Iterated Insights

Jared Edward Reser Ph.D. With GPT six Astra. 

Abstract

Autistic regression raises a developmental question: why do some children display social and communicative behaviors that subsequently diminish or disappear? This article develops a comparative neuroecological hypothesis extending the solitary-forager framework. In mammals that become relatively solitary, early family contact gives way to different social and ecological demands. Such transitions can involve changes in affiliation, reward, learning, and neuroendocrine regulation. Research on hamsters demonstrates developmental declines in social interaction alongside regional changes in oxytocin-receptor binding; other rodent studies show that receptor distributions and hormone-dependent learning processes vary across development. I propose that some autistic trajectories may involve altered timing, intensity, or coordination of related developmental processes, potentially shifting attention from socially guided learning toward more self-directed exploration. The proposal distinguishes attentional autonomy from practical independence and selective attachment from generalized social motivation. Human regression research, receptor studies, and alternative developmental theories delimit the account. Behavioral resemblance, shared regulatory mechanisms, and evolutionary adaptation are treated as separate evidential claims. The hypothesis generates longitudinal and comparative predictions concerning learning, preserved capacities, developmental timing, and region-specific signaling. Its principal contribution is to move the solitary-mammal comparison from adult traits toward the developmental processes that produce changing social phenotypes.

Keywords: autism; developmental regression; neuroecology; solitary forager hypothesis; oxytocin receptor; heterochrony; social learning; natal coat; childhood hair pigmentation

1. The developmental problem

An initially expressed behavior need not remain equally useful throughout life. Attachment, proximity seeking, exploration, and social tolerance operate within changing developmental circumstances. A young mammal may depend on family contact for survival and learning, while its later ecology favors a different balance between affiliation and independent activity. Explaining an adult social phenotype therefore requires attention to the trajectory through which it develops.

This perspective raises a question about autistic regression. Could some decreases in social engagement reflect altered deployment of developmental processes that normally reorganize the relationship between family dependence and self-directed activity? The question does not assume that autistic children become competent solitary foragers when they regress. It asks whether a changing social phenotype may sometimes involve regulation of attention and motivation that has informative counterparts elsewhere in mammalian development.

The solitary-forager hypothesis proposed that some autism-associated characteristics could be understood in relation to relatively independent subsistence and reduced reliance on continual social coordination (Reser, 2011). Its original formulation already considered maternal learning, subsequent independence, and a potentially abbreviated developmental learning trajectory. The comparative account was extended to naturally solitary mammals and their social neurobiology (Reser, 2014). The present article develops a more specific proposition: the developmental transition itself, rather than only resemblance between adult phenotypes, may be relevant to some regressive trajectories.

Three explanatory levels must remain distinct. A behavioral analogy concerns similar changes in engagement. A mechanistic hypothesis concerns shared or corresponding regulatory processes. An evolutionary hypothesis concerns the historical function of those processes or the variants that modify them. Evidence at one level can motivate investigation at another without establishing it. In particular, resemblance to an adaptive animal transition does not by itself demonstrate that human regression is adaptive.

This is a hypothesis-generating synthesis rather than a systematic review or a claim to explain every form of autism. Its initial target is a subgroup characterized by early social-communicative decline, examined alongside changes in attention and learning. Broad deterioration across motor, cognitive, and self-care functions requires additional explanation.

2. The prevalence and timing of autistic regression

Regression is sufficiently common to require a central place in developmental accounts of autism. Tan and colleagues (2021) synthesized 75 studies involving 33,014 participants and estimated a pooled prevalence of 30%, with a weighted mean onset of 19.8 months. Between-study heterogeneity was very high, and estimates varied with the definition of regression. Approximately one-third is therefore a useful empirical anchor, not a universal rate applicable to every population or measurement method.

Prospective research reveals changes that retrospective categorical reports may miss. In Ozonoff and colleagues’ (2018) cohort, 32 children eventually received an autism diagnosis. Declining social-behavior trajectories were identified in 88% by examiner dimensional ratings and 69% by parent dimensional ratings. A retrospective categorical measure identified regression in 29% of the 31 children with available interviews. These figures concern a small, predominantly familial-risk subgroup. They do not establish that 88% of all autistic children undergo abrupt loss of language or other acquired skills. Their importance is methodological: the apparent frequency and form of regression depend partly on what is measured and how often it is assessed.

Three observations should be separated. A skill may disappear; a behavior may occur less frequently despite some retained capacity; or development may plateau while age-related expectations increase. All can concern families and affect functioning, but they imply different mechanisms. Likewise, apparently typical early development does not establish the absence of earlier subtle differences. Repeated within-child measurements are more informative than a simple distinction between normal beginnings and later abnormality.

Outcomes also vary. Pickles and colleagues (2022) identified reported language regression in 90 of 408 children with relevant interview data and followed communication development to age ten. Trajectories were heterogeneous, although regression remained associated with lower expressive communication. A history of regression therefore neither guarantees irreversible decline nor implies that lost capacities are merely dormant.

