and spatial development of axonal maturation and myelination of 813 white matter in the developing brain. AJNR Am J Neuroradiol 30, 290-296 (2009). 814 40. Smyser, C.D., Snyder, A.Z. & Neil, J.J. Functional connectivity MRI in infants: Exploration of 815 the functional organization of the developing brain. NeuroImage 56, 1437-1452 (2011). 816 41. Fransson, P., Aden, U., Blennow, M. & Lagercrantz, H. The functional architecture of the 817 infant brain as revealed by resting-state fMRI. Cerebral cortex 21, 145-154 (2011). 818 42. Fransson, P., et al. Resting-state networks in the infant brain. Proceedings of the National 819 Academy of Sciences 104, 15531-15536 (2007). 820 43. Elton, A., Alcauter, S. & Gao, W. Network connectivity abnormality profile supports a 821 categorical-dimensional hybrid model of ADHD. Hum Brain Mapp 35, 4531-4543 (2014). 822 44. Fair, D.A., et al. The maturing architecture of the brain's default network. Proceedings of the 823 National Academy of Sciences 105, 4028-4032 (2008). 824 45. Tau, G.Z. & Peterson, B.S. Normal development of brain circuits. Neuropsychopharmacology 825 35, 147-168 (2010). 826 46. Gao, W., Alcauter, S., Smith, J.K., Gilmore, J.H. & Lin, W. Development of human brain 827 cortical network architecture during infancy. Brain Structure and Function 220, 1173-1186 (2015). 828 47. Stiles, J. & Jernigan, T.L. The basics of brain development. Neuropsychology review 20, 327-829 348 (2010). 830 48. Dubois, J., et al. Primary cortical folding in the human newborn: an early marker of later 831 functional development. Brain 131, 2028-2041 (2008). 832 49. Finlay, B.L. & Uchiyama, R. The timing of brain maturation, early experience, and the human 833 social niche. in Evolution of Nervous Systems 2e 123–148 (Oxford: Elsevier, 2017). 834 50. Rogers, C.E., et al. Regional cerebral development at term relates to school-age social-835 emotional development in very preterm children. J Am Acad Child Adolesc Psychiatry 51, 181-191 836 (2012). 837

32

51. Woodward, L.J., Clark, C.A., Bora, S. & Inder, T.E. Neonatal white matter abnormalities an 838 important predictor of neurocognitive outcome for very preterm children. PloS one 7, e51879 839 (2012). 840 52. Curley, J.P. & Champagne, F.A. Influence of maternal care on the developing brain: 841 Mechanisms, temporal dynamics and sensitive periods. Frontiers in neuroendocrinology 40, 52-66 842 (2016). 843 53. Johnson, M.H. Functional brain development in humans. Nature Reviews Neuroscience 2, 844 475-483 (2001). 845 54. Feldman, R. Parent-infant synchrony and the construction of shared timing; physiological 846 precursors, developmental outcomes, and risk conditions. J Child Psychol Psychiatry 48, 329-354 847 (2007). 848 55. Tomoda, A., et al. Reduced prefrontal cortical gray matter volume in young adults exposed 849 to harsh corporal punishment. Neuroimage 47, T66-T71 (2009). 850 56. Whittle, S., et al. Positive parenting predicts the development of adolescent brain structure: 851 a longitudinal study. Developmental cognitive neuroscience 8, 7-17 (2014). 852 57. Teicher, M.H., Anderson, C.M. & Polcari, A. Childhood maltreatment is associated with 853 reduced volume in the hippocampal subfields CA3, dentate gyrus, and subiculum. Proceedings of 854 the National Academy of Sciences 109, E563-E572 (2012). 855 58. Luby, J.L., et al. Maternal support in early childhood predicts larger hippocampal volumes at 856 school age. Proceedings of the National Academy of Sciences 109, 2854-2859 (2012). 857 59. Champagne, F.A., et al. Maternal care associated with methylation of the estrogen receptor-858 alpha1b promoter and estrogen receptor-alpha expression in the medial preoptic area of female 859 offspring. Endocrinology 147, 2909-2915 (2006). 860 60. McGowan, P.O., et al. Epigenetic regulation of the glucocorticoid receptor in human brain 861 associates with childhood abuse. Nat Neurosci 12, 342-348 (2009). 862 61. Pena, C.J., Neugut, Y.D., Calarco, C.A. & Champagne, F.A. Effects of maternal care on the 863 development of midbrain dopamine pathways and reward-directed behavior in female offspring. 864 Eur J Neurosci 39, 946-956 (2014). 865 62. Insel, T.R. Oxytocin--a neuropeptide for affiliation: evidence from behavioral, receptor 866 autoradiographic, and comparative studies. Psychoneuroendocrinology 17, 3-35 (1992). 867 63. Webb, A.R., Heller, H.T., Benson, C.B. & Lahav, A. Mother's voice and heartbeat sounds elicit 868 auditory plasticity in the human brain before full gestation. Proc Natl Acad Sci U S A (2015). 869 64. Teicher, M.H., Samson, J.A., Anderson, C.M. & Ohashi, K. The effects of childhood 870 maltreatment on brain structure, function and connectivity. Nat Rev Neurosci 17, 652-666 (2016). 871 65. Suomi, S.J. Early determinants of behaviour: evidence from primate studies. British Medical 872 Bulletin 53, 170-184 (1997). 873 66. Arling, G.L. & Harlow, H.F. Effects of social deprivation on maternal behavior of rhesus 874 monkeys. Journal of comparative and physiological psychology 64, 371-377 (1967). 875 67. Harlow, H.F. Total Social Isolation: Effects on Macaque Monkey Behavior. Science 148, 666 876 (1965). 877 68. Champagne, F.A., Francis, D.D., Mar, A. & Meaney, M.J. Variations in maternal care in the 878 rat as a mediating influence for the effects of environment on development. Physiol Behav 79, 359-879 371 (2003). 880

