Psychology by BogdanovaThe Encyclopedia of Psychology for the Artificial Era

RESEARCH DOSSIER

Developmental and Biological Influences on Sexual Orientation

Abstract. This fundamental research monograph examines developmental and biological influences on human sexual orientation, distinguishing patterns of attraction from reported sexual behavior and identity. A critical narrative synthesis integrates historical definitions, population-based twin studies, genome-wide association and familial linkage analyses, clinical observations of prenatal androgen exposure, the fraternal birth-order association, maternal immune hypotheses, and prospective longitudinal studies of childhood characteristics and later attraction. The review evaluates causal identification, phenotype validity, heritability, genetic predictability, endocrine and immunological mechanisms, prospective measurement, sampling, replication, and cultural generalizability. Genetic and developmental factors contribute to population variation, but individual orientation cannot be meaningfully determined from the genetic variants presently associated with same-sex behavior; the clinical and immune evidence supports more limited, mechanism-specific hypotheses. Conflicting conclusions of recent 2D:4D meta-analyses demonstrate the importance of publication-bias corrections and biomarker validation. The monograph identifies unresolved causal steps and proposes a research agenda of harmonized, privacy-protective longitudinal measures. This is a critical narrative scholarly synthesis, not an original experiment, preregistered systematic review, clinical guideline, or meta-analysis.

Keywords: sexual orientation; developmental psychology; biological influences; genetic architecture; twin studies; same-sex sexual behavior; sexual attraction; prenatal androgens; congenital adrenal hyperplasia; fraternal birth order; maternal immune hypothesis; NLGN4Y; prospective cohort; sexual orientation measurement; heritability; 2D:4D; longitudinal development; causal inference

Sexual Orientation as a Developmental Research Object

Sexual orientation refers to relatively enduring patterns of sexual and romantic attraction, and the scientific analysis of its origins requires a clear account of the outcome being explained. Attraction toward men, women, multiple genders, or no one, patterns of sexual behavior, self-identification, and the meanings assigned to those experiences are related yet separable variables. Sexual attraction is an especially relevant phenotype for etiological inquiry because behavioral history depends on opportunity, circumstances, relationships, and disclosure. A survey item asking whether a respondent has ever had a same-sex sexual partner is an informative behavioral measure but does not exhaust the subject of sexual orientation. The broad causal review of Bailey et al. (2016) explicitly places attraction at the center of the construct, while genomic analyses such as Ganna et al. (2019) investigate a distinct operational outcome. The scientific question must therefore be expressed in terms of the measured phenotype before any biological or developmental inference is drawn.

This dossier examines influences on the development and expression of orientation-related patterns across individuals and populations. Its central questions concern familial aggregation; the genetic architecture of same-sex behavior; evidence about prenatal endocrine and immunological development; childhood markers and longitudinal trajectories; and the causal inferences that each observational, quasi-experimental, or experimental design can support. The outcome of interest is variation in human attraction and related patterns, including bisexuality and developmental change. The relevant explanatory levels extend from populations and inheritance to gestational biology, individual maturation, measurement, and social circumstances. The review integrates these levels without assuming that a statistical predictor of a later report must itself constitute a causal developmental mechanism Bailey et al. (2016) Roselli (2018).

The scientific scope is deliberately narrower than a general encyclopedia of sexuality. Detailed conceptual distinctions among identity, attraction, and behavior belong to the adjacent dossier Sexual Orientation, Attraction, and Identity Structure. Repeated change in attraction and self-description across adulthood is the principal subject of Sexual Fluidity and Orientation Across the Life Course. The present study draws on both areas where they are necessary for interpreting early predictors and outcome measures, but it concentrates on biological and developmental evidence. It also distinguishes the development of attraction from the development of gender identity, from general sexual response, and from social acceptance of different identities. Causal claims are evaluated within each level of analysis rather than inferred from the presence of a social or biological correlate Hu et al. (2016) Xu et al. (2021).

Historical Development of Scientific Questions and Measures

Scientific descriptions of sexual variation long preceded molecular genetics and modern imaging. The major mid-twentieth-century transition involved systematic interviews and the recognition that reported behavior and experience vary along more than a simple categorical division. Kinsey, Pomeroy, and Martin published Sexual Behavior in the Human Male in 1948 and investigated sexual histories with structured interviews Kinsey et al. (1948). The historical influence of this project lies in its insistence on describing distributions of experiences and patterns rather than taking everyday labels as directly observed biological categories. The sampling procedures, social conditions of disclosure, and focus on behavior also reveal why findings about reported sexual acts must be interpreted through the research design used to collect them.

During the late twentieth century, research questions shifted toward family resemblance, early developmental correlates, biological organization, and the reliability of adult recollections of childhood experiences. The birth-order analysis by Blanchard & Bogaert (1996) tested an unusually specific family-structure prediction: numbers of older brothers were associated with male homosexuality, whereas other sibling categories did not explain the effect in the same manner. Subsequent comparative and immunological work asked whether this association had a gestational mechanism. In parallel, behavior-genetic investigators employed twin designs to estimate the sources of observed population variance; the population-based Swedish study of Långström et al. (2010) represented an important attempt to reduce self-selection and small-sample limitations. These traditions developed different causal questions: who resembles whom, which pregnancy exposures precede a later outcome, and which phenotypes are consistently measurable.

The emergence of molecular genetic methods changed the evidentiary scale and precision of the hypotheses. Genome-wide association studies test relationships between measured genetic variants and reported outcomes across large samples. The extensive study by Ganna et al. (2019) demonstrated polygenic contributions to same-sex sexual behavior while explicitly rejecting useful prediction of individual behavior from the variants measured. Follow-up reviews have emphasized the complexity of genetic architecture, the distinction between historical evolutionary hypotheses and directly tested mechanisms, and limits of extending one sexual behavior phenotype to all dimensions of orientation Felesina & Zietsch (2025). Simultaneously, prospective developmental cohorts increasingly replaced some retrospective assumptions with repeated measurement of childhood behavior and later attraction Xu et al. (2019) Xu et al. (2021).

