RESEARCH DOSSIER
Sexual Fluidity and Orientation Across the Life Course
Abstract. Sexual fluidity is investigated here as within-person variation across time or context in attraction, sexual identity, erotic responsiveness, and relational expression. This fundamental research monograph reconstructs the historical formation of multidimensional sexuality measurement; examines intensive longitudinal cohorts, large population panels, adolescent and adult developmental studies, and prospective–retrospective measurement comparisons; and critically evaluates the statistical and theoretical interpretation of reported change. The central questions concern which dimensions change, how change relates to age, relationships, identity language, and social conditions, and what repeated observations can establish about psychological continuity. The review draws on original studies and established scholarly syntheses, with distinct attention to recruitment, starting categories, observation intervals, attrition, measurement invariance, and causal ambiguity. Evidence documents both widespread continuity and identifiable mobility in reported sexual identity and attraction across studied populations. Apparent mobility rates differ substantially by measurement schedule and baseline population; identity transitions do not automatically imply corresponding changes in attractions or behavior. The synthesis establishes a multidimensional life-course framework for critical scientific interpretation, outlines current evidence gaps concerning later adulthood and explanatory mechanisms, and identifies priorities for harmonized longitudinal research. The method is a critically structured narrative synthesis of existing scientific literature rather than a registered systematic review or meta-analysis.
Keywords: sexual fluidity; sexual orientation; sexual orientation identity; attraction; longitudinal sexuality research; life-course development; identity mobility; bisexuality; plurisexuality; sexual identity development; romantic relationships; within-person change; population panels; measurement invariance; repeated measures; retrospective recall; social context; sexuality research methods
Scientific Overview and Scope
Sexual fluidity across the life course denotes within-person variation over time or across contexts in sexual attraction, erotic responsiveness, sexual self-description, and sometimes the gender composition of romantic or sexual partnerships. These dimensions must be observed separately before their associations are interpreted. Sexual orientation is often experienced and reported with considerable continuity, yet continuity is compatible with variation in intensity, expression, labels, and relationship circumstances. A defensible scientific account begins with trajectories of specified variables rather than treating every categorical change as transformation of a single underlying essence. This monograph examines how people experience, name, and express attraction over adolescence, adulthood, and later life, and how longitudinal designs identify both stability and change. Its core materials are original repeated-measure cohorts, national panels, developmental investigations, and methodological comparisons, integrated through critical narrative scientific synthesis Bailey et al. (2016) Lilly et al. (2024) Hansen et al. (2024).
The research questions concern which dimensions vary, when variation is observed, how much is attributable to measurement schedules and social categories, and how researchers distinguish developmental processes from changing possibilities for disclosure and partnership. A respondent who selects a different identity label after six years, someone who reports a temporary shift in the relative intensity of attractions, and someone whose partner's gender differs from a prior partner's gender represent distinguishable observations. Their histories can coexist without contradiction. Individual continuity is estimated from repeated observations; population prevalence is a cross-sectional statistic. Treating these quantities as interchangeable obscures the actual temporal problem and inflates apparent disagreement among studies Diamond et al. (2020) Savin-Williams & Ream (2007) Mock & Eibach (2012).
The monograph gives priority to psychological mechanisms and evidence quality. It addresses trajectories of attraction and identity, contextual responsiveness, identity development, recall, categorization, cultural norms, and the statistical interpretation of movement between labels. Evidence on biological contributions to the origin of orientation is discussed only where it clarifies developmental inference; its comprehensive treatment belongs to a neighboring research object. Clinical pathology, diagnostic categorization, and interventions to alter orientation do not define sexual fluidity. The central scientific outcome is a structured explanation of why stable patterns and meaningful variation are both observed, how to measure each without collapsing distinct constructs, and what longitudinal evidence can support about development Bailey et al. (2016) Rosario et al. (2006) Ranker et al. (2025a).
Conceptual History and the Formation of the Research Problem
Earlier approaches frequently placed sexual orientation on a continuum or represented sexuality through a single enduring classification. The important historical advance was the recognition that experiences, fantasies, desire, behavior, emotional preference, social identity, and relationships could diverge within one person's history. Klein, Sepekoff, and Wolf's multivariable proposal explicitly treated orientation as dynamic across time and across facets, furnishing a conceptual ancestor to later multidimensional measurement. Such descriptive frameworks were valuable for organizing phenomena, although their components did not by themselves establish that each facet changed or determine the mechanisms of any change. Scientific progress required repeated observation of persons and a clear distinction between what participants felt, did, and called themselves Klein et al. (1985) Bailey et al. (2016).
Developmental investigators subsequently tracked sexual identity milestones, attraction, and behavior at multiple waves, enabling researchers to ask whether categories functioned as stable indicators or as evolving self-descriptions. Rosario and colleagues studied identity development among sexual-minority youth, while Savin-Williams and Ream investigated the stability of different orientation components in adolescence and young adulthood. These efforts shifted the inquiry from retrospective life-story labels toward observations embedded in developmental periods and social circumstances. They also revealed a persistent inferential problem: an identity transition might reflect a changed attraction profile, a newly available word, disclosure to a researcher, or a previously unexpressed understanding. Longitudinal measurement reduces some ambiguities while introducing its own problems of attrition, interval length, and question wording Rosario et al. (2006) Savin-Williams & Ream (2007).
Lisa Diamond's ten-year study of women who initially identified as lesbian, bisexual, or unlabeled became a critical reference because it compared rival accounts of bisexuality with repeated data rather than treating later labels as proof of a presumed transitional sequence. The observed patterns provided support for bisexuality as an orientation and for individual fluidity while challenging a model in which bisexuality merely precedes identification as lesbian or heterosexual. Subsequent work compared distinct operational forms of fluidity, making clear that the term itself does not denote one uniform trait. The history of this field is therefore the history of improved questions: from whether sexuality can change to which observable aspects change, at what temporal resolution, in which populations, and under which measurement conditions Diamond (2008) Diamond et al. (2020).
Multidimensional Sexual Orientation and Its Observable Components
Sexual attraction describes the direction, intensity, and sometimes context-dependent patterning of desire toward people of different genders. Sexual identity refers to the terms by which an individual understands or communicates their sexuality. Sexual behavior refers to enacted sexual activities and the characteristics of partners, while romantic attraction and intimate attachment may follow partially distinct temporal courses. A single survey category cannot represent this entire configuration. In particular, observed behavior is constrained by access to partners, relationship status, personal decisions, safety, and opportunity, and therefore cannot be assumed to disclose the full distribution of an individual's attractions Klein et al. (1985) Savin-Williams & Ream (2007) Bailey et al. (2016).
Operational distinctions matter when reports appear discrepant. A respondent may experience attractions to more than one gender while selecting a lesbian, gay, bisexual, pansexual, queer, heterosexual, or no label; another may select the same label across waves while the relative intensity of attractions varies. Relational histories also accumulate across years even when contemporary attractions differ from earlier ones. Measures should indicate the recall period, the people or genders to whom attraction refers, whether emotional and sexual attractions are queried separately, whether labels are open-ended or constrained, and whether behavior means ever, past year, or current. Without those specifications, changes in a questionnaire's output cannot be identified with changes in its target construct Diamond et al. (2020) Ott et al. (2011) Hansen et al. (2024).
Convergence among dimensions is a research question. The correlation between attraction and identity may vary by age, disclosure environment, and wording, while the relationship between identity and partnership can be mediated by opportunity and social recognition. Researchers should report joint transitions—for example, whether a label changed while the attraction distribution remained similar—rather than assuming that a single identity change implies complete psychological reorganization. This multidimensional perspective also protects the interpretation of bisexuality and other plurisexual orientations: attractions involving multiple genders can be stable as a pattern even when experiences and partnerships vary. Variation within a dimension and continuity of the broader orientation profile can coexist Diamond (2008) Diamond et al. (2020) Savin-Williams et al. (2012).