33

69. Champagne, F.A. Epigenetic mechanisms and the transgenerational effects of maternal 881 care. Front Neuroendocrinol 29, 386-397 (2008). 882 70. Champagne, F. & Meaney, M.J. Like mother, like daughter: evidence for non-genomic 883 transmission of parental behavior and stress responsivity. Progress in brain research 133, 287-302 884 (2001). 885 71. Pena, C.J., Neugut, Y.D. & Champagne, F.A. Developmental timing of the effects of maternal 886 care on gene expression and epigenetic regulation of hormone receptor levels in female rats. 887 Endocrinology 154, 4340-4351 (2013). 888 72. Feldman, R. The adaptive human parental brain: implications for children's social 889 development. Trends in neurosciences 38, 387-399 (2015). 890 73. Granat, A., Gadassi, R., Gilboa-Schechtman, E. & Feldman, R. Maternal Depression and 891 Anxiety, Social Synchrony, and Infant Regulation of Negative and Positive Emotions. (2016). 892 74. Herba, C.M. Maternal depression and child behavioural outcomes. The Lancet Psychiatry 1, 893 408-409 (2014). 894 75. Raby, K.L., Roisman, G.I., Simpson, J.A., Collins, W.A. & Steele, R.D. Greater Maternal 895 Insensitivity in Childhood Predicts Greater Electrodermal Reactivity During Conflict Discussions 896 With Romantic Partners in Adulthood. Psychological science (2015). 897 76. Feldman, R. Parent–infant synchrony: Biological foundations and developmental outcomes. 898 Current directions in psychological science 16, 340-345 (2007). 899 77. Carey, S. & Spelke, E. Science and Core Knowledge. Philosophy of Science 63, 515-533 900 (1996). 901 78. Spelke, E.S. & Kinzler, K.D. Core knowledge. Developmental science 10, 89-96 (2007). 902 79. Clark, A. Whatever next? Predictive brains, situated agents, and the future of cognitive 903 science. The Behavioral and brain sciences 36, 181-204 (2013). 904 80. Hohwy, J. The predictive mind (Oxford University Press, 2013). 905 81. Rao, R.P. & Ballard, D.H. Predictive coding in the visual cortex: a functional interpretation of 906 some extra-classical receptive-field effects. Nature neuroscience 2, 79-87 (1999). 907 82. Friston, K. The free-energy principle: a unified brain theory? Nat Rev Neurosci 11, 127-138 908 (2010). 909 83. Clark, A. Are we predictive engines? Perils, prospects, and the puzzle of the porous 910 perceiver. Behavioral and Brain Sciences 36, 233-253 (2013). 911 84. Barrett, L.F. & Simmons, W.K. Interoceptive predictions in the brain. Nat Rev Neurosci 16, 912 419-429 (2015). 913 85. Chanes, L. & Barrett, L.F. Redefining the Role of Limbic Areas in Cortical Processing. Trends 914 Cogn Sci 20, 96-106 (2016). 915 86. Friston, K. A theory of cortical responses. Philosophical Transactions of the Royal Society of 916 London B: Biological Sciences 360, 815-836 (2005). 917 87. Gopnik, A. The philosophical baby. London: Bodley Head (2009). 918 88. Siegelman, N. & Frost, R. Statistical learning as an individual ability: Theoretical perspectives 919 and empirical evidence. Journal of Memory and Language 81, 105-120 (2015). 920 89. Krogh, L., Vlach, H.A. & Johnson, S.P. Statistical learning across development: flexible yet 921 constrained. Frontiers in psychology 3, 598 (2012). 922 90. Saffran, J.R., Aslin, R.N. & Newport, E.L. Statistical learning by 8-month-old infants. Science 923 274, 1926-1928 (1996). 924