Historical work in this field also demonstrates the importance of retaining the original language and sample structure of an experiment or survey. A clinical cohort selected for atypical prenatal androgen exposure is informative about developmental biology but is not interchangeable with a population survey. A retrospective measure of childhood gender nonconformity offers different inferential strength from prospectively recorded childhood play and activity. A linkage signal in particular families is a different form of evidence from a GWAS association, and a population variance estimate has no direct interpretation as a fixed proportion of any individual's orientation. The historical narrative therefore tracks progressive specification of the unit of analysis and increasingly explicit limitations on what data can establish Hines (2011) Sanders et al. (2021).

The Explanatory Levels: Genes, Gestation, Development, and Context

Developmental explanations operate at different temporal and causal scales. Genetic variants are inherited or arise through mutation; gene regulation and developmental physiology unfold during gestation and postnatal life; relationships and experiences affect opportunities for recognizing, expressing, and describing attraction. An exposure can shape a biological pathway without determining a unique adult outcome. A developmental process can also be influenced by multiple sources at different times. To study these possibilities, researchers must define the hypothesized causal pathway, its timing, its measurable intermediate variables, and its counterfactual alternative. The existence of a genetic association cannot establish the physiology that mediates it, and the association between a childhood characteristic and adult attraction cannot identify the origin of their correlation Hines (2011) Roselli (2018).

An adequate account distinguishes causation of predispositions, development of patterns of attraction, and self-reported identification at a particular age. The first may involve factors that cannot be observed directly in ordinary human cohorts; the second requires longitudinal or developmentally grounded inference; and the third depends upon language, reflection, and disclosure. Each presents a different measurement problem. Even a repeated categorical self-label can be stable while attraction varies in degree, and a person can experience attraction while never reporting corresponding behavior. The Add Health longitudinal analysis by Hu et al. (2016) demonstrates the value of tracking attraction and relationship histories separately rather than treating partnership status as a substitute for desire. The prospective birth-cohort work of Xu et al. (2021) further shows why developmental time and specific measures of childhood interests should be studied directly.

Population heterogeneity is a primary explanatory feature rather than an inconvenience to be averaged away. Findings in men and women can differ, and the historical literature often has comparatively sparse evidence on bisexual patterns, asexuality, and populations outside Western industrialized contexts. Genetic discovery samples likewise differ in ancestry, recruitment channel, cohort age, and willingness to report behavior. A causal synthesis must identify which subgroup and outcome support each inference and then state what remains unknown about transfer. The comparison of Samoan and Western evidence for fraternal birth order illustrates the scientific value of testing a phenomenon where cultural categories and fertility patterns differ Semenyna et al. (2017) Semenyna et al. (2023).

Critical Narrative Evidence Method and Limits of the Review

This monograph uses a critical narrative synthesis anchored in identifiable original studies, foundational research, longitudinal cohorts, population-based twin models, large genetic datasets, endocrinological clinical research, and contemporary methodological reviews. It evaluates findings by phenotype validity, temporal order, recruitment, confounding, statistical uncertainty, biological plausibility, replication, and applicability to the question actually asked. The strategy is to place explanatory claims beside the design that produced them and to compare convergent and conflicting evidence across research traditions. This work presents no original participant data, genetic analysis, new clinical experiment, preregistered systematic search, or quantitative meta-analysis. It does not assign pooled effects to studies whose published methods and estimates have not been independently combined.

For genetic evidence, the relevant quantities include family and twin resemblance, heritability under a specified model, locus-level association, and genome-wide estimates with their sample and phenotype definitions. For prenatal hypotheses, the crucial distinction is between direct or clinical exposure evidence, indirect surrogate measures, and proposed pathways. For developmental cohorts, it is temporal precedence coupled with adjustment for measured confounders and scrutiny of attrition. For neuroimaging, anatomical or functional group differences must be evaluated for sample size, multiple comparisons, and the inability of cross-sectional scans to recover prenatal developmental causes. This framework privileges convergence across designs while preserving discrepancies that reveal unanswered questions Ganna et al. (2019) Meyer-Bahlburg et al. (2008) Xu et al. (2019) Shiramizu et al. (2026).

The evidentiary synthesis has two further constraints. First, sex assignment, gender identity, partner gender, and sexual attraction can be recorded differently in historical and contemporary datasets; their definitions must be described accurately before comparing samples. Second, scientifically observed distributions give no warrant for inferring the orientation or developmental history of an identified individual. Research about group-level variance addresses explanations, not tests for assigning identities. The 2019 genome-wide study directly reported that its findings lacked meaningful individual predictive utility Ganna et al. (2019). These constraints are embedded in the interpretation of each later chapter, including those that evaluate plausible mechanistic hypotheses with presently incomplete human causal evidence.

Family Resemblance and the Logic of Twin Studies

Family studies ask whether relatives resemble one another in an orientation-related outcome more than would be expected under an appropriate comparison model. Resemblance can reflect genetic transmission, shared gestational or household contexts, and correlated reporting or recruitment. Twin studies commonly compare monozygotic twins, who share virtually all inherited sequence variants, with dizygotic twins, who share on average approximately half of segregating variants. Structural models partition population variance into additive genetic components, environmental factors shared within families, and individual-specific influences plus measurement error. The model is a statistical reconstruction conditional upon its assumptions. It does not reveal the developmental events responsible for an individual person's attractions, and the inferred variance components can differ across birth cohorts, social environments, phenotypes, and measurement methods Långström et al. (2010).

Långström, Rahman, Carlström, and Lichtenstein used a population-based Swedish adult-twin survey and analyzed lifetime same-sex partner measures in 3,826 same-sex twin pairs Långström et al. (2010). Their models estimated genetic variance contributions of approximately 0.34–0.39 in men and 0.18–0.19 in women, with wide confidence intervals. The remaining components involved individual-specific and, especially for women, some shared environmental variance. These figures are often repeated in popular discussions as if they describe the genetic determination of attraction itself. Their actual referent was variation in same-sex sexual behavior in the studied population, under the fitted biometric model. The authors explicitly emphasized uncertainty and the large contribution of nonshared influences, a category that includes both biological and social experiences along with error.

The equal-environments assumption, the selection of participating twins, and measurement error all matter for twin inference. If resemblance differs because one twin type experiences more similar exposure relevant to the measured behavior, a model may attribute some similarity to inherited variation. Conversely, unreliable self-report can inflate the residual individual-specific component. Opposite-sex twins, gender-specific models, and repeated measurements can provide complementary checks, but they introduce their own sampling and interpretive questions. Twin estimates also reflect the environment in which a trait is expressed: changing social barriers to disclosure or behavior can change observed variance and classification without requiring that genetic architecture change. Robust interpretation therefore compares twin outcomes with molecular evidence and developmental observation instead of treating one model as a complete etiological theory Långström et al. (2010) Bailey et al. (2016).