Four Operational Forms of Sexual Fluidity
The term sexual fluidity has been applied to distinct empirical phenomena, and its usefulness depends on transparent definition. Diamond and colleagues directly compared four types in a study of 76 women: overall erotic responsiveness to a less-preferred gender, situational variation in that responsiveness, divergence between the gender patterning of attraction and partnering, and day-to-day instability in attractions. These measures identify different properties of a sexual profile. The first concerns an overall degree of responsiveness, the second context sensitivity, the third cross-domain discrepancy, and the fourth temporal volatility at short intervals. Label changes constitute another commonly studied phenomenon that should be named sexual identity mobility rather than silently treated as equivalent to all four Diamond et al. (2020).
Operational separation changes scientific interpretation. Someone with consistently bisexual attractions can show little short-term variability; someone with a predominantly same-gender attraction profile can report context-dependent responsiveness without revising their identity. A person who enters a different-gender relationship may exhibit a behavior-attraction discrepancy even while attraction remains stable. These are analytically different configurations with distinct causal candidates. A scale that simply asks whether respondents consider themselves sexually fluid measures a self-understanding or metabelief, not necessarily daily attraction variability. Conversely, frequent fluctuations recorded in daily diaries need not produce a categorical label transition or an autobiographical judgment that one's orientation has changed Diamond et al. (2020) Katz-Wise & Hyde (2015).
Scientific precision requires naming the outcome in every result: label transition, attraction change, situational responsiveness, behavior-attraction discrepancy, or partnership transition. Researchers should also distinguish magnitude, frequency, direction, and persistence. An occasional small change around an otherwise continuous profile differs from repeated transitions across broad categories; neither provides an automatic ranking of authentic identity. A useful synthesis tests whether distinct forms share predictors and how they combine across the life course, instead of treating fluidity as a unitary disposition. This conceptual rule governs the evidence appraisal throughout the monograph and explains why estimates derived from different research programs should not be combined as though they measured one common rate Diamond et al. (2020) Ranker et al. (2025a) Lilly et al. (2024).
Life-Course Time, Developmental Context, and Cohort
Life-course analysis studies change as a function of age, historical time, social institutions, and the sequencing of relationships and opportunities. Chronological age alone cannot identify developmental causation. Adolescents may encounter new attraction experiences while learning available social labels, adults may enter and leave partnerships, and older individuals may reinterpret past experiences under contemporary cultural vocabularies. A panel starting in 2014 and ending in 2022 also spans changes in public terminology and disclosure environments. The same measured difference can therefore involve age-related development, period-specific social shifts, or cohort differences in exposure to concepts and institutions Hansen et al. (2024) Hu & Denier (2023) Lilly et al. (2024).
The timing of measurement defines what can be observed. A survey repeated every seven years may register only the end points of a trajectory, missing interim movements and returns; a diary repeated daily reveals local variability but may sample a short and highly distinctive life interval. Within-person processes also operate at multiple scales: moments of attraction, months of self-reflection, years of identity description, and decades of cumulative relational history. An adequate developmental model specifies the time scale before estimating a transition. If an individual moves between two categories and later returns, a two-wave panel might record no change. The absence of a detected transition is therefore conditional on the observations made Diamond et al. (2020) Ott et al. (2011) Ranker et al. (2025a).
An important consequence is that no single percentage can summarize how often sexual orientation changes over a human life. Estimates depend on initial category composition, age distribution, survey modes, response options, interval lengths, and the dimension assessed. Long-term repeated-measure studies show substantial continuity in many populations alongside observable mobility in defined subgroups. Evaluating the whole life course requires comparable measures across early, middle, and later adulthood and explicit treatment of differential attrition and changing social context. The strongest available evidence supports a plural developmental landscape in which continuity and variability have distributions rather than one invariant sequence Mock & Eibach (2012) Lilly et al. (2024) Hansen et al. (2024).
Theoretical Accounts of Stability and Variation
Several theoretical programs explain different parts of the observed phenomena. Stable-orientation accounts emphasize persistence in broad patterns of attraction, while dynamical systems approaches consider how individual responses emerge through interactions among dispositions, experience, context, and time. Identity-development perspectives study the acquisition, revision, and social use of self-descriptions; social-context models study disclosure constraints, relational possibilities, and normative expectations. These models occupy distinct explanatory levels, and a credible synthesis identifies what each predicts about separate measured variables. Diamond's dynamical analysis draws attention to changing desire and the possibility that temporal variation itself carries psychological information Diamond (2012) Bailey et al. (2016) Rosario et al. (2008).
Theories become more informative when their implications are framed as competing observable predictions. A model of attraction reorganization would predict persistent shifts across repeated attraction measures, even when labels remain unchanged. A semantic-reclassification account predicts label transitions with substantial continuity in reported attractions. A context-responsiveness account predicts within-person fluctuations associated with situations or relationship characteristics, while a reporting-safety account predicts disclosure-linked changes when privacy and social support vary. These predictions can be evaluated using repeated attraction ratings, identity interviews, social-context measures, and temporal ordering, preferably in cohorts with planned follow-up and measurement-invariance testing. Existing observational work does not invariably distinguish them Diamond et al. (2020) Ranker et al. (2025a) Hansen et al. (2024).
Developmental inference should also respect the difference between individual experiences and average associations. A statistical association between age and identity mobility does not imply that age causes orientation change in a particular person. Likewise, a partnership event can coincide with a newly reported attraction while leaving their causal sequence undetermined. Theories should seek mechanisms that predict temporal covariance, allow for bidirectional influences, and specify falsifiable boundaries. The cumulative record is most intelligible when broad continuity, attraction variability, social labeling, and relational history are modeled together without forcing one of them to serve as a proxy for the others Hu & Denier (2023) Lilly et al. (2024) Diamond (2012).
Adolescence and the Development of Sexual Self-Understanding
Adolescence is a developmental interval during which sexual and romantic attractions, access to social categories, peer relationships, and opportunities for self-description can change together. The scientific problem is to observe each process without interpreting an early category as a prediction that an individual must follow a particular later sequence. In repeated studies of gay, lesbian, and bisexual youths, Rosario and colleagues identified more than one pattern of identity continuity and transition, including consistent lesbian or gay identification, consistent bisexual identification, and movement between labels. These patterns established that identity trajectories among young people require prospective description. They did not show that every identity transition reflects a corresponding reorganization of attraction or that all adolescents share one developmental path Rosario et al. (2006) Rosario et al. (2008).
Large cohorts add an important population perspective. Ott and colleagues used repeated reports from adolescents and young adults to measure movement across identity categories, showing why analysts should calculate both the number of transitions and the identities from which those transitions originate. A young person beginning in a widely used identity category has a different statistical probability of displaying a change from someone initially reporting a rare or ambiguous label. Response options such as mostly heterosexual, bisexual, questioning, or unlabeled can reveal aspects of experience that binary response formats conceal. For this reason, rates of adolescent mobility are not portable across studies unless sample composition, categories, developmental ages, and wave intervals are sufficiently similar Ott et al. (2011) Savin-Williams & Ream (2007).