34

91. Kirkham, N.Z., Slemmer, J.A. & Johnson, S.P. Visual statistical learning in infancy: evidence 925 for a domain general learning mechanism. Cognition 83, B35-B42 (2002). 926 92. Tenenbaum, J.B., Kemp, C., Griffiths, T.L. & Goodman, N.D. How to grow a mind: statistics, 927 structure, and abstraction. Science 331, 1279-1285 (2011). 928 93. Sterling, P. & Laughlin, S. Principles of neural design (MIT Press, 2015). 929 94. Carpenter, R. Homeostasis: a plea for a unified approach. Advances in Physiology Education 930 28, 180-187 (2004). 931 95. Gu, X. & FitzGerald, T. Interoceptive inference: homeostasis and decision-making. Trends 932 Cogn. Sci 18, 269-270 (2014). 933 96. Seth, A.K. Interoceptive inference, emotion, and the embodied self. Trends in cognitive 934 sciences 17, 565-573 (2013). 935 97. Seth, A.K., Suzuki, K. & Critchley, H.D. An interoceptive predictive coding model of conscious 936 presence. Front Psychol 2, 395 (2012). 937 98. Finlay, B.L. & Syal, S. The pain of altruism. Trends in cognitive sciences 18, 615-617 (2014). 938 99. Lummaa, V., Vuorisalo, T., Barr, R.G. & Lehtonen, L. Why cry? Adaptive significance of 939 intensive crying in human infants. Evolution and human behavior 19, 193-202 (1998). 940 100. Davis, E.P., et al. Exposure to unpredictable maternal sensory signals influences cognitive 941 development across species. Proceedings of the National Academy of Sciences, 201703444 (2017). 942 101. Stein, B.E., Stanford, T.R. & Rowland, B.A. Development of multisensory integration from 943 the perspective of the individual neuron. Nature reviews. Neuroscience 15, 520-535 (2014). 944 102. Petanjek, Z., Judaš, M., Kostović, I. & Uylings, H.B.M. Lifespan Alterations of Basal Dendritic 945 Trees of Pyramidal Neurons in the Human Prefrontal Cortex: A Layer-Specific Pattern. Cerebral 946 cortex 18, 915-929 (2008). 947 103. Alcauter, S., et al. Development of thalamocortical connectivity during infancy and its 948 cognitive correlations. J Neurosci 34, 9067-9075 (2014). 949 104. Alcauter, S., Lin, W., Keith Smith, J., Gilmore, J.H. & Gao, W. Consistent anterior-posterior 950 segregation of the insula during the first 2 years of life. Cerebral cortex 25, 1176-1187 (2015). 951 105. Trachtenberg, J.T. & Stryker, M.P. Rapid Anatomical Plasticity of Horizontal Connections in 952 the Developing Visual Cortex. The Journal of Neuroscience 21, 3476-3482 (2001). 953 106. Barrett, L.F. The theory of constructed emotion: an active inference account of 954 interoception and categorization. Soc Cogn Affect Neurosci (2017). 955 107. Sepulcre, J., Sabuncu, M.R., Yeo, T.B., Liu, H. & Johnson, K.A. Stepwise connectivity of the 956 modal cortex reveals the multimodal organization of the human brain. Journal of Neuroscience 32, 957 10649-10661 (2012). 958 108. Xu, P., et al. Different topological organization of human brain functional networks with 959 eyes open versus eyes closed. Neuroimage 90, 246-255 (2014). 960 109. Sterzer, P. & Kleinschmidt, A. Anterior insula activations in perceptual paradigms: often 961 observed but barely understood. Brain Structure and Function 214, 611-622 (2010). 962 110. Singer, T. & Lamm, C. The social neuroscience of empathy. Ann N Y Acad Sci 1156, 81-96 963 (2009). 964 111. Andrews-Hanna, J.R., Smallwood, J. & Spreng, R.N. The default network and self-generated 965 thought: component processes, dynamic control, and clinical relevance. Ann N Y Acad Sci 1316, 29-966 52 (2014). 967 112. Shamay-Tsoory, S.G. The neural bases for empathy. The Neuroscientist 17, 18-24 (2011). 968