Later deterioration must be considered separately. In a selected clinical series of 20 autistic patients with late regression, the mean onset was 13 years and 17 had catatonia (Ghaziuddin, 2021). These are not population prevalence estimates. They demonstrate that substantial later change can involve clinical processes that should not be assimilated automatically to a toddler-age learning transition. Timing is an explanatory variable, not sufficient evidence of a shared cause.

3. Developmental transitions in mammalian social behavior

3.1 Affiliation changes in naturally solitary hamsters

Beery, Lee, and Cooke (2025) provide a particularly relevant comparison. Syrian hamster behavior was assessed at 2.5, 4, and 8 weeks. Social interest and interaction declined across this interval, while play showed a juvenile peak. Separate tissue comparisons examined oxytocin-receptor binding in approximately two-week-old and eight- to nine-week-old Syrian and Siberian hamsters. Binding declined in several regions, including the endopiriform nucleus in both species; the largest Syrian-hamster reduction occurred in hippocampal CA2.

The study connects behavioral maturation with receptor reorganization around a dispersal-like transition. It does not establish that receptor changes caused the behavioral changes: age, separation, and laboratory housing were intertwined. Nor did it demonstrate loss of acquired skills. Its contribution is narrower and important: diminished affiliation can occur within a species-typical developmental trajectory and coincide with regional neurobiological change.

3.2 Behavioral timing is not reducible to puberty

Paul and colleagues (2018) manipulated reproductive maturation in Siberian hamsters through photoperiod and prepubertal gonadectomy. The timing of the transition from juvenile play toward more aggressive interaction was not simply delayed along with puberty. Gonadal status nevertheless affected behavioral levels. These results separate regulation of developmental timing from hormonal modulation of behavioral expression.

The authors proposed that this separation could allow dispersal-related milestones to occur at an appropriate age despite seasonal variation in reproductive maturation. That ecological interpretation was not itself a demonstrated fitness effect. Nevertheless, the experiment shows why a developmental account should not assume that weaning, puberty, exploration, and social reorganization are governed by one clock.

3.3 Chemical regulation of learning windows

Upton and Sullivan (2010) examined rat pups’ transition between attachment-related odor preference and conditioned fear learning. Increasing or decreasing corticosterone changed learning within a bounded interval, approximately postnatal days 6 through 15. The transition normally occurring around day 10 could therefore be modified experimentally. Induced early fear learning did not permanently eliminate preference learning, which could reappear under different hormonal conditions.

This is causal evidence that a chemical signal can regulate which learning processes are available at a developmental stage. It is not a demonstration of autistic regression or solitary dispersal. Its relevance lies in the general mechanism: a developing nervous system can change the significance of experience through age-dependent physiological regulation.

3.4 Selective attachment and general social motivation can diverge

Comparative research also discourages treating sociability as a single quantity. Beery and colleagues (2021) found that selective relationships and willingness to work for social access were not uniformly coupled across vole species and sexes. Familiarity preferences could persist despite relatively weak social reinforcement. This was not a longitudinal study of infant-to-adult regression, but it identifies a distinction that the present hypothesis requires.

Reduced attraction to social interaction in general need not entail an absence of familiar bonds. Attachment, social recognition, social reward, tolerance, and the practical use of social information should be measured separately. Their developmental coordination, or partial decoupling, may matter more than a global characterization of an animal or child as social or asocial.

4. Childhood blond hair and primate natal coats

Childhood blond-hair darkening provides a useful comparison because it makes age-dependent phenotypic change immediately visible. A child can have light hair at one developmental stage and substantially darker hair later without this change necessarily reflecting damage or requiring a change in the inherited DNA sequence. An account based only on adult pigmentation would omit part of the phenotype that requires explanation.

The observation has empirical support. Kukla-Bartoszek and colleagues (2018) studied 476 children aged six to thirteen and compared available early-childhood hair information with later coloration. Among the subgroup described as blond in early childhood, 143 children, or 70.8%, were brown-haired at the later assessment. This is a sample-specific proportion, not an estimate for all blond children. Earlier longitudinal twin research, covering three months to six years, found marked age-related changes and high monozygotic concordance, supporting genetic influence on the timing of color change (Matheny & Dolan, 1975).

Neither study directly measured the transcriptional events responsible for childhood darkening. The appropriate inference is that pigmentation is developmentally patterned, not that a particular gene-expression switch has already been identified. A separate molecular study illustrates the broader regulatory principle: Guenther and colleagues (2014) showed that variation in a KITLG enhancer can alter hair pigmentation. That experiment concerned regulatory alleles, not the mechanism by which an individual child’s blond hair darkens. Regulatory variation between individuals and developmental regulation within an individual are related questions, but not interchangeable findings.

Primate natal coats add a comparative ecological dimension. In many species, infants initially display coloration that differs from the adult coat. Caro and colleagues (2022) examined natal coloration using a 286-species phylogeny and found an association between distinctive infant coats and infanticide prevalence. They proposed that conspicuous infant coloration might elicit maternal care and shorten a vulnerable developmental phase. The association was empirical; the maternal-care mechanism remained an interpretation rather than an experimental demonstration.