The interpretive challenge extends beyond surveys. Young people may develop a vocabulary for existing attractions, change the emphasis they place on romance and sexuality, or encounter relationships that make formerly abstract possibilities personally relevant. These are plausible psychological pathways, not mechanisms established by identity labels alone. Researchers can investigate them by combining confidential repeated measures with qualitative accounts of meaning, including when participants first recognized an attraction, first used a term privately, and first disclosed it publicly. Temporal ordering must be established independently in order to distinguish recognition, expression, and experience. The guiding principle is longitudinal descriptive accuracy in a population undergoing multiple simultaneous developmental transitions Rosario et al. (2006) Ranker et al. (2025a) Bailey et al. (2016).
Short-Interval Studies and Emerging Adulthood
Short intervals reveal changes that coarse panels cannot resolve. Ranker and colleagues followed 1,628 U.S. adolescents and young adults ages 14–25 through two survey waves separated by approximately two months; 11.4% selected a different sexual orientation identity at follow-up. The result is a two-wave category-transition proportion in a longitudinal online cohort, and its interpretation depends on the baseline distribution, survey options, and participant recruitment. Identity mobility was particularly common among respondents who initially selected uncertainty or a write-in category. Movement between terms therefore cannot be interpreted as an equal degree of change in attraction for every participant. The study adds resolution to an undermeasured interval while leaving the general-population rate and long-term persistence of its transitions unsettled Ranker et al. (2025a).
Repeated reports at this life stage often occur alongside changing education, social networks, family independence, and partnerships. Those circumstances offer potential correlates for studies of attraction, self-identification, and disclosure, but the direction of influence is rarely recoverable from two observations. The timing of a label change may follow the introduction of a new term or an interpersonal conversation, precede a partnership, or coincide with a more sustained alteration in reported attraction. A valid design collects those events separately and avoids inferring temporal order from a single cross-sectional life-story question. The scientific goal is to determine when short-term change represents transient measurement variation, developing self-understanding, or a persistent reorganization of experience Ott et al. (2011) Rosario et al. (2008) Ranker et al. (2025a).
Two-wave rates are also sensitive to statistical noise. A respondent who interprets adjacent categories as nearly synonymous may switch between them with little underlying change, whereas a person who repeatedly reports a more widely spaced attraction profile may retain the same identity term. Analysts should distinguish semantic-neighbor changes, changes involving uncertainty, and wider category transitions; they should assess whether repeated wording and available categories were identical. Multiwave designs help establish persistence and returns, and cognitive interviewing can identify what categories mean to participants. These safeguards allow short-interval results to enrich life-course theory instead of being treated as evidence that all orientation dimensions are volatile Ranker et al. (2025a) Ranker et al. (2025b) Hansen et al. (2024).
The Ten-Year Longitudinal Evidence on Women's Sexuality
Diamond's original longitudinal investigation followed 79 women who identified as lesbian, bisexual, or unlabeled across five waves spanning ten years. About two thirds changed their original identity labels at least once, and about one third changed labels at least twice. The study found support for both a bisexual-orientation account and individual fluidity, while its data did not support a simple hypothesis that bisexuality is merely a temporary transitional category. Most importantly, attraction trajectories and identity labels were examined together: broad distributions of same-sex and other-sex attraction showed continuity even where wave-to-wave attraction reports varied. The sample was selected for its theoretical relevance and was not a probability sample of all women Diamond (2008).
The design was unusually informative because it separated repeat identity descriptions from patterns of attraction and observed people over a substantial developmental interval. It also exposed the limits of a binary stable-or-changed classification. A participant may identify as unlabeled after previously selecting bisexual while reporting attractions to more than one gender at both assessments. That move describes identity language and the participant's relation to available categories; it does not by itself establish the appearance or disappearance of bisexual attractions. Meanwhile, a respondent who never changes a category may experience meaningful shifts in relative attraction intensity. Observational inference requires both facts to be retained in the same account Diamond (2008) Diamond (2012).
An enduring methodological lesson is the distinction between generalizable prevalence and evidence for a psychological possibility. A selected cohort can demonstrate that repeated nonmonotonic identity trajectories occur and can test specified competing explanations within its sample. It cannot estimate how common the same trajectories are in an entire national population or establish that women as a group are more fluid than men. Subsequent larger panels provide those comparative baselines, although often using cruder questions. The historical value of intensive follow-up therefore lies in its multidimensional depth and theory testing, while population panels contribute sampling breadth and more stable denominators Diamond (2008) Lilly et al. (2024) Hansen et al. (2024).
Population-Based Evidence in the United States
National panels reveal that identity continuity is widespread even when mobility is scientifically consequential. Mock and Eibach analyzed 2,560 participants in the U.S. MIDUS longitudinal survey with approximately ten years between measurements. More than 97% initially identified as heterosexual, and somewhat more than 2% reported a different orientation identity at the second wave. The authors' subgroup analyses demonstrated important variation by original category and gender. Because the baseline population was overwhelmingly heterosexual and sexual-minority cells were small, the overall percentage is strongly driven by the dominant baseline category; it cannot serve as the change probability for every identity group Mock & Eibach (2012).
Hansen and colleagues examined nationally representative U.S. panel data collected between 2014 and 2022, including 19,469 adults and 970 adolescents with as many as eight sexual-orientation and gender-identity observations. Their work makes repeat reporting a methodological problem in national statistics as well as a developmental subject. It shows how demographic measurement can register movement between categories over survey waves, and how interpretation must account for ages, recruitment, and the fact that categorical changes in identity are not measurements of attraction. In a context where questionnaires serve public health and demographic purposes, classification stability affects denominators and potentially the interpretation of trends in sexual-minority population estimates Hansen et al. (2024).
The Add Health analysis of Savin-Williams, Joyner, and Rieger provides a different view of young adult respondents. Among 12,287 participants observed across two waves about six years apart, stability was more common than change, with distinctive patterns in the mostly heterosexual and bisexual categories. The mostly heterosexual category was larger than all other nonheterosexual identities combined in the population studied, emphasizing that option design shapes the observed distribution. A defensible comparison across MIDUS, Add Health, and modern multiyear panels therefore requires matched definitions, explicit population composition, and intervals. None of these studies alone identifies a universal life-course probability of orientation change Savin-Williams et al. (2012) Mock & Eibach (2012) Hansen et al. (2024).
Cross-National Longitudinal Panels
Hu and Denier analyzed 22,673 respondents to the United Kingdom Household Longitudinal Study, each observed on two occasions across survey windows in 2011–2013 and 2017–2019. They reported that 6.6% changed the sexual orientation identity they selected between observations. Their analyses suggested elevated mobility at younger and older ends of the measured age distribution relative to middle ages. This pattern is scientifically notable because the assumption that all identity mobility belongs to adolescence is too narrow. It remains an association from two surveys separated by several years, so the investigators cannot observe intermediate changes or establish whether age, cohort, or period processes produced the reported profile Hu & Denier (2023).
Lilly and colleagues examined repeated sexual identity reports in a New Zealand national panel that included 45,856 adult respondents aged 18–99. Across seven years, 5.7% changed identity at least once. Reported change was bidirectional, with particularly high mobility among those initially selecting plurisexual categories. Despite women's higher representation in plurisexual identities, the study did not find that women were more fluid than men in identity reporting. These results challenge a generalized sex-difference claim while leaving room for gender-specific contextual effects and differences in other dimensions of attraction or behavior. The distinction matters because a claim about label mobility cannot settle a claim about situational erotic responsiveness Lilly et al. (2024) Diamond et al. (2020).
The United Kingdom and New Zealand analyses converge in showing that adult identity mobility exists, is patterned by initial category, and remains a minority of reports when expressed as an overall panel percentage. Their differences also matter. Cultural terminology, demographic composition, panel recruitment, survey modes, number of observations, missingness, and historical timing vary across national settings. Country is not itself a psychological mechanism; it summarizes many institutions and communicative settings that could mediate identification. A mature comparative literature would harmonize questions and conduct measurement-equivalence analyses before treating cross-national percentage gaps as substantive differences in fluidity Hu & Denier (2023) Lilly et al. (2024) Hansen et al. (2024).