35

113. Ganzel, B.L., Morris, P.A. & Wethington, E. Allostasis and the human brain: Integrating 969 models of stress from the social and life sciences. Psychol Rev 117, 134-174 (2010). 970 114. Bickart, K.C., Dickerson, B.C. & Barrett, L.F. The amygdala as a hub in brain networks that 971 support social life. Neuropsychologia 63, 235-248 (2014). 972 115. Schultz, W. & Dickinson, A. Neuronal coding of prediction errors. Annual review of 973 neuroscience 23, 473-500 (2000). 974 116. Lindquist, K.A., Satpute, A.B., Wager, T.D., Weber, J. & Barrett, L.F. The brain basis of 975 positive and negative affect: evidence from a meta-analysis of the human neuroimaging literature. 976 Cerebral cortex, bhv001 (2015). 977 117. Bar, M. The proactive brain: memory for predictions. Philos Trans R Soc Lond B Biol Sci 364, 978 1235-1243 (2009). 979 118. Bar, M. Predictions: a universal principle in the operation of the human brain. Introduction. 980 Philos Trans R Soc Lond B Biol Sci 364, 1181-1182 (2009). 981 119. Murphy, G. The big book of concepts (MIT press, 2004). 982 120. Barsalou, L.W. Ad hoc categories. Mem Cognit 11, 211-227 (1983). 983 121. Barsalou, L.W., Kyle Simmons, W., Barbey, A.K. & Wilson, C.D. Grounding conceptual 984 knowledge in modality-specific systems. Trends Cogn Sci 7, 84-91 (2003). 985 122. Hollerman, J.R. & Schultz, W. Dopamine neurons report an error in the temporal prediction 986 of reward during learning. Nat Neurosci 1, 304-309 (1998). 987 123. Angelaki, D.E., Gu, Y. & DeAngelis, G.C. Multisensory integration: psychophysics, 988 neurophysiology, and computation. Current opinion in neurobiology 19, 452-458 (2009). 989 124. Feldman, R., Rosenthal, Z. & Eidelman, A.I. Maternal-preterm skin-to-skin contact enhances 990 child physiologic organization and cognitive control across the first 10 years of life. Biol Psychiatry 991 75, 56-64 (2014). 992 125. Levin, B.E. Metabolic imprinting: critical impact of the perinatal environment on the 993 regulation of energy homeostasis. Philosophical Transactions of the Royal Society of London B: 994 Biological Sciences 361, 1107-1121 (2006). 995 126. Bauman, D. Regulation of nutrient partitioning during lactation: homeostasis and 996 homeorhesis revisited. Ruminant physiology: digestion, metabolism, growth and reproduction, 311-997 328 (2000). 998 127. Arrieta, M.C., Stiemsma, L.T., Amenyogbe, N., Brown, E.M. & Finlay, B. The intestinal 999 microbiome in early life: health and disease. Frontiers in immunology 5, 427 (2014). 1000 128. Nakata, T. & Trehub, S.E. Infants’ responsiveness to maternal speech and singing. Infant 1001 Behavior and Development 27, 455-464 (2004). 1002 129. Tomasello, M. Joint attention as social cognition. Joint attention: Its origins and role in 1003 development, 103-130 (1995). 1004 130. Amso, D. & Scerif, G. The attentive brain: insights from developmental cognitive 1005 neuroscience. Nat Rev Neurosci 16, 606-619 (2015). 1006 131. Baron-Cohen, S. The development of a theory of mind in autism: deviance and delay? 1007 Psychiatr Clin North Am 14, 33-51 (1991). 1008 132. Belmonte, M.K., et al. Autism as a disorder of neural information processing: directions for 1009 research and targets for therapy[ast]. Molecular psychiatry 9, 646-663 (2004). 1010 133. Trehub, S.E. & Gudmundsdottir, H.R. Mothers as singing mentors for infants. in The Oxford 1011 Handbook of Singing (2015). 1012