Molecular Genetics and Polygenic Architecture

The molecular-genetic approach searches for statistical associations between particular inherited variants and measured traits. Early candidate-gene and linkage approaches generated hypotheses about segments of the genome potentially shared among families, although sparse family samples and uncertain replication limited causal interpretation. More recent genome-wide association studies evaluate many variants across large datasets while controlling for population structure and extensive multiple testing. A significant genomic association is a reproducible statistical signal to investigate further, not by itself a mechanism of neural development. The relevant phenotype, ancestry, sex-stratified analysis, and discovery sample determine what the findings describe. Results from family linkage and population association also answer different questions and should not be collapsed into a single narrative of gene discovery Sanders et al. (2017) Sanders et al. (2021).

The landmark study by Ganna and colleagues included 477,522 participants and assessed same-sex sexual behavior, using data from large research cohorts Ganna et al. (2019). Five loci reached genome-wide significance in the combined analyses. Aggregate tested genetic variation accounted for an estimated 8% to 25% of variation in the behavior measures, depending on analysis and assumptions, and genetic influences overlapped only partly between men and women. The authors concluded that the measured variants could not meaningfully predict an individual's sexual behavior. These findings strongly support a distributed or polygenic architecture of the specific behavioral phenotype. They do not show that a given individual's orientation is genetically fixed, that all attractions share the same determinants, or that a single biological pathway has been identified.

A particularly difficult issue is phenotype substitution. The frequently used question of ever having had a same-sex partner depends on individual attraction but also on opportunity, relationship choice, legal and social context, and readiness to disclose intimate behavior. It can classify a person who has occasional same-sex behavior differently from someone with enduring same-sex attraction but no corresponding experiences. Large cohorts reduce sampling noise but cannot repair a mismatch between scientific construct and survey variable merely by increasing sample size. An association with the behavioral outcome may arise through pathways connected to sexual orientation, general behavioral propensity, contextual opportunity, or the interaction of several processes. A strong genomic discovery program should replicate findings against multiple attraction, behavior, and identity measures and report their interrelationships Ganna et al. (2019) Bailey et al. (2016).

Genome-wide linkage meta-analysis has maintained attention to earlier male sibling findings, including support at chromosome 8 and Xq28 in combined linkage material Sanders et al. (2021). Such regions represent linkage evidence within relevant familial datasets rather than individualized determinants. Sanders and colleagues' earlier case-control GWAS also demonstrated the small and ancestrally restricted samples of the pre-biobank phase Sanders et al. (2017). Contemporary synthesis by Felesina and Zietsch discusses polygenicity and possible evolutionary maintenance models Felesina & Zietsch (2025). Evolutionary models require particular care: a theoretical account of how genetic variation could persist in populations is distinct from a directly demonstrated causal pathway in any historical population. Pleiotropy, changing fitness environments, selection, and correlated traits must be tested rather than assumed.

Heritability, Genetic Prediction, and Causal Interpretation

Heritability is the proportion of observed population variance attributed to genetic variance under a particular model and population definition. It is not the share of an individual's orientation caused by genes. High heritability can coexist with powerful environmental interventions for many traits, and a moderate estimate does not imply that the residual variance consists of childhood parenting or conscious choices. Gene–environment correlations can arise when genetic propensities influence exposure to environments, and interactions can arise when the same inherited propensity has different phenotypic consequences across contexts. The behavioral science problem is to identify these pathways with independent measures and longitudinal or experimental leverage. The existing twin and GWAS findings are useful population evidence but cannot alone quantify such detailed pathways Långström et al. (2010) Ganna et al. (2019).

Polygenic scores aggregate associations estimated from discovery samples. Their validity depends on discovery phenotype and representativeness, the ancestry and environment of the target sample, and measurement invariance. A low predictive capacity is especially consequential for a construct that is multidimensional, developmental, and privacy-sensitive. In the sexual orientation literature, the GWAS authors explicitly cautioned against individual prediction Ganna et al. (2019). Any attempt to use a genetic score to infer someone's identity would compound the original phenotype mismatch with population transport problems and low individual discrimination. Scientific reporting should therefore present aggregate associations as population results and explain why such evidence is informative about polygenic influences without permitting personal classification.

The genetic evidence also does not resolve prenatal hormone, maternal immune, or life-course mechanisms by itself. Different variants could act through fetal development, later endocrine function, general learning processes, behavioral propensities, or numerous intermediary pathways. The existence of a genetic contribution and a prenatal association can be mutually compatible while requiring distinct causal evidence. A scientifically complete account therefore does not place genes and environment into competing, exhaustive containers. It asks which measurable components of the pathway link initial conditions, development, attraction, and behavior, and how the evidence differs across biological sexes, ages, cultures, and outcomes Hines (2011) Roselli (2018) Felesina & Zietsch (2025).

Evidence Quality Across Genetic Study Designs

The largest datasets offer improved precision and opportunities for independent replication, but their recruitment and measurement define the boundaries of generalization. Volunteer genetic cohorts may overrepresent certain ages, ancestry backgrounds, education levels, and degrees of comfort reporting intimate experiences. Retrospective survey items also combine memories, personal interpretation, and contemporary social context. Genetic associations can be altered by population stratification, relatedness, and technical artifacts if controls are inadequate. Modern association procedures address many of these problems statistically, but neither genome-wide significance nor sample size alone establishes that a result measures the exact psychological construct named in a headline Ganna et al. (2019).

The evidentiary hierarchy varies with the proposed claim. A family model is appropriate for estimating familial covariance; a GWAS identifies variant associations; a prospective endocrine dataset may test temporal exposure correlations; functional molecular studies can investigate mechanisms; and repeated self-report can estimate stability of a psychological outcome. Neither one design nor a combination of correlations proves a fully specified etiological mechanism. Repeated failure to find a large-effect single locus is compatible with a highly polygenic trait, whereas consistent modest associations do not authorize genetic determination of a particular person. The appropriate scientific synthesis retains the observed signal, the limitations of design, and the causal steps still missing Sanders et al. (2021) Felesina & Zietsch (2025).