Midlife, Later Adulthood, and Relationship Histories
Midlife observations complicate developmental narratives that place all meaningful transitions early. In the MIDUS cohort, participants were on average in their late forties at baseline and a minority reported different identity categories across a decade, showing that identity continuity and change remain observable beyond the early adult years. The study design cannot determine whether transitions followed new attractions, reframing of longstanding feelings, changing partnerships, or greater willingness to report a category to survey researchers. Its heterosexual-heavy sample further limits precision for the trajectories of bisexual, gay, lesbian, and other minority categories Mock & Eibach (2012).
Later-life sexuality is shaped by accumulated relationships and the opportunity to reinterpret past experiences. Social vocabulary and norms can evolve over a person's lifetime; categories available in youth may differ from those used in a later interview. In such circumstances the same person may narrate a continuous history using a new term, while another person may describe previously unrecognized desire as a recent development. Both narratives are meaningful data about self-understanding, yet they require corroborating temporal observations before they can be classified as attraction change. Hu and Denier's observed age pattern highlights the need for richer late-life panels with interval-specific questions and inclusive response options Hu & Denier (2023) Klein et al. (1985).
A genuinely life-course research program should sample across adult decades, record relationship formation and dissolution, allow reports of nonpartnered attraction, and follow participants even when their self-descriptions change. It should also distinguish chronological age from historical cohort, since participants aged seventy in a current survey formed their earlier identities under very different public categories from those available to today's adolescents. Partnerships provide informative contexts, but changing a partner's gender is neither necessary nor sufficient to establish changed attraction. Future research on later adulthood must maintain participant autonomy and avoid imposing a single developmental endpoint on a diverse collection of histories Lilly et al. (2024) Hu & Denier (2023) Bailey et al. (2016).
Attraction, Identity, and Behavior as Divergent Trajectories
A recurring error in sexual-orientation research is the assumption that changes in reported identity, attraction, and behavior are interchangeable evidence of one latent event. They are correlated but differently constrained processes. Identity is partly a self-descriptive practice, attraction is a reported pattern of experience, and behavior reflects decisions and available relational circumstances. The three may converge for a person at a particular time without implying that their histories must remain synchronized. Savin-Williams and Ream's component-based longitudinal framework and Diamond's intensive work demonstrate why researchers should preserve these differences in both questionnaires and interpretation Savin-Williams & Ream (2007) Diamond (2008).
Discordance is especially informative when measured repeatedly. A person may report attraction to several genders and have partners of only one gender over a defined interval, yielding a behavior-attraction difference with no required change in underlying attraction. Another may use an inclusive identity label while reporting that most recent attractions are toward one gender. The presence of either pattern warrants investigation of opportunity, preferences, disclosure, cultural meaning, and relationship continuity; it establishes no deception or incoherence. Diamond and colleagues explicitly treat gender-pattern discrepancies between attraction and partnering as one distinct operational form of fluidity. They show why these differences should not automatically be merged into temporal instability of feelings Diamond et al. (2020).
The strongest analytic strategy records parallel series: attraction intensity and relative distribution by gender, identity terms and the meaning assigned to them, recent and lifetime behavior, romantic commitments, and relevant situational variables. Investigators can then determine whether variation is dimension-specific, coincident, lagged, or persistent. This approach also reduces the tendency to infer orientation from a single partner's gender or to interpret a temporary absence of sexual activity as absence of attraction. The aim is to retain the lived and measured complexity of trajectories while making the estimands scientifically transparent Klein et al. (1985) Ranker et al. (2025b) Hansen et al. (2024).
Bisexuality, Plurisexuality, and the Meaning of Category Transitions
Bisexuality and other plurisexual identities are central to the history of fluidity research because they make multidimensional attraction visible in identity language. The scientific distinction is straightforward: the possibility of attraction involving more than one gender does not mean that those attractions must have equal intensity, occur simultaneously, or lead to equal numbers of partners. A pattern of broad attraction may be continuous over years despite changing relationships or differently weighted experiences. Diamond's ten-year cohort directly challenged the assumption that bisexuality is merely a stage on the way to a later exclusive identity, providing evidence of sustained bisexual or unlabeled identifications and patterns of attraction Diamond (2008).
Population research finds that participants starting in certain plurisexual categories change categorical identity labels more frequently than those starting in heterosexual categories. In New Zealand, for example, identity change was particularly common among those beginning in plurisexual categories; in Add Health, the bisexual category displayed less label stability than several other categories. Multiple explanations remain possible: a broader range of attractions, different social meanings of labels, varying availability of terms, and a larger number of adjacent descriptive alternatives can all affect transition counts. A mathematical category transition is not a diagnosis of weak identity, an expectation of future heterosexuality, or evidence that bisexual orientation is temporary Lilly et al. (2024) Savin-Williams et al. (2012).
Identity transitions should be analyzed using full transition matrices. Movement from bisexual to pansexual, queer, or unlabeled may describe a relatively small change in the language of self-presentation, whereas movement between categories spanning different attraction profiles may have a different psychological interpretation. Even this distinction requires participants' own definitions and contemporaneous attraction data. Research that collapses bisexual, pansexual, queer, and questioning respondents into an unspecified residual category can manufacture apparent stability or obscure movement within a group. Scientific care with plurisexual measurement is therefore a condition for accurate descriptions of both continuity and fluidity Ranker et al. (2025a) Ott et al. (2011) Diamond et al. (2020).
Gender, Sexual Differences, and Generalization
Gender is an important moderator to investigate, yet the nature of the dependent variable must be specified. Earlier investigations of women's sexual trajectories were influential in generating hypotheses about contextual responsiveness and attraction variability. However, evidence from a selected female cohort cannot establish that all women are more fluid than all men. Studies comparing identity transitions in probability-oriented adult panels provide a different empirical basis. The New Zealand national analysis found that women did not exhibit higher sexual identity mobility than men, even though reporting of plurisexual identities differed by gender. This demonstrates why sex-comparative claims require matched samples and matched measures of the same phenomenon Diamond (2008) Lilly et al. (2024).
Other studies document differences that depend on initial identity, age, and sample. In MIDUS, gender interacted with initial sexual orientation identity; in youth surveys, identities classified as uncertain or open-ended often had especially high mobility. Gender categories themselves may also be more inclusive in contemporary studies than in earlier binary-coded panels, making direct historical comparisons difficult. A robust model tests interactions among gender, age, attraction distribution, starting label, and sampling design. Aggregate changes can arise from unequal proportions of people in categories with different transition probabilities even when within-category rates are similar Mock & Eibach (2012) Ranker et al. (2025a) Hansen et al. (2024).
Claims about the biological determination of orientation require separate evidence and cannot be inferred from gender differences in category movement. Equally, contextual variation is not proof that an orientation was wholly socially produced. The best scientific practice reports distributions, confidence intervals, measurement constraints, and the size and direction of observed differences before proposing mechanisms. A life-course account should leave room for diverse trajectories within each gender, including strong continuity, experiences of fluidity, and changes in identity language. The target of explanation is measured variability among persons rather than an essentialized property assigned to a whole sex or gender group Bailey et al. (2016) Diamond et al. (2020) Lilly et al. (2024).