36

134. MacLean, P.C., et al. Mother–infant mutual eye gaze supports emotion regulation in infancy 1013 during the Still-Face paradigm. Infant Behavior and Development 37, 512-522 (2014). 1014 135. Mantis, I., Stack, D.M., Ng, L., Serbin, L.A. & Schwartzman, A.E. Mutual touch during 1015 mother–infant face-to-face still-face interactions: Influences of interaction period and infant birth 1016 status. Infant Behavior and Development 37, 258-267 (2014). 1017 136. Ramsay, D.S. & Woods, S.C. Clarifying the roles of homeostasis and allostasis in physiological 1018 regulation. Psychological review 121, 225 (2014). 1019 137. Muenzinger, K.F. & Fletcher, F.M. Motivation in learning. VI. Escape from electric shock 1020 compared with hunger-food tension in the visual discrimination habit. Journal of comparative 1021 psychology 22, 79 (1936). 1022 138. Petrinovich, L. & Bolles, R. Deprivation states and behavioral attributes. Journal of 1023 comparative and physiological psychology 47, 450 (1954). 1024 139. Okanoya, K. Evolution of Vocal Communication: From Animal Calls to Human Speech. in 1025 Evolution of the Brain, Cognition, and Emotion in Vertebrates 207-224 (Springer, 2017). 1026 140. Scott, J.P. Critical periods in the development of social behavior in puppies. Psychosomatic 1027 medicine 20, 42-54 (1958). 1028 141. Li, S.S.Y. & McNally, G.P. The conditions that promote fear learning: Prediction error and 1029 Pavlovian fear conditioning. Neurobiology of Learning and Memory 108, 14-21 (2014). 1030 142. Preuss, T.M. The human brain: rewired and running hot. Ann N Y Acad Sci 1225 (2011). 1031 143. Spocter, M.A., et al. Neuropil distribution in the cerebral cortex differs between humans and 1032 chimpanzees. Journal of Comparative Neurology 520, 2917-2929 (2012). 1033 144. Barrett, L.F. The theory of constructed emotion: an active inference account of 1034 interoception and categorization. Social cognitive and affective neuroscience 12, 1-23 (2017). 1035 145. Finlay, B. & Uchiyama, R. The timing of brain maturation, early experience, and the human 1036 social niche. (2017). 1037 146. Hauser, M.D., Chomsky, N. & Fitch, W.T. The faculty of language: what is it, who has it, and 1038 how did it evolve? Science 298, 1569-1579 (2002). 1039 147. Bloom, P. Precis of How children learn the meanings of words. The Behavioral and brain 1040 sciences 24, 1095-1103; discussion 1104-1034 (2001). 1041 148. Lupfer, G., Frieman, J. & Coonfield, D. Social transmission of flavor preferences in two 1042 species of hamsters (Mesocricetus auratus and Phodopus campbelli). Journal of comparative 1043 psychology 117, 449-455 (2003). 1044 149. Galef, B.G. & Laland, K.N. Social learning in animals: empirical studies and theoretical 1045 models. AIBS Bulletin 55, 489-499 (2005). 1046 150. Uller, T. Developmental plasticity and the evolution of parental effects. Trends in ecology & 1047 evolution 23, 432-438 (2008). 1048 151. Wolf, J.B. & Brodie, E.D. The coadaptation of parental and offspring characters. Evolution 1049 52, 299-308 (1998). 1050 152. Shweder, R.A. & Shweder, R.A. Cultural psychology – what is it? 1051 Cultural psychology (Cambridge University Press, 1990). 1052 153. Atzil, S. & Gendron, M. Bio-Behavioral Synchrony and the Development of Emotion: A Role 1053 for Conceptualization. Current Opinion in Psychology (2017). 1054 154. Gendron, M., Roberson, D. & Barrett, L.F. Cultural Variation in Emotion Perception Is Real: A 1055 Response to Sauter, Eisner, Ekman, and Scott (2015). Psychological science (2015). 1056

37

155. Russell, J.A. Culture and the categorization of emotions. Psychological bulletin 110, 426-450 1057 (1991). 1058 156. Andrews-Hanna, J.R. The brain's default network and its adaptive role in internal mentation. 1059 The Neuroscientist : a review journal bringing neurobiology, neurology and psychiatry 18, 251-270 1060 (2012). 1061 157. Lombardo, M.V., et al. Shared neural circuits for mentalizing about the self and others. 1062 Journal of cognitive neuroscience 22, 1623-1635 (2010). 1063 158. Gao, W., et al. Functional Network Development During the First Year: Relative Sequence 1064 and Socioeconomic Correlations. Cerebral cortex 25, 2919-2928 (2015). 1065 159. Gao, W., et al. Evidence on the emergence of the brain's default network from 2-week-old 1066 to 2-year-old healthy pediatric subjects. Proceedings of the National Academy of Sciences 106, 1067 6790-6795 (2009). 1068 160. Supekar, K., et al. Development of functional and structural connectivity within the default 1069 mode network in young children. NeuroImage 52, 290-301 (2010). 1070 161. Blakemore, S.J., den Ouden, H., Choudhury, S. & Frith, C. Adolescent development of the 1071 neural circuitry for thinking about intentions. Soc Cogn Affect Neurosci 2, 130-139 (2007). 1072 162. Alcauter, S., et al. Frequency of spontaneous BOLD signal shifts during infancy and 1073 correlates with cognitive performance. Dev Cogn Neurosci 12, 40-50 (2015). 1074 163. Uddin, L.Q., Supekar, K.S., Ryali, S. & Menon, V. Dynamic reconfiguration of structural and 1075 functional connectivity across core neurocognitive brain networks with development. J Neurosci 31, 1076 18578-18589 (2011). 1077 164. Wegner, D.M. Transactive memory: A contemporary analysis of the group mind. in Theories 1078 of group behavior 185-208 (Springer, 1987). 1079 165. Syal, S. & Finlay, B.L. Thinking outside the cortex: social motivation in the evolution and 1080 development of language. Developmental science 14, 417-430 (2011). 1081 166. Dunbar, R.I.M. The social brain hypothesis. Evolutionary Anthropology: Issues, News, and 1082 Reviews 6, 178-190 (1998). 1083 167. Gunnar, M.R. & Sullivan, R.M. The neurodevelopment of social buffering and fear learning: 1084 integration and crosstalk. Social Neuroscience 12, 1-7 (2017). 1085 168. Coan, J.A., Schaefer, H.S. & Davidson, R.J. Lending a hand: social regulation of the neural 1086 response to threat. Psychological science 17, 1032-1039 (2006). 1087 169. Master, S.L., et al. A picture's worth: partner photographs reduce experimentally induced 1088 pain. Psychological science 20, 1316-1318 (2009). 1089 170. Lantolf, J.P., Thorne, S.L. & Poehner, M.E. Sociocultural theory and second language 1090 development. Theories in second language acquisition: An introduction, 207-226 (2015). 1091 171. Padilla, A.M. & Perez, W. Acculturation, social identity, and social cognition: A new 1092 perspective. Hispanic Journal of Behavioral Sciences 25, 35-55 (2003). 1093 172. Adolphs, R. The social brain: neural basis of social knowledge. Annu Rev Psychol 60, 693-716 1094 (2009). 1095 173. Frith, C.D. The social brain? Philos Trans R Soc Lond B Biol Sci 362, 671-678 (2007). 1096 174. Whitacre, J.M., Rohlfshagen, P., Bender, A. & Yao, X. Evolutionary mechanics: new 1097 engineering principles for the emergence of flexibility in a dynamic and uncertain world. Natural 1098 computing 11, 431-448 (2012). 1099