A final issue concerns the inference from absence of evidence. A statistical estimate with broad confidence intervals can be inconclusive even when its point estimate is substantial, and a small null clinical comparison may lack power. Equally, a postulated neural pathway cannot be treated as established simply because it is biologically conceivable. Registered replication, representative populations, deliberately comparable phenotypes, and clear separation of exploratory from confirmatory analyses are essential standards for future research. These principles become even more important when genetically correlated outcomes invite ethical misinterpretation or misuse. The methodological goal is precise causal knowledge about population variation while respecting the fact that attraction and identity cannot be read off a genome Ganna et al. (2019) Bailey et al. (2016).

Prenatal Hormones and Developmental Organization

Prenatal endocrine hypotheses investigate how hormonal signals during fetal development may influence the formation of neural and behavioral systems later involved in attraction. The foundational distinction is between organizational effects, which influence developmental trajectories during sensitive periods, and activational effects, which modify responses as hormonal conditions change later in life. Animal experiments have demonstrated hormonal organization of mating and reproductive behaviors, and these findings motivated hypotheses about human sexual development. Translating results requires care because animal measures such as mounting, lordosis, partner preference, and laboratory choice do not correspond one-to-one with human romantic and sexual attraction. Roselli's neurobiological review describes the value of these mechanistic models and the limits of moving across species and distinct psychological outcomes Roselli (2018).

In humans, researchers generally cannot randomly assign prenatal androgen exposure, so causal evidence comes from unusual biological conditions, clinically ascertained groups, natural variation, and indirect correlates. A convincing developmental hypothesis must specify the hormone, the gestational window, the type of receptor signaling, the neural intermediate, and the adult measure. A difference in average interest between clinical and comparison groups can suggest the involvement of an exposure without explaining all variation within either group. Hormone concentration, receptor sensitivity, timing, medical management, and postnatal experience may be correlated or jointly relevant. Hines reviewed evidence that prenatal testosterone affects sex-typical childhood interests and may contribute to sexual orientation, while also emphasizing the contribution of experience and the uncertainty about the precise neural routes Hines (2011).

The scope of endocrine effects should be described as probabilistic. A hormone can increase the likelihood of a subsequent pattern without specifying a unique identity or attraction. Furthermore, endocrine effects on childhood play are not equivalent to endocrine effects on adult orientation: the two outcomes can covary through common causes while retaining different developmental pathways. Puberty adds further questions about the emergence of conscious attraction and behavior under later changing hormone levels, but the presence of puberty-related changes does not reveal when the underlying disposition developed. Research must compare prospective biomarkers, postnatal outcomes, and relevant confounders to establish a stronger developmental explanation Hines (2011) Roselli (2018).

Congenital Adrenal Hyperplasia as a Naturalistic Research Model

Congenital adrenal hyperplasia, particularly forms involving 21-hydroxylase deficiency, can result in elevated prenatal androgen exposure in some 46,XX individuals. This clinical circumstance provides a naturalistic opportunity to study whether atypical androgen signaling is associated with later childhood interests and sexual attraction. Meyer-Bahlburg, Dolezal, Baker, and New examined women with classical and nonclassical congenital adrenal hyperplasia across degrees of prenatal androgen excess Meyer-Bahlburg et al. (2008). The research is important because it compares groups distinguished by an endocrine developmental condition instead of relying solely on adult recollection of personality or childhood behavior. Findings have been interpreted as supporting an androgen-related contribution to variation in attraction, while the clinical context introduces interpretive constraints.

Clinical ascertainment, treatment history, genital development, medical interventions, relationships, and socialization can vary in conjunction with the underlying endocrine condition. These influences may affect reported behavior, experience, and opportunities to form relationships. Researchers must therefore distinguish prenatal exposure from later illness burden and medical or interpersonal experiences rather than attributing every group difference to fetal hormones. The presence of heterosexual, bisexual, and same-sex attraction across relevant groups also demonstrates that androgen exposure cannot operate as an individually deterministic switch. Measurement of orientation must specify attraction, fantasy, identity, or behavior; clinical study samples may be small, and selection for care can reduce representativeness of the broader population Meyer-Bahlburg et al. (2008) Bailey et al. (2016).

The CAH literature is more directly informative about endocrine-developmental hypotheses than highly indirect measures such as finger-length ratios. Its causal evidentiary strength still depends upon comparison groups, exposure classification, adequate sample size, and control of correlated experiences. It does not supply an equivalent experimental model for every orientation phenotype in men and women. Generalizing from a specific endocrine condition to all sexual orientations is therefore an extrapolation requiring corroboration. An integrated research agenda would collect prospectively recorded clinical endocrine histories, validated attraction measures, and longitudinal data while monitoring the effects of care, family environment, stigma, and cultural vocabulary on responses Hines (2011) Roselli (2018).

Digit Ratios and the Problem of Indirect Prenatal Biomarkers

The ratio of second-to-fourth finger lengths, known as 2D:4D, has often been used as a proposed proxy for prenatal hormonal signaling. Its appeal lies in the possibility of measuring a relatively accessible adult physical characteristic that may preserve developmental information. Yet the inferential chain is long: a digit ratio must validly index the relevant exposure, the exposure must alter developmental systems involved in attraction, and differences in orientation-defined groups must reflect that pathway rather than confounding or biased reporting. Distributions of digit ratios overlap substantially and are affected by variation unrelated to the psychological outcome. A population difference in 2D:4D is therefore far from an individual-level endocrine test Hines (2011) Shiramizu et al. (2026).

Swift-Gallant, Aung, Salia, Breedlove, and Puts conducted an updated meta-analysis that included 51 studies and 227,648 participants, including published and unpublished datasets Swift-Gallant et al. (2025). Their synthesis reported small mean digit-ratio differences between some homosexual and heterosexual comparison groups and found different patterns by sex and hand. This evidence contributed a broad, explicit quantitative evaluation, but the interpretation depended on methods of assessing and correcting publication bias. The relevant outcome was a comparison of group means across heterogeneous samples; even a statistically reliable mean difference would not prove that the digit ratio directly measures fetal endocrine exposure or causes adult attraction.