Within-species observations are especially informative about timing. Bădescu and colleagues (2016) studied ursine colobus, whose coats change from white through grey to adult coloration. The complete transition averaged approximately 95 days among infants with full-transition observations, and its timing varied with sex and social-group conditions associated with infanticide risk. The study supports investigating coat maturation in relation to social circumstances, without proving that the color change itself caused a survival advantage.

The proposed comparison is therefore substantive but bounded. Childhood hair darkening demonstrates that an early phenotype can be developmentally temporary. Primate natal coats demonstrate that the timing and possible social function of a temporary phenotype can be investigated comparatively. Together, they motivate attention to developmental sequences rather than fixed adult traits. They do not establish that human blond childhood hair is a homologous natal-coat adaptation, that it elicits caregiving, or that hair pigmentation and autistic regression share a molecular pathway.

The relevance to autism is a question about developmental organization: might an early social phenotype also be modified by a regulated transition whose timing and consequences differ among individuals? The analogy justifies asking the question. Evidence from social behavior, learning, and neural regulation must determine its answer.

5. From family learning to self-directed exploration

The hypothesis proposes a possible change in the relative organization of learning. Early proximity to caregivers supplies access to nourishment, protection, familiar sensory patterns, and opportunities for observation. As motor and exploratory capacities expand, an individual can acquire information through a broader range of actions. A species’ social ecology could influence how strongly continued development depends on social guidance, independent exploration, or a combination of the two.

Human foraging development illustrates why autonomy should not be equated with absence of social learning. Salali and colleagues (2019) documented developmental changes in imitation, observation, practice, and play among BaYaka children. Learning became increasingly embedded in practice and peer activity, not detached from social life. The relevant contrast is therefore between differing balances and sources of guidance, rather than a universal switch from social to nonsocial learning.

Applied to autism, the specific proposal is that some children may undergo a change in the motivational weighting of social signals relative to environmental regularities. This could reduce spontaneous orientation toward other people while preserving some familiar attachments and some capacities for learning. Attentional autonomy would not imply nutritional independence, broad adaptive competence, or readiness to leave caregivers. A developmental component could change before the other capacities that ordinarily coordinate with it.

Some infant findings make this question tractable. Miller and colleagues (2021) observed 147 infants, including 17 later diagnosed with autism. Unusual visual inspection at nine months predicted lower social engagement at twelve months after accounting for earlier engagement; the reverse relationship was not significant. The association was observational and was not shown to differ reliably by diagnostic outcome. Unusual inspection is also not evidence of improved systemizing or ecological learning.

A complementary result limits a simple reallocation account. Ozonoff and colleagues (2010) found social-communicative decline without an accompanying increase in the duration of ordinary object-looking. Reduced social engagement therefore cannot be assumed to release a fixed quantity of attention that automatically transfers to objects. The stronger research question concerns particular learning operations, rewards, and information preferences, measured directly.

For the proposed functional interpretation to succeed, a specified subgroup would need to show more than withdrawal. There should be a coherent pattern of preserved or improved self-directed learning, together with evidence that changing social priorities help explain that pattern. If the predicted competencies do not emerge, the adaptive-reallocation account would lose support even if developmental regulation remains involved.

6. Receptors and developmental regulation

The animal comparison becomes mechanistically useful when it identifies a measurable regulatory process. Oxytocin receptors provide one candidate, but the relevant proposition concerns regional developmental organization, not a simple equation between more oxytocin and more sociability.

Hammock and Levitt (2013) measured oxytocin-receptor ligand binding at several postnatal ages in mice. A transient neocortical pattern, especially in layers II and III, peaked around postnatal day 14 and was much lower in adulthood. Because mice remain social, this result is also an essential comparative control. Developmental receptor decline alone cannot identify a program for solitude. The question is whether its timing, anatomical distribution, or functional consequences differ systematically with social ecology.

Human data establish developmental variation in the same receptor system. Rokicki and colleagues (2022) analyzed gene-expression data from 57 clinically unremarkable donors across 16 brain regions and found an early-childhood OXTR expression peak, with regional and sex-related differences. Their comparative macaque analysis also indicated different timing. These were cross-sectional developmental data, not observations of children undergoing autistic regression. A broad childhood expression peak cannot be converted into a demonstrated receptor decrease at eighteen to twenty-four months.

Direct autism tissue studies add a different kind of evidence. Freeman and colleagues (2018) examined 22 autism and 22 comparison specimens. Oxytocin-receptor binding was lower in the ventral pallidum but higher in the nucleus basalis of Meynert in autism. An exploratory age analysis suggested an altered childhood binding trajectory in the ventral pallidum. These findings support investigation of regional regulation, while contradicting an undifferentiated account of globally reduced receptor availability.

Dayley and colleagues (2026) examined adjacent sections from the same specimen collection. OXTR messenger RNA was higher in autism specimens in both regions. Binding and messenger RNA were associated in nucleus-basalis cholinergic neurons, but not in the ventral pallidum. The follow-up was not an independent replication, and the contrast between its transcriptional findings and the earlier binding results is instructive. Gene sequence, transcription, receptor availability, ligand binding, and downstream signaling must be distinguished.