38

175. Boyd, R., Richerson, P.J. & Henrich, J. The cultural niche: Why social learning is essential for 1100 human adaptation. Proceedings of the National Academy of Sciences 108, 10918-10925 (2011). 1101 176. Heyes, C. & Pearce, J.M. Not-so-social learning strategies. Proceedings of the Royal Society 1102 B: Biological Sciences 282 (2015). 1103 177. Champagne, F.A. & Meaney, M.J. Transgenerational effects of social environment on 1104 variations in maternal care and behavioral response to novelty. Behav Neurosci 121, 1353-1363 1105 (2007). 1106 178. Francis, D., Diorio, J., Liu, D. & Meaney, M.J. Nongenomic transmission across generations of 1107 maternal behavior and stress responses in the rat. Science 286, 1155-1158 (1999). 1108 179. Lorenz, K. Der Kumpan in der Umwelt des Vogels. J. Ornithol 83, 289-413 (1935). 1109 180. Lorenz, K. My family and other animals. Leaders in the study of animal behavior: 1110 Autobiographical perspectives, 259-287 (1985). 1111 181. Morton, J. & Johnson, M.H. CONSPEC and CONLERN: a two-process theory of infant face 1112 recognition. Psychol Rev 98, 164-181 (1991). 1113 182. Braddick, O. Human Development: Faces in the Womb. Current Biology 27, R704-R706 1114 (2017). 1115 183. Cook, R., Bird, G., Catmur, C., Press, C. & Heyes, C. Mirror neurons: from origin to function. 1116 The Behavioral and brain sciences 37, 177-192 (2014). 1117 184. Turati, C., Di Giorgio, E., Bardi, L. & Simion, F. Holistic face processing in newborns, 3-month-1118 old infants, and adults: evidence from the composite face effect. Child Dev 81, 1894-1905 (2010). 1119 185. Gava, L., Valenza, E., Turati, C. & de Schonen, S. Effect of partial occlusion on newborns' face 1120 preference and recognition. Developmental science 11, 563-574 (2008). 1121 186. Turati, C., Bulf, H. & Simion, F. Newborns' face recognition over changes in viewpoint. 1122 Cognition 106, 1300-1321 (2008). 1123 187. Cassia, V.M., Turati, C. & Simion, F. Can a nonspecific bias toward top-heavy patterns 1124 explain newborns' face preference? Psychological science 15, 379-383 (2004). 1125 188. Simion, F., Leo, I., Turati, C., Valenza, E. & Dalla Barba, B. How face specialization emerges in 1126 the first months of life. Progress in brain research 164, 169-185 (2007). 1127 189. Turati, C. Why Faces Are Not Special to Newborns: An Alternative Account of the Face 1128 Preference. Current Directions in Psychological Science 13, 5-8 (2004). 1129 190. Gartstein, M.A. & Rothbart, M.K. Studying infant temperament via the revised infant 1130 behavior questionnaire. Infant Behavior and Development 26, 64-86 (2003). 1131 191. Huffman, L.C., et al. Infant temperament and cardiac vagal tone: Assessments at twelve 1132 weeks of age. Child Dev 69, 624-635 (1998). 1133 192. Davidov, M., Knafo-Noam, A., Serbin, L.A. & Moss, E. The influential child: How children 1134 affect their environment and influence their own risk and resilience. Development and 1135 psychopathology 27, 947-951 (2015). 1136 193. Rothbart, M.K. & Ahadi, S.A. Temperament and the development of personality. Journal of 1137 abnormal psychology 103, 55 (1994). 1138 194. George, O., Le Moal, M. & Koob, G.F. Allostasis and addiction: role of the dopamine and 1139 corticotropin-releasing factor systems. Physiology & behavior 106, 58-64 (2012). 1140 195. Koob, G.F. & Le Moal, M. Drug addiction, dysregulation of reward, and allostasis. 1141 Neuropsychopharmacology 24, 97-129 (2001). 1142