Partnership, Intimacy, and Changing Social Context
Romantic and sexual relationships are among the contexts in which people recognize, disclose, and express attraction. A new partnership may change the frequency with which a person encounters possibilities associated with a particular gender, while ending a long relationship may create opportunities for experiences previously absent from a life history. Such events may alter reported behavior without changing an enduring attraction profile; they may also accompany new self-descriptions or reported attractions. The inference depends on repeated observations before and after the relevant relationship events, not on the visible gender of one partner Diamond et al. (2020) Hu & Denier (2023).
Relationship status can influence the feasibility of behavior and the public identity others attribute to a person. Bisexual adults in a different-gender partnership may be externally classified as heterosexual; a person in a same-gender partnership may be externally classified as gay or lesbian regardless of their own description. Survey-based identity measures must therefore seek self-report and avoid inferring identity from partnership records. Questions about relationship histories also need time anchors so that lifetime patterns are not confused with current attractions. A longitudinal model can include relationship formation, dissolution, duration, and exclusivity as contextual covariates while allowing attraction and identity to vary independently Klein et al. (1985) Savin-Williams & Ream (2007) Diamond (2008).
Social and cultural conditions enter through vocabulary, visibility, legal recognition, privacy, and norms concerning sexuality. Changing those conditions can make certain self-descriptions easier or harder to report without requiring a change in private desire. They can also shape the relationships people enter and the experiences available to them. These mechanisms are plausible and testable, but observational correlations between an identity change and a shift in social environment do not uniquely identify causation. Researchers can strengthen inference through event-centered repeated sampling, within-person analysis, and direct measurement of perceived safety and identity meaning Hansen et al. (2024) Rosario et al. (2008) Hu & Denier (2023).
Stigma, Identity Disclosure, and Context-Sensitive Reporting
A respondent's answer about sexual orientation is a report produced in a specific setting. The wording of a question, whether other people can see the response, institutional trust, and anticipated stigma may all affect willingness to select a label. This becomes especially important when surveys are administered by different modes or when participants change living circumstances. An apparent identity shift can reflect a genuine reinterpretation, a newly disclosed identity, or a change in reporting conditions. Researchers who wish to study attraction development should measure those possibilities rather than default to one explanation. The large U.S. repeated demographic study explicitly examined variation by survey mode as part of the measurement problem Hansen et al. (2024).
Sexual-minority youth may experience multiple developmental tasks simultaneously, including deciding whether and when to identify publicly with a particular community. Rosario's work illustrates why identity development is studied as a sequence of recognitions, affiliations, and self-descriptions rather than as one initial classification. A report of uncertainty can be a meaningful and durable self-description, not merely missing information. When categories omit terms participants use, repeated selection of another option can obscure an otherwise coherent trajectory. Write-in responses and open-ended interviews can reveal such issues, but they require a transparent coding plan that respects the respondent's intended meaning Rosario et al. (2006) Rosario et al. (2008) Ranker et al. (2025a).
Studies of reporting context must avoid equating social pressure with an individual's attraction mechanism. The empirical pathway from context to reported answer may operate through willingness to disclose; a different pathway from context to experience would require repeated attraction data demonstrating within-person change. A third pathway may involve altered interpretation of the survey term itself. These mechanisms could coexist and require independent indicators. Privacy-protective designs, confidential interviews, cognitive testing, and sensitive interpretation improve both participant welfare and scientific validity, particularly when a recorded transition could carry unwanted implications for a respondent's identity Bailey et al. (2016) Hansen et al. (2024) Diamond et al. (2020).
Autobiographical Memory and Retrospective Narratives
Retrospective accounts are central to how people make sense of their sexuality, but memory-based reports do not reproduce a time series of historical observations. Ranker and colleagues directly compared changes detected from prospective surveys with participants' retrospective recollection of change across the same four-month interval. Among 1,235 U.S. youths ages 14–25, prospective comparison detected a change in sexual orientation identity reports for 12.8%, while 5.0% retrospectively recalled an identity change; attraction change was detected for 21.9% prospectively and recalled by 12.4%. These differences demonstrate a measurement effect, not a simple proof that one method reveals the person's ultimate orientation Ranker et al. (2025b).
Why recall and repeated reports disagree remains a substantive question. Respondents may consider minor shifts too unimportant to remember, may integrate adjacent labels into one continuous self-understanding, or may reconstruct past experiences using current vocabulary. Conversely, repeated surveys can register response noise that respondents appropriately do not regard as meaningful life changes. A rigorous assessment compares both forms with carefully defined outcomes and ideally supplements them with interviews about perceived significance and contemporaneous records of attraction. The key insight is that narrative coherence and repeated measurement provide different kinds of evidence about personal continuity Ranker et al. (2025b) Diamond (2008).
Memories of identity change can have psychological significance independently of their agreement with survey-coded transitions. Someone may tell a life story emphasizing continuity while having used several adjacent terms; another may emphasize a formative discovery while having reported similar attraction patterns earlier. Researchers should analyze narrative identity as a process of autobiographical meaning-making and avoid ranking participants' descriptions by the magnitude of survey category differences. This separation also guards against imposing retrospective narratives on people whose categories have changed for cultural or linguistic reasons. Longitudinal designs and qualitative narratives are strongest when their differences are treated as evidence for further inquiry rather than as a reason to discount either source wholesale Ranker et al. (2025b) Klein et al. (1985) Hu & Denier (2023).
Research Designs for Sexual Fluidity Across Time
Sexual fluidity research is a longitudinal measurement problem before it is a causal explanation. Cross-sectional surveys estimate the distribution of sexual identity or attraction categories at one time, whereas panels permit an estimate of within-person transition conditional on continued participation. Diary designs can capture fluctuations in attraction over days, intensive interviews can clarify meaning, and repeated national surveys can estimate population-level identity mobility. Each design defines a different target quantity. The researcher must specify the population, observation interval, outcome variable, number of waves, and sampling framework before interpreting how frequently change occurs. Studies differing on these terms answer different questions even when they use the same phrase sexual fluidity Diamond et al. (2020) Lilly et al. (2024) Ott et al. (2011).
Two-wave designs estimate whether endpoints differ. They cannot establish how many unobserved transitions occurred between endpoints, how long a state persisted, or whether a respondent returned to an earlier category. Three or more waves permit patterns such as stable identification, one transition, return, and recurrent change to be distinguished, subject to the resolution of the measurement schedule. More waves increase observational detail but also create more opportunities for category movement caused by random response variation. Analysts must therefore report the observation schedule and consider both the probability of any observed transition and the frequency of changes relative to the number of available wave-to-wave comparisons Diamond (2008) Hansen et al. (2024) Ranker et al. (2025a).
Sampling frames also determine the interpretation of the evidence. Probability-based national panels are useful for population summaries, but rare initial identity categories may yield few participants and wide uncertainty intervals. Deliberately recruited sexual-minority samples provide greater depth for analyzing identity trajectories but cannot directly estimate national prevalences. Online youth cohorts can measure short intervals while reflecting the characteristics of those able and willing to participate. Triangulation across these designs is appropriate when differences in target populations are made explicit; pooling their percentages without harmonization would disguise selection and measurement differences. Scientific synthesis should preserve the distinct strengths of intensive longitudinal, community, and nationally representative studies Diamond (2008) Mock & Eibach (2012) Ranker et al. (2025a).
Construct Validity, Survey Categories, and Measurement Invariance
Construct validity begins with the question of what a survey item means. A single item asking respondents to identify as heterosexual, bisexual, gay, lesbian, or other captures a categorical report under the offered choices. It does not measure current attraction magnitude, duration of feelings, romantic preference, or exclusive behavior. Its content also changes when response options change. In longitudinal studies, a revised list of categories can produce more observed label movement even if the underlying experiences of some respondents remain continuous. Repeated research should retain stable core items, document revisions, and offer carefully designed ways to communicate self-descriptions outside a fixed list Hansen et al. (2024) Ott et al. (2011) Savin-Williams et al. (2012).