A 2026 reanalysis by Shiramizu, Bartoš, Jones, and Pollet examined the previously compiled data using robust Bayesian meta-analysis that incorporated uncertainty about publication bias, heterogeneity, and the existence of an effect Shiramizu et al. (2026). Their analysis found little evidence for reliable sexual-orientation group differences in 2D:4D once these uncertainties were modeled, although some narrowly specified comparisons remained under examination. This direct methodological disagreement is scientifically informative. It shows why syntheses of observational biomarkers require transparent decisions about inclusion, small-study effects, unpublished data, and model averaging. A secondary trait should not be promoted into a validated prenatal biological assay because a particular meta-analytic correction produced a nominally significant result.

The broader lesson extends to handedness, dermatoglyphics, birth size, and other physical or historical proxies sometimes interpreted as markers of early biological development. Such variables may be associated with the outcome without revealing the proposed causal pathway. A robust developmental investigation first validates the proxy against the exposure it is intended to measure, then tests the exposure–outcome pathway with independent samples and appropriate temporal controls. Where exposure validation is weak, the evidence supports at most a correlational hypothesis. The conflicting 2025 and 2026 digit-ratio syntheses illustrate why reliability and validity of the measuring instrument belong inside the biological explanation itself Swift-Gallant et al. (2025) Shiramizu et al. (2026).

Neural Correlates and the Limits of Reverse Developmental Inference

Brain imaging can investigate contemporary structural, functional, or connectivity patterns associated with measured attraction or behavior. A cross-sectional difference between groups is informative about a correlate at the time of measurement. It cannot by itself establish whether the neural feature preceded attraction, developed through repeated experiences, arose through shared causes, or reflects covariates such as age, mental health, medication, sexual activity, or sampling. Human neurodevelopment is interactive across time, and group-level imaging effects can be small relative to within-group heterogeneity. Consequently, a claim that a scan reveals the prenatal origin of orientation would exceed the temporal resolution of the design Roselli (2018).

The available literature includes studies of hypothalamic regions and later structural or network-level comparisons. Historical neuroanatomical hypotheses were influential because they proposed measurable substrates for orientation-related differences, yet anatomical sex differences, neural sexual response, gender identity, and orientation require distinct operational definitions. Neuroimaging analyses should specify the exact behavioral phenotype, preregistered or exploratory comparisons, control variables, sample composition, and correction for multiple testing. Replication in independent datasets and longitudinal measures would be necessary to establish that a candidate neural feature is developmentally antecedent rather than a correlate of lived experience Roselli (2018) Bailey et al. (2016).

A mechanistic model connecting prenatal endocrine variation to later attraction requires evidence along multiple steps: exposure measurement, receptor or cellular processes, changes in developmentally relevant circuits, and adult attraction measured independently of the exposures and intermediate variables. Animal experiments may test some of these links experimentally; human clinical observations and neuroimaging can evaluate others. Cross-species analogies can guide hypotheses but leave gaps about human subjective attraction and relational experience. The synthesis thus retains neural development as an important research pathway while evaluating particular imaging findings at the level that their design can support Hines (2011) Roselli (2018).

Fraternal Birth Order as a Developmental Epidemiological Association

The fraternal birth-order effect is a statistical association between the number of biological older brothers a male individual has and the probability of a male-oriented sexual attraction or homosexual classification. The original research problem was whether specific sibling categories predict orientation independently of total family size and other family-composition variables. Blanchard and Bogaert reported an association in their 1996 investigation that became the foundation for a larger research program Blanchard & Bogaert (1996). In that sample, each additional older brother was associated with an approximately one-third increase in the odds of the relevant male homosexual classification. An odds increment describes a population relationship under the study conditions; it is neither a thirty-three-percent probability that an additional brother will cause homosexuality nor an estimate of the proportion of individuals whose orientation was caused by family composition.

The interpretation requires a precise distinction between biological older brothers and brothers raised in the same household. If the observed association follows maternal pregnancy history even when boys were not reared with the older siblings, an intrauterine mechanism becomes more plausible than explanations centered exclusively on ordinary sibling socialization. Yet a retrospective household and sibling survey can be affected by differential recall, family-size selection, survival, parental fertility, and the classification of sexual orientation. Designs should therefore collect the sex and birth order of biological siblings, later siblings, nonbiological siblings, maternal reproductive history, and the manner in which the outcome was assessed. The causal question is whether the relevant exposure lies in the gestational history of successive male pregnancies or in other variables associated with that history Blanchard & Bogaert (1996) Bailey et al. (2016).

Fraternal birth order is a group-level probabilistic association. Most men with older brothers are heterosexual, and same-sex attraction is also observed in firstborn sons and in families without preceding sons. A birth-order association can coexist with polygenic and other developmental influences. It neither establishes a single biological cause of male sexual orientation nor explains orientation-related variation among women or all bisexual trajectories. Its scientific importance lies in the specificity of the predicted relationship and in the possibility of constructing independent tests, rather than in the idea that a person's position among siblings yields an individual diagnostic signal Blanchard & Bogaert (1996) Bailey et al. (2016).

The Maternal Immune Hypothesis and NLGN4Y

One proposed mechanistic account of the fraternal birth-order effect involves immune responses associated with successive male pregnancies. The hypothesis holds that maternal exposure to male-specific fetal proteins could induce antibodies that affect aspects of fetal neural development in later male pregnancies. A plausible test must establish both the presence of a relevant antigen-specific immune response and its temporal and functional relationship to subsequent offspring development. The Y-linked protein NLGN4Y, a member of the neuroligin family, has been investigated within this framework because of its involvement in neural synaptic organization and its sex-linked molecular identity. This mechanism is a specific research hypothesis arising from family-structure correlations rather than a settled explanation of all male sexual orientation Bogaert et al. (2018).

Bogaert and colleagues measured maternal antibodies against NLGN4Y and compared relevant groups defined by sons' orientation and reproductive histories Bogaert et al. (2018). They reported a pattern interpreted as consistent with the maternal immune hypothesis, including increased immune reactivity in the mothers of gay sons, particularly in specific older-brother circumstances. Their findings provide evidence about an immunological correlate and make the original epidemiological idea more experimentally tractable. However, maternal blood samples collected long after the relevant pregnancies cannot establish the exact timing of an antibody response, its concentration in fetal tissues, or a causal effect on developing neural circuitry. A measured antibody difference is also insufficient to show that a particular male infant acquired a specific adult pattern of attraction through that immune response.