39

196. Beauchaine, T.P., Neuhaus, E., Zalewski, M., Crowell, S.E. & Potapova, N. The effects of 1143 allostatic load on neural systems subserving motivation, mood regulation, and social affiliation. 1144 Development and psychopathology 23, 975-999 (2011). 1145 197. Buckner, R.L., Andrews-Hanna, J.R. & Schacter, D.L. The brain's default network: anatomy, 1146 function, and relevance to disease. Ann N Y Acad Sci 1124, 1-38 (2008). 1147 198. Young, L.J. & Barrett, C.E. Neuroscience. Can oxytocin treat autism? Science 347, 825-826 1148 (2015). 1149 199. Farmer, C., Thurm, A. & Grant, P. Pharmacotherapy for the core symptoms in autistic 1150 disorder: current status of the research. Drugs 73, 303-314 (2013). 1151 200. Pellicano, E. & Burr, D. When the world becomes ‘too real’: a Bayesian 1152 explanation of autistic perception. Trends in Cognitive Sciences 16, 504-510. 1153 201. Verly, M., et al. Altered functional connectivity of the language network in ASD: role of 1154 classical language areas and cerebellum. NeuroImage: Clinical 4, 374-382 (2014). 1155 202. Jaffe-Dax, S., Frenkel, O. & Ahissar, M. Dyslexics' faster decay of implicit memory for sounds 1156 and words is manifested in their shorter neural adaptation. Elife 6 (2017). 1157 203. Jaffe-Dax, S., Raviv, O., Jacoby, N., Loewenstein, Y. & Ahissar, M. Towards a computational 1158 model of Dyslexia. BMC Neuroscience 16, O12 (2015). 1159 204. Leerkes, E.M., Su, J., Calkins, S.D., O'Brien, M. & Supple, A.J. Maternal physiological 1160 dysregulation while parenting poses risk for infant attachment disorganization and behavior 1161 problems. Development and psychopathology, 1-13 (2016). 1162 205. TASKER, F. Lesbian Mothers, Gay Fathers, and Their Children: A Review. Journal of 1163 Developmental & Behavioral Pediatrics 26, 224-240 (2005). 1164 206. Bornstein, M.H. & Bradley, R.H. Socioeconomic status, parenting, and child development 1165 (Routledge, 2014). 1166 207. Merz, E.C., Tottenham, N. & Noble, K.G. Socioeconomic Status, Amygdala Volume, and 1167 Internalizing Symptoms in Children and Adolescents. Journal of Clinical Child & Adolescent 1168 Psychology 47, 312-323 (2018). 1169 208. Kolb, B., Gibb, R. & Robinson, T.E. Brain Plasticity and Behavior. Current Directions in 1170 Psychological Science 12, 1-5 (2003). 1171 209. Bagot, R.C., et al. Variations in postnatal maternal care and the epigenetic regulation of 1172 metabotropic glutamate receptor 1 expression and hippocampal function in the rat. Proc Natl Acad 1173 Sci U S A 109 Suppl 2, 17200-17207 (2012). 1174 210. Liu, D., et al. Maternal care, hippocampal glucocorticoid receptors, and hypothalamic-1175 pituitary-adrenal responses to stress. Science 277, 1659-1662 (1997). 1176 211. Caldji, C., Diorio, J. & Meaney, M.J. Variations in maternal care alter GABA(A) receptor 1177 subunit expression in brain regions associated with fear. Neuropsychopharmacology 28, 1950-1959 1178 (2003). 1179 212. Francis, D.D., Young, L.J., Meaney, M.J. & Insel, T.R. Naturally occurring differences in 1180 maternal care are associated with the expression of oxytocin and vasopressin (V1a) receptors: 1181 gender differences. J Neuroendocrinol 14, 349-353 (2002). 1182 213. Weaver, I.C. Epigenetic programming by maternal behavior and pharmacological 1183 intervention. Nature versus nurture: let's call the whole thing off. Epigenetics 2, 22-28 (2007). 1184 214. Meaney, M.J. Maternal care, gene expression, and the transmission of individual differences 1185 in stress reactivity across generations. Annual review of neuroscience 24, 1161-1192 (2001). 1186