The term neuromodulatory transition should consequently be used with precision. Oxytocin is a neuropeptide, OXTR is its receptor, and corticosterone is a steroid hormone that affects neural function. Measuring one does not measure the others. Vasopressin, opioid, dopamine, and stress-related systems remain plausible additional research targets, but their general relevance to social behavior does not establish a developmental mechanism for autistic regression.

The missing empirical connection is specific: no study discussed here follows an individual child through regression while demonstrating that a corresponding neural regulatory change initiates a shift toward independent ecological learning. This is the central prediction to investigate, not a result to assume.

Table 1. Different kinds of developmental evidence

Study system

Developmental comparison

Evidence obtained

Solitary hamsters

Juvenile versus adult; behavior at 2.5, 4, and 8 weeks

Behavioral and receptor-binding associations, not receptor causation

Siberian hamsters

Social maturation with experimentally altered puberty

Partial separation of developmental timing from gonadal modulation

Rat pups

Learning regulation around postnatal days 6 to 15

Causal hormonal manipulation of learning within a sensitive window

Mice

Postnatal receptor mapping through adulthood

Transient regional receptor binding, not evidence of solitude

Human brain tissue

Cross-sectional age and diagnostic comparisons

OXTR transcription and binding patterns, not within-child regression mechanisms

Sources: Beery et al. (2025); Paul et al. (2018); Upton and Sullivan (2010); Hammock and Levitt (2013); Freeman et al. (2018); Rokicki et al. (2022); Dayley et al. (2026).

7. Why a transition might occur at a particular age

A timing hypothesis must explain more than coincidence between a symptom and a familiar developmental milestone. Biological changes can be governed by chronological maturation, accumulated experience, physiological thresholds, or interactions among them. Different components of development can therefore become temporarily misaligned.

Heterochrony refers to changes in developmental timing or rate. Applied here, it could involve an earlier reduction in the priority assigned to social cues, a greater reduction, a longer duration, or an altered relationship between motivational change and the acquisition of other capacities. These possibilities predict different trajectories and should not be combined into an explanation that accommodates every outcome after the fact.

The human version of the hypothesis concerns a possible partial transition. A child could remain dependent on caregivers for food, protection, and co-regulation while becoming less reliant on their cues to organize attention. If developmental systems that ordinarily coordinate affiliation and exploration became differently timed, the resulting profile would not have to resemble either ordinary human maturation or complete adult competence in a solitary species.

The proposed functional logic is that early family engagement may be useful for establishing a foundation of learning, after which increasing self-directed practice could become valuable. However, chronological proximity to the second birthday does not demonstrate that such a foundation has been completed. The amount and kind of prior learning would need to be measured. A fixed-age switch and an experience-dependent transition are different hypotheses.

Cross-species comparison should therefore align functional milestones rather than convert weeks into months through a single scaling rule. Relevant measures include exploration, changing sources of information, separation responses, foraging practice, and the independence of particular learning tasks. Weaning and puberty provide context, but neither is an automatic proxy for cognitive autonomy.

7.1 A hypothesis of shared developmental machinery

An explicit extension of the solitary-forager account is that the ancestral regulatory machinery that retunes mammalian brains as early family dependence changes may also contribute to some forms of autistic regression. The proposed relationship is therefore mechanistic and evolutionary as well as behavioral. Similarities between developmental trajectories could arise partly because related neural systems are reorganized by inherited, partly conserved gene-regulatory networks. Different timing, intensity, or coordination of these networks could produce different outcomes across species and among individuals.

This possibility can be formulated as a hypothesis of deep homology. In evolutionary developmental biology, deep homology concerns the reuse and modification of ancient generative machinery in the production of traits whose eventual forms or functions may differ. Shubin, Tabin, and Carroll (2009) described how diverse structures can emerge through the modification of pre-existing genetic regulatory circuits. Applied here, the relevant proposed continuity concerns developmental regulation of social motivation and learning rather than identical adult behavior.

The expression “same complex of genes” should accordingly refer to an ancestrally related regulatory system, not necessarily an unchanged gene list or identical DNA sequences. Evolution may preserve core relationships while modifying regulatory inputs, cellular targets, and developmental schedules. Finding a shared receptor would identify a candidate component; demonstrating corresponding regulatory interactions and causal functions would provide stronger evidence for the proposed homology.

7.2 A process that may extend across mammalian social systems

The concentration of current evidence in laboratory rodents should not be treated as a boundary around the biological hypothesis. Hamsters provide experimentally accessible examples of changing affiliation, but related regulatory processes could operate in many solitary mammals. Their wider distribution is a prediction requiring comparative investigation, rather than a conclusion established by the hamster findings alone.

The same possibility extends to social species. A developing animal may become less dependent on its parents without becoming generally less social. Motivation can be redirected toward peers, potential mates, or other members of a group. Thus, solitary and social species need not differ only in the strength of a single decline in sociability. They may differ in how related developmental components reorganize the targets of social interest and the balance between socially guided and self-directed learning.