A stringent causal program would require prospective maternal sampling across pregnancies, biologically validated antibody assays, evidence of placental transfer and target engagement, developmental assessments that avoid treating gender identity as a proxy for orientation, and independent replication across populations. Where direct prospective human testing is unavailable, animal models may assess molecular pathways but need careful interpretation of their behavioral outcomes. Another difficulty is subgroup selection: the maternal immune pathway, even if causally established for some male births, would account for only a portion of variation in the broader population. Distinct genetic, endocrine, developmental, and social processes may still be involved. The appropriate conclusion at present is that maternal anti-NLGN4Y findings support a candidate mechanistic explanation of one association while leaving the central human causal steps open Bogaert et al. (2018) Roselli (2018).

Triangulation of Birth-Order, Endocrine, and Genomic Evidence

Independent studies can help assess whether different proposed explanations concern separate pathways or partially overlapping mechanisms. Fraternal birth order derives from an epidemiological pattern of successive pregnancies; congenital adrenal hyperplasia provides endocrine-developmental variation; twin models estimate family resemblance; and genome-wide analyses identify inherited variant associations. These are distinct sources of evidence. Their coexistence supports a multilevel research architecture, but none of the datasets demonstrates that all four results operate through a shared developmental process. An integrated model would require measures that place genetic background, pregnancy order, immune biomarkers, prenatal hormones, and validated attraction outcomes into a temporally explicit design Långström et al. (2010) Meyer-Bahlburg et al. (2008) Bogaert et al. (2018) Ganna et al. (2019).

Triangulation is especially helpful where causal experiments on pregnant human participants would be inappropriate or infeasible. Researchers can test predictions that differ across proposed mechanisms: biological versus social sibling order, maternal versus paternal half-sibling patterns, endocrine clinical subgroups versus population distributions, and childhood phenotypes versus adult attraction. Evidence that several imperfect designs converge can increase confidence in a narrowly defined explanation, but common reporting biases or overlapping samples can also produce false convergence. The correct comparison therefore specifies both what each method uniquely tests and the confounds the methods share. Causal interpretation should remain proportional to the strongest independently supported link in the proposed sequence Bailey et al. (2016) Hines (2011).

The research field benefits from treating effect sizes and mechanisms as separate achievements. A replicated association can be real even when the causal pathway is unresolved. A biologically plausible pathway can be interesting even before an effect has been replicated in prospective human data. In contrast, using a candidate mechanism as if it were an established predictive test would misrepresent both. Future work can strengthen inference through preregistered outcome measures, harmonized gestational records, blinded biomarker assessment, and transparent correction for the number of hypotheses tested. This evidence-oriented approach also makes it possible to revise explanations when independent findings, such as fertility analyses or improved biomarker meta-analyses, constrain earlier claims Semenyna et al. (2023) Shiramizu et al. (2026).

Childhood Developmental Correlates and Prospective Cohort Evidence

Research on childhood development frequently examines gender-related interests, play patterns, and behavioral nonconformity as statistical correlates of later attraction. These measures must be specified without turning a child’s manner of dress, play, friends, or self-expression into an individual prediction of sexual orientation. Prospective research is especially valuable because childhood characteristics are recorded before later attraction reports, reducing the retrospective reconstruction that can accompany adult accounts of childhood. The Avon Longitudinal Study of Parents and Children has supplied such information through repeated developmental assessments. Xu, Norton, and Rahman analyzed prospective early-life and adolescent data to test associations between measured childhood characteristics and later orientation-related outcomes Xu et al. (2019). The study design permits stronger conclusions about temporal order than a single retrospective adult questionnaire while retaining possible confounding and attrition.

Childhood gender nonconformity can be statistically associated with later same-sex attraction, but the pathway remains complex. A common developmental antecedent may influence both childhood interests and later attraction; social responses to those interests may affect disclosure or the language used in reporting; and the measured characteristics may be imperfect indicators of broader developmental processes. Prospective association alone cannot choose between these accounts. Xu, Norton, and Rahman subsequently studied whether prospectively recorded childhood gender nonconformity predicted stability and change in self-reported sexual orientation from adolescence to young adulthood within the same cohort Xu et al. (2021). Their work supports analysis of within-person trajectories and the difference between prediction of a later report and identification of a unique biological cause.

Childhood gender-typed behavior, gender identity, and sexual orientation remain distinct constructs. This distinction is necessary when interpreting clinical samples and studies where investigators retrospectively group children by later orientation. A child's preferences can reflect multiple interacting influences and may change without implying an underlying orientation of any specific kind. Researchers should avoid using nonconformity scores as if they were clinical screening instruments for sexual orientation. The scientifically sound use of these variables is to test population-level hypotheses with validated measures, independently measured later outcomes, and explicit models of selection, reporting context, and cultural change Hines (2011) Xu et al. (2019) Xu et al. (2021).

Adolescence, Attraction, and Continuity Over Time

Puberty creates biological changes in endocrine activity, sexual motivation, and bodily experience while social contexts also influence recognition and description of attraction. Researchers must distinguish the emergence of self-aware attraction, the first reported identity, and the first sexual behavior from a presumed onset of an underlying orientation-related disposition. A developmental panel can reveal how people describe their attractions across survey waves, but the timing of initial reporting is affected by question wording, private experience, social safety, and opportunities for relationships. A robust model therefore describes development across multiple measures rather than fixing a universal age at which orientation becomes fully formed Bailey et al. (2016) Hu et al. (2016).

Hu, Xu, and Tornello used longitudinal data to investigate the stability of reported same-sex and both-sex attraction between adolescence and young adulthood Hu et al. (2016). Their analysis demonstrates why repeated attraction reports are essential to distinguish stable patterns from reported changes and why attraction toward more than one sex cannot be adequately represented by an exclusively heterosexual–homosexual binary. A reported shift between waves can reflect a change in attraction, increased self-understanding, a more accurate or safer disclosure, or measurement variability. The study illuminates longitudinal patterns; by itself it cannot identify which causal process produced any individual transition. The same limitation applies to studies that compare identity labels but omit attraction measurements.