40

215. Champagne, F.A., et al. Variations in nucleus accumbens dopamine associated with 1187 individual differences in maternal behavior in the rat. J Neurosci 24, 4113-4123 (2004). 1188 216. Weaver, I.C.G., et al. Epigenetic programming by maternal behavior. Nat Neurosci 7, 847-1189 854 (2004). 1190 217. Murgatroyd, C. & Spengler, D. Epigenetics of early child development. Frontiers in psychiatry 1191 2, 16 (2011). 1192 218. Branchi, I. & Cirulli, F. Early experiences: Building up the tools to face the challenges of adult 1193 life. Developmental Psychobiology 56, 1661-1674 (2014). 1194 219. Cirulli, F., et al. Early life stress as a risk factor for mental health: role of neurotrophins from 1195 rodents to non-human primates. Neurosci Biobehav Rev 33, 573-585 (2009). 1196 220. Branchi, I., et al. Early interactions with mother and peers independently build adult social 1197 skills and shape BDNF and oxytocin receptor brain levels. Psychoneuroendocrinology 38, 522-532 1198 (2013). 1199 221. Jiao, Y., et al. A key mechanism underlying sensory experience-dependent maturation of 1200 neocortical GABAergic circuits in vivo. Proc Natl Acad Sci U S A 108, 12131-12136 (2011). 1201 222. Pineda, R.G., et al. Alterations in brain structure and neurodevelopmental outcome in 1202 preterm infants hospitalized in different neonatal intensive care unit environments. J Pediatr 164, 1203 52-60 e52 (2014). 1204 223. Fifer, W.P. & Moon, C.M. The role of mother's voice in the organization of brain function in 1205 the newborn. Acta Paediatr Suppl 397, 86-93 (1994). 1206 224. Lynch, C.J., et al. Default mode network in childhood autism: posteromedial cortex 1207 heterogeneity and relationship with social deficits. Biological psychiatry 74, 212-219 (2013). 1208 225. Rudie, J.D., et al. Altered functional and structural brain network organization in autism. 1209 NeuroImage: Clinical 2, 79-94 (2013). 1210 1211

Allostasis deviation Allostasis regulation

C

(a) Interoception of allostasis- via primary interoceptive cortex in Posterior Insula (PI).(b) Exteroception of mommy- via tactile, visual, auditory, olfactory sensory Cortices.

Anterior Insula (AI)-this neural hub integrates sensory projections from primary exteroceptive and interoceptivecortices into a multi-modal representation of mommy: a neural model that associates social information with allostasis. With experience, the anterior insula is potentiated to predict allostatic changes 84

based on social stimuli.

Prediction signalsPrediction error signals

PI

AI

Prediction errors

Probabilistic modelsA group of sensations

with common features/goal/action

Predictions

Cognitive Science Framework Predictive Coding Framework

Organization

Construction

Perception

Sensory Input

CategoriesA group of sensations

with common features/goal/action

Concepts

Xa

Mommy`s voice

Auditory

Hearing mommy

Prediction error

Xv

Mommy`s face

Visual

Seeing mommy

(prediction error)

Xi

Glucouse levels

Interoceptive

Representation of feeding

(change in glucose levels)

(predicted sensation)

Exteroceptive

Representation of mommy

(new auditory information)

(predicted sensation)

(new visual information)

(predicted voice)

(predicted face)

Mommy Concept

(new sensory information regarding caregiver)

(predicted glucose regulation)

(new information regarding allostasis)

(predicted mommy's figure)

(predicted allostatic regulation)

Social Development: In humans, newborns’ allostasis is always regulated by acaretaker. Therefore, the sensory patterns that are relevant for newborns’ allostasisare social, and infants construct rudimentary social concepts such as ‘face’ and‘mommy’. Social concepts are conditioned with positive outcome of allostasisregulation. Through allostasis, social care becomes rewarding, and the infantattaches to the caretaker, and becomes increasingly driven towards socialinteractions. Gradually, more complex social and non-social information, skill andbehaviours are learned.

Concepts: To prepare for allostatic demands, the brain organizes environmentalsensory input into concepts. During development, infants detect statisticalregularities in the environment and from a meaningless stream of information,populations of instances are grouped into mentally represented concepts. Sensorypatterns relevant for allostatic will efficiently becomes concepts.

Allostasis: is the elementary ongoing process of optimizing the body’s internalmilieu. The brain is constantly regulating allostasis, while predicting allostaticneeds, and adjusting behaviour and metabolism towards survival, growth, andreproduction.

PIAI