Under this account, the relevant machinery could also participate in ordinary human development and might be present throughout the human population. What varies would be its deployment, including the developmental stage at which particular components change and their coordination with communication, exploration, and practical competence. Some autistic trajectories could reflect a different organization of processes that also function in nonautistic development. This does not predict that every person experiences autistic regression or that a universal human transition has already been identified.

The candidate system is therefore best understood as a mechanism for developmental reorganization rather than a unitary switch for solitude. Reduced affiliation could be one outcome, redirection toward different social partners another, and increased reliance on self-directed exploration a third. Whether these outcomes share an integrated ancestral program or only partly overlapping modules is itself an empirical question.

7.3 Evidence for conserved systems and changing social priorities

Comparative neurobiology supports the general premise that diverse social behaviors can use ancient shared machinery. O’Connell and Hofmann (2012) analyzed expression profiles for 10 neurochemical genes across 12 brain regions in 88 species representing five vertebrate lineages. They found extensive conservation in the neural and gene-expression organization relevant to social decision-making. This establishes a comparative foundation for investigating inherited social-regulatory systems, although it does not identify a conserved developmental transition from family dependence to independence.

Human developmental transcriptomics offers a complementary starting point. Rokicki and colleagues (2022) identified genes whose spatial and developmental expression patterns were closely associated with OXTR, including DRD2, CD38, and OPRK1. These findings encourage investigation of coordinated systems rather than an isolated oxytocin receptor. Co-expression does not establish direct regulation or a common behavioral function, but it provides candidate relationships that can be tested across developmental stages and species.

Behavioral evidence in social mice further broadens the comparison. Harda and colleagues (2026) found that the rewarding value of sibling interaction was relatively high around postnatal day 33, lower around day 38, and higher again around day 43. Neither social contact seeking in a partition test nor cocaine-conditioned preference showed the same developmental profile. The authors proposed that changes in the reward value of familiar kin might contribute to natal dispersal and suggested possible evolutionary conservation with human adolescence. The observed result was a selective, temporary change in reward, not a general loss of sociability or a demonstrated genetic mechanism.

Human imaging research also identifies changing social priorities during ordinary development. In a cross-sectional sample aged approximately seven to sixteen, Abrams and colleagues (2022) found that younger children showed stronger responses to their mother’s voice than to unfamiliar voices in reward and social-valuation regions, whereas older adolescents showed the opposite relative pattern. The authors interpreted this reorientation in relation to the adaptive development of independence. The study measured neural responses, not the genes controlling the transition, and the reorientation favored other people rather than solitude.

An autism study provides a useful refinement. Abrams and colleagues (2026), studying participants aged seven to seventeen, found age-related patterns that differed between autistic and typically developing groups. Autistic adolescents showed decreasing neural engagement with nonfamilial voices and increasing engagement with their mother’s voice as age increased. These findings make it important to distinguish familiar attachment from generalized social motivation. They do not measure toddler regression, but they show why the proposed mechanism should allow selective changes in social priorities rather than require uniform detachment from caregivers.

These studies address different levels of the argument: evolutionary conservation, coordinated gene expression, changing reward, and developmental differences in neural responses. Together they motivate a comparative hypothesis. They do not yet demonstrate that one conserved gene-regulatory network causes both mammalian social reorientation and autistic regression. The adolescent findings also concern a later developmental window than most early autistic regression; the relationship between those windows must be tested rather than assumed.

7.4 Adaptive developmental functions and autistic regression

The proposed adaptive logic concerns changing requirements across development. Early family proximity can support protection, nourishment, attachment, and initial learning. As an individual’s capacities expand, a different allocation of attention and motivation may support exploration and the acquisition of information through its own actions. In social species, increasing autonomy can also involve learning from a broader set of partners. The hypothesis is that selection has shaped regulatory processes that coordinate such changes with species-specific ecological demands.

This formulation preserves the possibility that a widely distributed developmental system is adaptive while separating that claim from the consequences of a particular regressive trajectory. An ancestral regulatory process could contribute to impairment if its components change at an unusual time or become poorly coordinated with other abilities. Alternatively, some patterns might involve functional trade-offs, with altered social priorities accompanying preserved or enhanced learning in specified domains. These alternatives make different predictions and should not be treated as interchangeable explanations.

The adaptive-reallocation version requires evidence that the proposed change supports identifiable competencies under identifiable conditions. Withdrawal alone is insufficient to establish a benefit. Similarly, the substantial frequency of autistic regression makes developmental regulation important to investigate, but frequency does not determine evolutionary function. The central empirical task is to explain the relationship among changing motivation, retained or lost skills, and the acquisition of new capacities.

7.5 Predictions of the deep homology account

The hypothesis predicts that comparisons aligned to functional developmental milestones will reveal corresponding regulatory changes in homologous neural cell types across multiple species. Relevant milestones include changing responses to separation, shifts in the value of familiar and unfamiliar partners, and increasing competence in self-directed tasks. These comparisons should control for phylogeny and distinguish maturation from laboratory housing and separation effects.