Attraction, behavior, and labels can follow partly different trajectories. Some people experience relatively stable attraction while reporting changing identities; others report changes in attraction over time; still others never engage in behavior corresponding to a reported attraction. The developmental task for research is to estimate the frequency and predictors of these trajectories while examining their dependence on age, measurement intervals, population recruitment, and social opportunity. The adjacent study of Sexual Fluidity and Orientation Across the Life Course addresses longitudinal variability in greater detail. Here, it matters because developmental and biological hypotheses must explain the phenotype at the time it was measured and should be assessed against repeated measurements whenever possible Hu et al. (2016) Xu et al. (2021).

Sex Differences, Bisexual Patterns, Culture, and Generalizability

Research samples differ in the numbers of male and female participants, and some early theories were developed primarily from studies of men. A finding about fraternal birth order in male androphilia cannot simply be applied to women, and findings from congenital adrenal hyperplasia in 46,XX clinical populations cannot be generalized to every male developmental pathway. Likewise, a genome-wide association with same-sex behavior measured across sexes may exhibit partially distinct genetic associations by sex without implying distinct biological determinism. Ganna and colleagues observed imperfect genetic correlation between male and female reported same-sex behavior, a result requiring careful attention to the behavior-specific outcome Ganna et al. (2019). Population distributions must be reported in the terms and measurements used by their source studies.

Bisexuality and attraction involving more than one gender pose a particular construct challenge. A dichotomous indicator of any same-sex behavior collapses potentially different patterns of attraction into a single outcome, and one-dimensional measurement may omit intensity, timing, and relationships to multiple genders. For populations in which sex and gender are described through culturally specific categories, translation of Western instruments can also erase relevant distinctions. Researchers need multidimensional, privacy-protective measures and procedures for testing whether scores represent comparable constructs across groups. A genetic association or childhood correlate that is identified with one broad survey item should not be presented as an established mechanism for bisexuality or asexuality without directly measured evidence Bailey et al. (2016) Hu et al. (2016).

The external validity problem becomes more acute when data from large Western biobanks are treated as universal. Ancestry differences influence genome-wide associations and the portability of genetic scores. Historical cohort differences affect willingness to disclose behavior and the availability of relationships. Cross-cultural reproductive findings, including Samoan investigations, can challenge simple fertility and sibling explanations while raising their own questions about local categories and sampling Semenyna et al. (2017) Semenyna et al. (2023). Responsible generalization requires explicit descriptions of recruitment, age, ethnicity or ancestry where relevant to the method, sexual orientation measure, legal and social context, and the population to which the authors intend to infer. Large samples improve precision for the population and phenotype actually studied.

Research Ethics, Clinical Interpretation, and the Limits of Prediction

Scientific investigation of sexual orientation involves sensitive information about attraction, behavior, identity, genetic background, and family history. Research designs should protect confidentiality, avoid unnecessary collection of identifying data, separate consent for genomic or biological material from consent for publication or data sharing, and account for the potential consequences of inadvertent disclosure. Ethical quality also depends on how results are communicated. A population association with prenatal conditions is not a clinical diagnostic test, and a genomic association with behavior is not an instrument for inferring an identified individual's identity. Ganna and colleagues explicitly reported that genetic findings lacked meaningful individual predictive value Ganna et al. (2019). The scientific credibility of a monograph depends on keeping such boundaries clear even where biological hypotheses attract broad public interest.

The causal origins of an orientation do not determine its moral worth or the legitimacy of a person's self-description. Etiological research is pursued to understand developmental processes and population diversity, and the evaluation of a proposed mechanism concerns evidence and measurement. Natural variability in sexual attraction is not, in itself, a condition requiring medical correction. Clinical interpretation becomes relevant when someone is experiencing distress, coercion, stigma-related harm, or a separate medical concern; the developmental evidence reviewed here does not justify an attempt to alter someone's orientation. In scientific terms, the finding that a predictor is associated with an outcome does not establish a benefit from intervening on that predictor, particularly when causal pathways and potential harms are unresolved Bailey et al. (2016).

A reliable future research program needs more representative longitudinal cohorts, multiple attraction measures, ancestry-aware genomic analysis, independent replication, and transparent evidence about recruitment and attrition. For prenatal hypotheses, prospective biomarkers and validated mechanisms would strengthen causal accounts. For birth-order research, the biological structure of families and appropriate cultural comparisons should be recorded directly. For clinical endocrine research, treatment and social experience should be distinguished from prenatal hormonal exposure. For secondary markers such as digit ratios, the marker must first be validated as a measure of the proposed exposure. Publication of analyses, models, and limitations enables subsequent researchers to separate durable findings from historically appealing but weakly identified explanations Hines (2011) Semenyna et al. (2023) Shiramizu et al. (2026).

Conclusion

The collective evidence supports a multifactorial developmental account of variation in sexual orientation-related phenotypes. Family and twin designs indicate that genetic and individual-specific influences contribute to population variation in reported same-sex behavior, although biometric estimates depend on measurement and model assumptions Långström et al. (2010). Large-scale association analysis demonstrates a polygenic component to the specific same-sex sexual behavior measure while ruling out meaningful individual genetic prediction with the reported data Ganna et al. (2019). Clinical endocrine evidence indicates that prenatal hormonal environments can be associated with later attraction-related variation in selected populations Meyer-Bahlburg et al. (2008). Fraternal birth-order findings identify an epidemiological association in men, and maternal immune work supplies a testable candidate mechanism that requires more direct causal verification Blanchard & Bogaert (1996) Bogaert et al. (2018) Semenyna et al. (2023).

These findings also delineate what has not been resolved. No available single mechanism explains the full diversity of human attractions, developmental trajectories, and identity structures. Biometric heritability cannot reveal a person's cause; genome-wide associations do not supply a unique genetic program; clinical endocrine contrasts are vulnerable to correlated exposures; and maternal immune hypotheses have not fully demonstrated a prospective fetal antibody-to-attraction pathway. Secondary physical proxies are especially disputed, as the different conclusions of the 2025 digit-ratio meta-analysis and its 2026 bias-adjusted reanalysis illustrate Swift-Gallant et al. (2025) Shiramizu et al. (2026). Evidence about one sex or one behavioral measure does not automatically transfer to all orientations or social contexts.