Measurement invariance asks whether an instrument has comparable meaning across times, ages, cultures, and participant groups. Identity-label categories may be especially difficult to render equivalent across historical periods, because words such as queer, pansexual, or mostly heterosexual have changing public circulation. For attraction scales, comparable framing, time windows, and gender referents are necessary to evaluate continuity. Researchers can use cognitive interviews to establish how participants interpret items, include open-ended explanations for apparent shifts, and perform sensitivity analyses that distinguish strict category changes from transitions between semantically neighboring terms. Invariance failures should be treated as substantive findings about measurement rather than as mere inconvenience Klein et al. (1985) Hansen et al. (2024) Ranker et al. (2025b).
Reliable measurement also requires attention to response variability that does not reflect meaningful change. A person whose attraction profile lies near an arbitrary categorization threshold may cross it between surveys with small underlying differences. Coarsely measured variables can therefore generate abrupt label changes from continuous experiences. Investigators should preserve graded attraction ratings alongside categorical identity and report transition matrices using the original categories before any aggregation. Where possible, repeat items within brief intervals, assess test-retest reliability, and distinguish persistent trajectories from isolated observations. These practices increase clarity about whether reported variability belongs to the psychological construct, the reporting system, or both Diamond et al. (2020) Savin-Williams & Ream (2007) Ranker et al. (2025a).
Statistical Models of Continuity and Change
Transition matrices provide the most transparent foundation for analysis of categorical identity. A matrix lists the number and proportion of people in each initial category who occupy every available category at follow-up. Row denominators reveal conditional mobility: a study with predominantly heterosexual respondents can have a low overall transition proportion while showing substantial mobility within smaller baseline groups. The inverse problem also matters: a group represented by very few people can display an unstable estimate with considerable sampling uncertainty. Matrices should therefore include cell counts, confidence intervals where appropriate, missing-data conventions, and clear distinctions between changes in population prevalence and changes within individuals Savin-Williams et al. (2012) Lilly et al. (2024) Hu & Denier (2023).
With multiple waves, investigators can compare first-to-last change with any-wave movement and model the number, direction, and persistence of transitions. Markov models may estimate conditional probabilities of occupying a category in the next wave, provided the assumptions about time intervals and dependence are reasonable. Latent transition models can represent measurement error and movement among inferred states, but their interpretation is only as sound as the mapping from observed items to states. Growth-curve models or nonlinear mixed-effects models are more appropriate for continuous attraction ratings when the substantive question concerns magnitude or trajectory shape. Models should be selected from the defined construct and available data rather than from a preference for sophisticated terminology Ott et al. (2011) Diamond (2012) Hansen et al. (2024).
Subgroup analyses should account for unequal baseline distributions and attrition. Differences in overall rates across women and men, age bands, or countries may disappear or reverse after conditioning on starting category. Multilevel models can estimate within-person associations with contextual variables, but temporal correlations are not evidence that an event caused a later change. Researchers should preregister primary outcomes, report both absolute and conditional transition probabilities, provide appropriate uncertainty intervals, and distinguish exploratory interaction tests from hypotheses specified in advance. Transparent reporting is particularly important when small identity groups lead to unstable estimates or when repeated testing increases the chance of misleading subgroup claims Mock & Eibach (2012) Lilly et al. (2024) Ranker et al. (2025a).
Attrition, Selection, and Generalizability
Longitudinal participation is not random by default. Participants may discontinue when they relocate, experience relationship changes, face social stigma, or lose interest in surveys. If attrition is associated with both starting identity and likelihood of reporting change, complete-case estimates may be biased. Investigators should compare baseline characteristics of retained and lost participants, use plausible missing-data models, perform sensitivity analyses for departures from missing-at-random assumptions, and distinguish planned missingness from unplanned nonresponse. Neither a very large initial sample nor a long observation period eliminates the need for these checks Hansen et al. (2024) Hu & Denier (2023) Lilly et al. (2024).
Survey-mode differences create additional selection and reporting problems. The willingness to provide a particular identity answer may depend on whether the interview is conducted privately online, on paper, by telephone, or in another person's presence. Recontact can also select people who are more comfortable with repeated identity questions. These sources of variation matter in historical data assembled from multiple survey instruments. Researchers should report collection modes, participant privacy conditions, changes in item wording, and the extent to which demographic refresh samples are comparable with the original panel. Results can then be interpreted in relation to the actual sampling and measurement process Hansen et al. (2024) Ranker et al. (2025b).
Generalization should be explicitly bounded. The ten-year study of 79 women has substantial value for theory and rich longitudinal trajectories, but its initial selection does not authorize population rates. Nationally oriented panels offer broader estimates but often use only one identity question, limiting inference about attractions. Young-adult online cohorts address high-frequency transitions in participants who choose to enroll, a sample that may differ from youth who remain outside the study. A strong scientific conclusion integrates these complementary evidentiary roles while preserving uncertainty about other populations, jurisdictions, age ranges, and measurement periods Diamond (2008) Ranker et al. (2025a) Lilly et al. (2024).
Causal Interpretation and Alternative Explanations
A correlation between a contextual event and a subsequent identity report supports a hypothesis about timing, not an established cause of sexual orientation change. Relationship dissolution may coincide with a label transition because opportunities change; a label transition may precede a new relationship; both may follow a third factor such as social network change or growing familiarity with terminology. Observational panels usually lack random assignment of such experiences and may not measure the relevant confounders. Causal interpretation requires a clearly specified intervention or exposure, a temporally ordered outcome, plausible exchangeability assumptions, and sensitivity analysis. The object of causal inference must also be precise: effects on self-labeling differ from effects on private attraction Hu & Denier (2023) Hansen et al. (2024) Bailey et al. (2016).
Measurement-based alternatives include changes in survey choices, shifts in self-disclosure, recall differences, and classification around cut points. Actual attraction variability remains another possibility, and these pathways are not mutually exclusive. A participant may both learn a new identity term and experience changes in attraction; research must assess each contribution independently. Combining short-interval repeated reports with context measures and contemporaneous qualitative interviews is one promising approach. Long-term within-person observation would make it possible to test whether a transition persists across waves and whether it reliably precedes or follows specified life events, but such evidence remains uneven across the field Ranker et al. (2025b) Diamond et al. (2020) Diamond (2012).
Biological explanations of the origins of orientation and psychological accounts of its longitudinal expression address partly different questions. Evidence that a trait has developmental or biological correlates does not imply that all reported dimensions must be temporally invariant, and evidence of label fluidity does not refute enduring biological contributions. The proper inference depends on the measured phenotype and the research design. This monograph therefore treats proposed explanatory pathways as hypotheses wherever studies report associations without direct experimental identification. Scientific restraint here is substantive: it prevents authors from turning a repeated survey category into a claim about unobserved causal structures Bailey et al. (2016) Savin-Williams & Ream (2007) Lilly et al. (2024).
Evidence Quality, Independent Replication, and Meta-Interpretation
The evidence base has several convergent strengths. Independent longitudinal investigations in the United States, United Kingdom, and New Zealand observe some changes in reported identity over time, while repeatedly documenting considerable continuity at the population level. Intensive studies also demonstrate that a person's attraction, identity, and behavior can vary on different time scales. These results have appeared in diverse cohorts and under different instruments, so their broad qualitative convergence is meaningful. Yet direct numerical pooling would require harmonized response categories, wave spacing, baseline distributions, and missing-data treatments that are generally absent from the available set Mock & Eibach (2012) Hu & Denier (2023) Lilly et al. (2024) Diamond et al. (2020).