A stronger mechanistic test would identify upstream regulators whose manipulation changes the predicted combination of social priorities and learning functions in suitable animal models. Effects limited to generalized distress, motor suppression, or reduced activity would not establish a developmental independence mechanism. Conserved causal relationships would be more informative than overlap in broadly expressed receptor genes.

In human studies, the corresponding prediction is a difference in the developmental trajectory of candidate networks among regressive autism, nonregressive autism, and typical development. Ethically accessible measurements and experimental models would need validation against the neural processes they are intended to represent. A finding in blood or another peripheral tissue should not automatically be treated as evidence of the same change in a particular brain circuit.

The proposed homology would lose support if the apparent similarities were produced by unrelated mechanisms. A candidate regulatory change that occurred only after behavioral onset would be less plausible as an initiator, although it could still influence the subsequent trajectory. Failure to observe the predicted pattern of preserved or enhanced learning would specifically weaken the adaptive-reallocation explanation. These distinctions allow the mechanistic and functional components of the hypothesis to be evaluated separately.

The childhood blond-hair and primate natal-coat comparisons illustrate why temporary early-life phenotypes merit developmental and ecological explanation. They do not supply evidence for a shared pigmentation and social-regulation pathway. The deep-homology proposal concerns the possible ancestry of the social-regulatory machinery itself. Its contribution is to extend the solitary-mammal comparison into a testable account of how related developmental systems could generate different social and learning trajectories.

8. Relation to existing developmental theories

The proposed account overlaps with several established lines of theorizing while asking a more specific comparative question. Mottron and Gagnon (2023) proposed a developmental bifurcation involving declining social-information bias, altered information processing, a plateau, and later partial normalization. Mottron and colleagues’ (2026) perception-language model further connects reduced social priority with the recruitment of learning processes by recurrent structures outside ordinary socially guided language acquisition. These are substantive predecessors to the learning-reallocation component of the present hypothesis. They do not themselves establish its proposed connection to mammalian transitions toward independent activity.

Colombi and Brüne (2026) also proposed that attention may shift from caregivers toward objects as motor abilities expand and caregiver proximity decreases. Their hypothesis emphasizes missed critical-period practice and the possibility that ancestral caregiving ecologies buffered disability. This is a close developmental parallel, but a different functional explanation from a transition toward solitary activity.

Crespi (2013) developed an evolutionary heterochrony account emphasizing the retention of relatively juvenile characteristics. That framework demonstrates the relevance of developmental timing, but a delay in some processes and earlier or stronger deployment of others are different propositions. The present account requires specifying which component changes and in which direction.

Johnson, Jones, and Gliga (2015) proposed that some autistic attentional patterns and social withdrawal could constitute compensatory niche construction by a developing brain with atypical processing characteristics. They distinguished adaptation within the individual from evolutionary adaptation. Their discussion also considered regression as potentially reflecting failed homeostatic regulation. This provides an important alternative: a child’s behavior could improve the fit between input and processing capacity without expressing an ancestral strategy for independent foraging.

The distinctive question developed here is whether some changing social phenotypes recruit or retime regulatory processes that also contribute to normal mammalian transitions away from early family dependence. Reser’s earlier comparative framework supplies the evolutionary context; developmental-regression research specifies a new target. The proposal does not require claiming that altered attention, developmental plasticity, or functional withdrawal were previously unrecognized.

9. Predictions and a comparative research program

9.1 Define the developmental phenotype before testing its explanation

Prospective studies should separate lost competence, reduced spontaneous performance, and developmental plateau. Measures should include communication under supportive conditions, independent problem solving, sensory distress, motor abilities, and everyday functioning. A subgroup should be specified before testing the hypothesis, rather than selected afterward because its behavior resembles the proposed model.

9.2 Test coordinated change rather than withdrawal alone

The learning-reallocation version predicts a relationship between declining social priority and a specified change in independent learning. Appropriate tasks could compare acquisition of spatial regularities, causal relations, object properties, and socially demonstrated information. Task difficulty and motivation should be controlled. A preference for looking at an object is not equivalent to learning more from it.

The most informative evidence would be within-child change that precedes or accompanies altered learning performance and remains predictive after accounting for general developmental level, distress, and sensory differences. Failure to find the proposed coordination would weaken the functional explanation.

9.3 Measure attachment and social motivation separately

If the transition selectively reduces generalized social priority, familiar attachment, social recognition, or some forms of supported interaction may remain relatively preserved. These predictions require direct measurement and must not be inferred from caregiver impressions alone. Conversely, extensive loss across all relationship and recognition measures would require a broader mechanism than selective motivational reweighting.

9.4 Compare social and solitary species with developmental controls

A comparative program should include closely related species that differ in adult social organization, alongside age-matched social controls. Receptor mapping should be combined with behavior, cell-type information, and carefully controlled rearing conditions. Experimental work should distinguish maturation from separation and housing effects and follow appropriate welfare safeguards.