The strongest direction for future science is a coordinated longitudinal framework in which biological exposures are measured as close as possible to their relevant developmental windows and orientation-related outcomes are assessed repeatedly, sensitively, and multidimensionally. Such work would integrate family records, molecular data, endocrine histories, validated psychological measurement, and cultural context while explicitly modeling selection and uncertainty. This is a scientific program for explanatory precision rather than a promise of individual prediction. The fundamental conclusion is that sexual orientation is a multidimensional developmental phenomenon whose population variation reflects interacting biological and developmental influences; the evidentiary strength of each proposed influence depends upon its exact phenotype, design, replication, and demonstrated causal steps Bailey et al. (2016) Ganna et al. (2019) Xu et al. (2021).

References

  • Bailey, J. M., Vasey, P. L., Diamond, L. M., Breedlove, S. M., Vilain, E., & Epprecht, M. (2016). Sexual orientation, controversy, and science. Psychological Science in the Public Interest, 17(2), 45–101. https://doi.org/10.1177/1529100616637616
  • Blanchard, R., & Bogaert, A. F. (1996). Homosexuality in men and number of older brothers. American Journal of Psychiatry, 153(1), 27–31. https://doi.org/10.1176/ajp.153.1.27
  • Bogaert, A. F., Skorska, M. N., Wang, C., Gabrie, J., MacNeil, A. J., Hoffarth, M. R., VanderLaan, D. P., Zucker, K. J., & Blanchard, R. (2018). Male homosexuality and maternal immune responsivity to the Y-linked protein NLGN4Y. Proceedings of the National Academy of Sciences, 115(2), 302–306. https://doi.org/10.1073/pnas.1705895114
  • Felesina, T., & Zietsch, B. P. (2025). Emerging insights into the genetics and evolution of human same-sex sexual behavior. Trends in Genetics, 41(5), 402–411. https://doi.org/10.1016/j.tig.2024.12.005
  • Ganna, A., Verweij, K. J. H., Nivard, M. G., Maier, R., Wedow, R., Busch, A. S., Abdellaoui, A., Guo, S., Sathirapongsasuti, J. F., 23andMe Research Team, Lichtenstein, P., Lundström, S., Långström, N., Auton, A., Harris, K. M., Beecham, G. W., Martin, E. R., Sanders, A. R., Perry, J. R. B., . . . Zietsch, B. P. (2019). Large-scale GWAS reveals insights into the genetic architecture of same-sex sexual behavior. Science, 365(6456), eaat7693. https://doi.org/10.1126/science.aat7693
  • Hines, M. (2011). Gender development and the human brain. Annual Review of Neuroscience, 34, 69–88. https://doi.org/10.1146/annurev-neuro-061010-113654
  • Hu, Y., Xu, Y., & Tornello, S. L. (2016). Stability of self-reported same-sex and both-sex attraction from adolescence to young adulthood. Archives of Sexual Behavior, 45(3), 651–659. https://doi.org/10.1007/s10508-015-0541-1
  • Kinsey, A. C., Pomeroy, W. B., & Martin, C. E. (1948). Sexual behavior in the human male. W. B. Saunders. https://wellcomecollection.org/works/se8hb577
  • Långström, N., Rahman, Q., Carlström, E., & Lichtenstein, P. (2010). Genetic and environmental effects on same-sex sexual behavior: A population study of twins in Sweden. Archives of Sexual Behavior, 39(1), 75–80. https://doi.org/10.1007/s10508-008-9386-1
  • Meyer-Bahlburg, H. F. L., Dolezal, C., Baker, S. W., & New, M. I. (2008). Sexual orientation in women with classical or non-classical congenital adrenal hyperplasia as a function of degree of prenatal androgen excess. Archives of Sexual Behavior, 37(1), 85–99. https://doi.org/10.1007/s10508-007-9265-1
  • Roselli, C. E. (2018). Neurobiology of gender identity and sexual orientation. Journal of Neuroendocrinology, 30(7), e12562. https://doi.org/10.1111/jne.12562
  • Sanders, A. R., Beecham, G. W., Guo, S., Dawood, K., Rieger, G., Badner, J. A., Gershon, E. S., Krishnappa, R. S., Kolundzija, A. B., Duan, J., MGS Collaboration, Gejman, P. V., Bailey, J. M., & Martin, E. R. (2017). Genome-wide association study of male sexual orientation. Scientific Reports, 7, Article 16950. https://doi.org/10.1038/s41598-017-15736-4
  • Sanders, A. R., Beecham, G. W., Guo, S., Badner, J. A., Bocklandt, S., Mustanski, B. S., Hamer, D. H., & Martin, E. R. (2021). Genome-wide linkage study meta-analysis of male sexual orientation. Archives of Sexual Behavior, 50(8), 3371–3375. https://doi.org/10.1007/s10508-021-02035-3
  • Semenyna, S. W., VanderLaan, D. P., & Vasey, P. L. (2017). Birth order and recalled childhood gender nonconformity in Samoan men and fa'afafine. Developmental Psychobiology, 59(3), 338–347. https://doi.org/10.1002/dev.21498
  • Semenyna, S. W., Gómez Jiménez, F. R., VanderLaan, D. P., & Vasey, P. L. (2023). Male androphilia, fraternal birth order, and female fecundity in Samoa: A 10-y retrospective. Proceedings of the National Academy of Sciences, 120(50), e2313284120. https://doi.org/10.1073/pnas.2313284120
  • Shiramizu, V., Bartoš, F., Jones, B. C., & Pollet, T. V. (2026). Little evidence for group differences in 2D:4D ratios based on sexual orientation after adjusting for publication bias. Hormones and Behavior, 180, 105921. https://doi.org/10.1016/j.yhbeh.2026.105921
  • Swift-Gallant, A., Aung, T., Salia, S., Breedlove, S. M., & Puts, D. (2025). Sexual orientation is associated with 2D:4D finger length ratios in both sexes: An updated and expanded meta-analysis. Frontiers in Psychology, 16, 1559158. https://doi.org/10.3389/fpsyg.2025.1559158
  • Xu, Y., Norton, S., & Rahman, Q. (2019). Early life conditions and adolescent sexual orientation: A prospective birth cohort study. Developmental Psychology, 55(6), 1226–1243. https://doi.org/10.1037/dev0000704
  • Xu, Y., Norton, S., & Rahman, Q. (2021). Childhood gender nonconformity and the stability of self-reported sexual orientation from adolescence to young adulthood in a birth cohort. Developmental Psychology, 57(4), 557–569. https://doi.org/10.1037/dev0001164