Claims about the size of an effect must remain attached to a defined sample and design. The New Zealand estimate of 5.7% over seven years, the United Kingdom estimate of 6.6% across two measurement periods, and the 11.4% two-month youth-cohort result are not contradictory percentages competing to describe one universal parameter. They concern different populations and observation schedules; their identity options and sampling conditions differ. Such heterogeneity should be described before invoking psychological mechanisms to explain the numerical spread. Evidence quality also depends on adequate minority-subgroup counts and uncertainty reporting, not simply the total sample size Hu & Denier (2023) Lilly et al. (2024) Ranker et al. (2025a).
Replication should encompass operational details as well as broad conclusions. An exact replication might repeat response items and time intervals in a comparable panel; a conceptual replication could measure attraction and identity separately using different instruments. Stronger conclusions emerge when studies show that a predicted pattern survives both approaches. Current gaps include longitudinal coverage of later adulthood, frequent measures in probability-based samples, extensive gender-diverse and culturally diverse cohorts, and mechanisms with temporally resolved contextual data. Those are research priorities rather than claims that an entire subject lacks evidence: the longitudinal existence of identity mobility is well documented, while causes, detailed trajectories, and cross-population transportability remain open Hansen et al. (2024) Ranker et al. (2025b) Diamond et al. (2020).
Psychological Well-Being, Social Recognition, and Identity Trajectories
Changes in sexual self-description can affect social relationships and personal well-being, but the change itself does not constitute a clinical disorder. The likely consequences depend on context: whether a person has access to language that fits their experience, whether disclosure is safe, whether friends and family respond supportively, and whether institutions treat self-reports respectfully. Researchers should distinguish the psychological experience of revising a label from external reactions to that revision. A person may experience relief through a newly fitting self-description, distress through stigma, uncertainty during exploration, or several of these experiences at different times. These are empirical hypotheses that require independent measures of well-being and social context Rosario et al. (2006) Hansen et al. (2024).
Associations between reported change and health outcomes require particular care. A youth who changes a category may also experience discrimination, social rejection, relationship stress, or different access to affirmative care; the observed association may reflect those overlapping exposures rather than an intrinsic effect of fluidity. In longitudinal observational work, baseline distress can influence subsequent self-understanding or reporting, and contemporaneous circumstances can influence both. Studies should therefore assess prior well-being, relevant stressors, perceived support, changes in living situation, and identity-specific stigma before offering causal claims. Outcomes should be analyzed across participants who report both continuous and changing identities, without treating either trajectory as the preferred endpoint Rosario et al. (2008) Ott et al. (2011) Ranker et al. (2025a).
Research instruments can also influence welfare when participants perceive that a survey requires them to defend inconsistent answers. A respectful procedure allows individuals to change labels, decline a category, use multiple terms, or report uncertainty without suggesting that their responses invalidate earlier reports. For clinical research, the emphasis is on understanding the person's current experience and needs; no category-transition statistic provides a basis for prescribing an identity or expecting a future transition. This stance improves measurement validity because it reduces pressure to choose an externally desired answer and preserves the distinction between scientific observation and normative judgment Diamond (2008) Hansen et al. (2024) Bailey et al. (2016).
Research Ethics, Privacy, and Respectful Longitudinal Practice
Longitudinal identity data are sensitive because they link intimate self-reports across time and may reveal a trajectory that participants have not disclosed to people around them. Confidentiality is therefore integral to methodological quality. Investigators should use data-minimizing procedures, robust controls on longitudinal identifiers, carefully justified retention periods, and release policies that prevent indirect identification in small groups. They should explain which kinds of future contact and data linkage are planned and permit participants to answer sensitive questions without unnecessary disclosure. These are requirements of responsible research design, particularly when rare identities or geographic subgroups create disclosure risks Hansen et al. (2024) Ranker et al. (2025a).
Recruitment and survey language should permit meaningful self-description across ages and cultural contexts. A participant should be able to distinguish no attraction from an absence of recent partners and should have access to categories suitable for their self-understanding. Open response options can increase inclusion, although coding procedures then require transparent documentation and sufficient protection of identifiable narratives. Researchers should not use participants' historical answers to assert that a current self-description is mistaken, or interpret newly disclosed minority identity as proof of a chosen developmental trajectory. Trust is relevant to validity: participants who believe that answers will be judged or exposed may alter what they report Klein et al. (1985) Ranker et al. (2025b).
Scientific publication should protect the possibility of multiple life-course outcomes. Study design can describe changing and continuous trajectories without turning either into a prescriptive ideal. The language used to report subgroup differences matters because an association derived from one survey can easily be exaggerated into a claim about the essential nature of a gender or identity. Editors should retain study limitations beside numerical results, distinguish social labels from attraction, and avoid attributing unobserved motives or causal histories to respondents. These practices produce a more accurate record and reduce the risk that research findings are misused to deny a person's account of their own sexuality Bailey et al. (2016) Lilly et al. (2024) Diamond et al. (2020).
Digital Environments and the Artificial Era
Digital environments increasingly supply the language, communities, and tools through which people describe sexuality. Online communities can expose participants to identity terms that were unfamiliar in earlier periods, while digital surveys and private platforms change how people report sensitive experiences. Such changes may alter the public availability and personal meaning of labels. They do not establish that digital media cause a particular attraction trajectory. A research program concerned with the Artificial Era can study these environments as measurable conditions of identity expression, social interaction, and autobiographical continuity, while maintaining the human psychological phenomenon as its primary subject Hansen et al. (2024) Ranker et al. (2025b).
Human–artificial conversational systems introduce further questions about reflective self-description. A person may use an artificial interlocutor to discuss an emerging identity, recount relationships, or rehearse language for disclosure. Empirical research could examine whether these interactions influence confidence in self-labels, perceived safety, or the timing of disclosure, using consented longitudinal protocols with comparisons to other settings. Such hypotheses must be kept distinct from established findings on sexual fluidity. The existence of an artificial conversational partner provides a novel relational and communicative context; it does not supply evidence that sexual orientation changes as a result of that context Diamond et al. (2020) Rosario et al. (2008).
Future studies can distinguish digital vocabulary exposure, changes in social affiliation, and changes in attraction by measuring them separately over time. They can also investigate whether platform recommendation systems systematically foreground certain categories, whether privacy conditions affect the accuracy of reports, and how digital records influence retrospective self-narratives. These are empirically testable questions within Psychology of Human–Artificial Relations and Postsubjective Psychology when appropriately operationalized. They extend the study of life-course identity into contemporary information environments without importing unsupported conclusions about the psychological experience or personhood of artificial systems Ranker et al. (2025b) Hansen et al. (2024).
Integrative Scientific Synthesis
The research evidence supports an organized set of conclusions. First, sexual orientation is measured through several related dimensions whose temporal properties differ: attraction, identity, behavior, and relationship history do not always move in parallel. Second, identity change is demonstrable across adolescence and adulthood, including in nationally oriented samples, while stability remains common in many cohorts. Third, the apparent frequency of change depends strongly on age composition, starting labels, survey options, interval length, and observational design. Fourth, selected intensive cohorts and large panels contribute different strengths and should be synthesized on their respective terms. Together these findings establish a rigorous basis for studying variation without making a universal claim about an individual's future experience Diamond (2008) Lilly et al. (2024) Hansen et al. (2024).