Evidence for a conserved mechanism would require more than the involvement of oxytocin in two species. Corresponding regulatory relationships, developmental ordering, and circuit functions would need to be established. Similar outcomes produced by different mechanisms would support functional analogy or convergence instead.

9.5 Distinguish shared mechanisms from adaptive benefits

The mechanistic and evolutionary versions make different commitments. A conserved regulatory process could contribute to a disabling outcome when expressed at an unusual time or intensity. Demonstrating its involvement would not establish that the outcome was selected for.

The stronger adaptive account predicts demonstrable benefits under specified conditions, potentially involving independent acquisition of useful information or reduced costs of social dependence. Those benefits would need to be weighed against losses in communication, coordination, and practical functioning. Ecological competence must be measured across relevant tasks rather than inferred from repetitive behavior or low affiliation.

10. Scope and alternative explanations

The main difficulty is explaining true skill loss. A change in motivational priority could affect practice and spontaneous use, and a developmental change in learning could alter subsequent acquisition. Neither possibility automatically explains the disappearance of established language, motor skills, or daily-living abilities. Any adequate account must identify how the proposed regulatory process affects the particular competence that is lost.

Competing explanations include altered sensory processing, stress-related avoidance, changing demands on developing networks, and identifiable neurological or psychiatric conditions. These alternatives need not be mutually exclusive with developmental regulation. A regulatory mechanism could contribute to several pathways, only some of which resemble a coordinated learning transition.

The account is not a claim that autistic people lack attachments, that independence is universal, or that disability can be explained away as a beneficial specialization. Nor does an evolutionary hypothesis justify delaying assessment when a child loses skills or withholding communication, educational, or medical support. Scientific explanation and support needs address different questions.

11. Conclusion

Autistic regression invites investigation of how social engagement, learning, and biological regulation change together. Comparative mammalian research shows that affiliation can decline during ordinary development, that receptor distributions can reorganize, and that hormones can regulate the availability of particular learning processes. Human research identifies relevant developmental and regional variation in the same broad signaling systems, while leaving the mechanism of regression unresolved.

Childhood blond-hair darkening and primate natal-coat transitions provide a parallel way of recognizing the problem. A phenotype expressed early in life need not persist unchanged, and understanding its trajectory may require developmental as well as ecological explanation. This principle motivates the comparison without establishing shared molecular causes.

The proposed neuroecological hypothesis extends the solitary-forager framework by shifting attention from resemblance between adult traits to the processes that generate changing social phenotypes. Its decisive tests concern the coordination of social decline with specific learning changes, the developmental organization of shared mechanisms, and the distinction between preserved competence and genuine loss. A successful account would explain why a particular trajectory changes when it does, what is gained or lost, and under which conditions the comparison with mammalian social independence is biologically informative.

References

Bădescu, I., Wikberg, E. C., MacDonald, L. J., Fox, S. A., Vayro, J. V., Crotty, A., & Sicotte, P. (2016). Infanticide pressure accelerates infant development in a wild primate. Animal Behaviour, 114, 231–239. https://doi.org/10.1016/j.anbehav.2016.02.013

Beery, A. K., Lee, N. S., & Cooke, E. M. (2025). Declines in oxytocin receptor density and social behavior across a dispersal-like transition in solitary hamsters. Developmental Neurobiology, 85(4), e23005. https://doi.org/10.1002/dneu.23005

Beery, A. K., Lopez, S. A., Blandino, K. L., Lee, N. S., & Bourdon, N. S. (2021). Social selectivity and social motivation in voles. eLife, 10, e72684. https://doi.org/10.7554/eLife.72684

Caro, T., Brockelsbury, K., Ferrari, A., Koneru, M., Ono, K., Rowe, Z. W., Touche, E., Walker, H., & Stankowich, T. (2022). On the evolution of distinctive natal coat coloration in primates. American Journal of Biological Anthropology, 177(3), 530–539. https://doi.org/10.1002/ajpa.24468

Colombi, C., & Brüne, M. (2026). Ethno-evolutionary insights and missed critical periods hypothesis of Autism Spectrum Condition. Frontiers in Child and Adolescent Psychiatry, 5, 1838798. https://doi.org/10.3389/frcha.2026.1838798

Crespi, B. (2013). Developmental heterochrony and the evolution of autistic perception, cognition and behavior. BMC Medicine, 11, 119. https://doi.org/10.1186/1741-7015-11-119

Dayley, E. E., Durham, S., Palumbo, M. C., Lundell, J. F., & Freeman, S. M. (2026). Oxytocin receptor gene expression in the basal forebrain in autism: Association with receptor binding levels and single nucleotide polymorphisms. Journal of Neurodevelopmental Disorders, 18, 16. https://doi.org/10.1186/s11689-026-09678-0

Freeman, S. M., Palumbo, M. C., Lawrence, R. H., Smith, A. L., Goodman, M. M., & Bales, K. L. (2018). Effect of age and autism spectrum disorder on oxytocin receptor density in the human basal forebrain and midbrain. Translational Psychiatry, 8, 257. https://doi.org/10.1038/s41398-018-0315-3

Ghaziuddin, M. (2021). Catatonia: A common cause of late regression in autism. Frontiers in Psychiatry, 12, 674009. https://doi.org/10.3389/fpsyt.2021.674009

Guenther, C. A., Tasic, B., Luo, L., Bedell, M. A., & Kingsley, D. M. (2014). A molecular basis for classic blond hair color in Europeans. Nature Genetics, 46(7), 748–752. https://doi.org/10.1038/ng.2991

Hammock, E. A. D., & Levitt, P. (2013). Oxytocin receptor ligand binding in embryonic tissue and postnatal brain development of the C57BL/6J mouse. Frontiers in Behavioral Neuroscience, 7, 195. https://doi.org/10.3389/fnbeh.2013.00195

Johnson, M. H., Jones, E. J. H., & Gliga, T. (2015). Brain adaptation and alternative developmental trajectories. Development and Psychopathology, 27(2), 425–442. https://doi.org/10.1017/S0954579415000073

Kukla-Bartoszek, M., et al. (2018). Investigating the impact of age-depended hair colour darkening during childhood on DNA-based hair colour prediction with the HIrisPlex system. Forensic Science International: Genetics, 36, 26–33. https://doi.org/10.1016/j.fsigen.2018.06.007

Matheny, A. P., Jr., & Dolan, A. B. (1975). Sex and genetic differences in hair color changes during early childhood. American Journal of Physical Anthropology, 42(1), 53–56. https://doi.org/10.1002/ajpa.1330420106

Miller, M., Sun, S., Iosif, A.-M., Young, G. S., Belding, A., Tubbs, A., & Ozonoff, S. (2021). Repetitive behavior with objects in infants developing autism predicts diagnosis and later social behavior as early as 9 months. Journal of Abnormal Psychology, 130(6), 665–675. https://doi.org/10.1037/abn0000692

Mottron, L., Desrosiers, J., Lafrance Mcguire, O., Lestage, G., Krikorian, A.-L., Ostrolenk, A., & Gagnon, D. (2026). Autism-ness does not exist, but autism does, part 2: The perception-language chimera model accounts for the nature, chronology, and interrelation of autistic signs. Autism & Developmental Language Impairments. https://doi.org/10.1177/23969415261454212

Mottron, L., & Gagnon, D. (2023). Prototypical autism: New diagnostic criteria and asymmetrical bifurcation model. Acta Psychologica, 237, 103938. https://doi.org/10.1016/j.actpsy.2023.103938

Ozonoff, S., et al. (2010). A prospective study of the emergence of early behavioral signs of autism. Journal of the American Academy of Child & Adolescent Psychiatry, 49(3), 256–266.e2. https://doi.org/10.1016/j.jaac.2009.11.009

Ozonoff, S., Gangi, D., Hanzel, E. P., Hill, A., Hill, M. M., Miller, M., Schwichtenberg, A. J., Steinfeld, M. B., Parikh, C., & Iosif, A.-M. (2018). Onset patterns in autism: Variation across informants, methods, and timing. Autism Research, 11(5), 788–797. https://doi.org/10.1002/aur.1943

Paul, M. J., Probst, C. K., Brown, L. M., & de Vries, G. J. (2018). Dissociation of puberty and adolescent social development in a seasonally breeding species. Current Biology, 28(7), 1116–1123.e2. https://doi.org/10.1016/j.cub.2018.02.030

Pickles, A., et al. (2022). Predictors of language regression and its association with subsequent communication development in children with autism. Journal of Child Psychology and Psychiatry, 63(11), 1243–1251. https://doi.org/10.1111/jcpp.13565

Reser, J. E. (2011). Conceptualizing the autism spectrum in terms of natural selection and behavioral ecology: The solitary forager hypothesis. Evolutionary Psychology, 9(2), 207–238. https://doi.org/10.1177/147470491100900209

Reser, J. E. (2014). Solitary mammals provide an animal model for autism spectrum disorders. Journal of Comparative Psychology, 128(1), 99–113. https://doi.org/10.1037/a0034519

Rokicki, J., et al. (2022). Oxytocin receptor expression patterns in the human brain across development. Neuropsychopharmacology, 47, 1550–1560. https://doi.org/10.1038/s41386-022-01305-5

Salali, G. D., Chaudhary, N., Bouer, J., Thompson, J., Vinicius, L., & Migliano, A. B. (2019). Development of social learning and play in BaYaka hunter-gatherers of Congo. Scientific Reports, 9, 11080. https://doi.org/10.1038/s41598-019-47515-8

Tan, C., Frewer, V., Cox, G., Williams, K., & Ure, A. (2021). Prevalence and age of onset of regression in children with autism spectrum disorder: A systematic review and meta-analytical update. Autism Research, 14(3), 582–598. https://doi.org/10.1002/aur.2463

Upton, K. J., & Sullivan, R. M. (2010). Defining age limits of the sensitive period for attachment learning in rat pups. Developmental Psychobiology, 52(5), 453–464. https://doi.org/10.1002/dev.20448