The most important conceptual synthesis concerns scale and mechanism. Momentary or situational attraction responsiveness, recurrent changes in how someone labels their sexuality, and changes in partner gender are distinguishable time series. A person may show continuity in one while reporting variation in another. A broad pattern of attraction involving more than one gender can be persistent even when relative intensities change. Conversely, identity stability does not prove perfect invariance of attraction. Research should specify the construct and time scale before calculating a transition, use multidimensional measures, and model social reporting conditions alongside psychological processes Diamond et al. (2020) Savin-Williams & Ream (2007) Diamond (2012).
The evidence also sets boundaries on explanation. Repeated observations establish reported continuity and movement, not their entire causal history. Findings in youth, selected sexual-minority cohorts, and adult national panels may differ because the populations, temporal windows, and categories differ; discrepancies do not automatically reveal incompatible psychological mechanisms. Generalization requires an account of recruitment, retention, baseline categories, and meaning equivalence across populations. Theory is most productive when it generates specific, separately measurable predictions regarding attraction, identity, relationships, and context rather than treating every change as an instance of a single unexplained process Ranker et al. (2025b) Hu & Denier (2023) Mock & Eibach (2012).
Open Questions and a Life-Course Research Agenda
Research priorities begin with measurement harmonization. A common longitudinal core should include confidential identity self-description, graded sexual and romantic attraction, recent and lifetime partnering, contextual disclosure variables, and the participant's interpretation of important labels. Separate items should identify changes perceived as personally meaningful and changes detected by the measurement schedule. With explicit periods of reference and stable wording, studies can distinguish brief attraction fluctuations from persistent profile changes and track identity revisions that leave the underlying attractions approximately continuous. Shared methods would permit stronger comparison across countries and historical periods without forcing participants into categories that fail to describe their experiences Diamond et al. (2020) Hansen et al. (2024) Ranker et al. (2025b).
A second priority is coverage of the whole life course. Existing strong studies include adolescent cohorts, emerging-adult samples, and multiyear adult panels, but dense longitudinal measurements in middle and later adulthood are relatively limited. Multiple cohorts followed across changing social conditions could separate age, period, and cohort hypotheses more convincingly than two-wave comparisons. Oversampling smaller identity groups with appropriate population weights would improve the precision of transition matrices, while modern missing-data sensitivity analyses would make claims about generalizability more credible. Data collection should preserve the ability to examine experiences of gender-diverse participants and culturally varied populations without assuming that identical categories carry identical meanings everywhere Hu & Denier (2023) Lilly et al. (2024) Ranker et al. (2025a).
A third priority is stronger tests of explanatory mechanisms. Repeated assessments around relationship transitions, changes in social setting, disclosure milestones, and shifts in vocabulary could identify the temporal ordering of contextual and experiential changes. Intensive diary sampling joined to longer panels would allow researchers to test whether short-term variability predicts later identity revision or whether the two are largely independent. Studies should preregister key distinctions and publish transition distributions as well as aggregate rates. The unresolved problem is not whether some people report change, which is well established, but which processes produce which kinds of change and under what conditions they become persistent or personally significant Diamond (2012) Diamond et al. (2020) Ranker et al. (2025b).
Conclusion
Sexual fluidity and orientation across the life course constitute a coherent research field when its object is defined as the longitudinal organization of sexual attraction, identity, behavior, and relational histories. The evidence documents both substantial continuity and meaningful change across defined samples, ages, and time intervals. A ten-year intensive study provides detailed examples of multidimensional trajectories, while large U.S., British, and New Zealand panels establish that categorical identity mobility can be observed beyond adolescence. Short-interval studies reveal additional transitions that long-interval panels may miss, and prospective-retrospective comparisons demonstrate that memory and measurement schedules influence the reported rates Diamond (2008) Hu & Denier (2023) Lilly et al. (2024) Ranker et al. (2025a) Ranker et al. (2025b).
The most defensible scientific conclusion preserves the differences between changing a label, changing attractions, entering a relationship, and interpreting one's history. Those phenomena can interact, but none can be substituted for another without evidence. Findings about women, men, bisexual and plurisexual persons, and different age groups require conditional analysis rather than universal developmental stereotypes. The methodological foundation is repeated, privacy-protective measurement with explicit constructs, comparable categories, careful sampling, transparent uncertainty, and longitudinal designs suited to the duration of the process being studied Diamond et al. (2020) Hansen et al. (2024) Savin-Williams et al. (2012).
The resulting understanding is one of differentiated psychological trajectories. Individual histories may include sustained attraction profiles, fluctuating experiences, changes in self-description, continuous identities expressed under changing circumstances, and multiple combinations of these possibilities. A scientifically responsible monograph recognizes the observed diversity of paths, assigns causal explanations only to evidence capable of supporting them, and frames remaining questions as testable research problems. Sexual fluidity is therefore best investigated as a set of precisely described longitudinal phenomena embedded in developmental and social time. This framework supports future comparative science without prescribing how any person should identify or how their sexuality should develop Bailey et al. (2016) Diamond (2012) Lilly et al. (2024).
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
- Diamond, L. M. (2008). Female bisexuality from adolescence to adulthood: Results from a 10-year longitudinal study. Developmental Psychology, 44(1), 5–14. https://doi.org/10.1037/0012-1649.44.1.5
- Diamond, L. M. (2012). The desire disorder in research on sexual orientation in women: Contributions of dynamical systems theory. Archives of Sexual Behavior, 41(1), 73–83. https://doi.org/10.1007/s10508-012-9909-7
- Diamond, L. M., Alley, J., Dickenson, J., & Blair, K. L. (2020). Who counts as sexually fluid? Comparing four different types of sexual fluidity in women. Archives of Sexual Behavior, 49(7), 2389–2403. https://doi.org/10.1007/s10508-019-01565-1
- Hansen, C., Viox, M. H., Fordyce, E., Johns, M. M., Avripas, S., & Michaels, S. (2024). The longitudinal measurement of sexual orientation and gender identity: A study of identity change in a nationally representative sample of U.S. adults and adolescents. LGBT Health, 11(7), 522–530. https://doi.org/10.1089/lgbt.2023.0260
- Hu, Y., & Denier, N. (2023). Sexual orientation identity mobility in the United Kingdom: A research note. Demography, 60(3), 659–673. https://doi.org/10.1215/00703370-10769825
- Katz-Wise, S. L., & Hyde, J. S. (2015). Sexual fluidity and related attitudes and beliefs among young adults with a same-gender orientation. Archives of Sexual Behavior, 44(5), 1459–1470. https://doi.org/10.1007/s10508-014-0420-1
- Klein, F., Sepekoff, B., & Wolf, T. J. (1985). Sexual orientation: A multi-variable dynamic process. Journal of Homosexuality, 11(1–2), 35–49. https://doi.org/10.1300/J082v11n01_04
- Lilly, K. J., Satherley, N., Sibley, C. G., Barlow, F. K., & Greaves, L. M. (2024). Fixed or fluid? Sexual identity fluidity in a large national panel study of New Zealand adults. The Journal of Sex Research, 61(9), 1351–1366. https://doi.org/10.1080/00224499.2023.2289517
- Mock, S. E., & Eibach, R. P. (2012). Stability and change in sexual orientation identity over a 10-year period in adulthood. Archives of Sexual Behavior, 41(3), 641–648. https://doi.org/10.1007/s10508-011-9761-1
- Ott, M. Q., Corliss, H. L., Wypij, D., Rosario, M., & Austin, S. B. (2011). Stability and change in self-reported sexual orientation identity in young people: Application of mobility metrics. Archives of Sexual Behavior, 40(3), 519–532. https://doi.org/10.1007/s10508-010-9691-3
- Ranker, L. R., Korkodilos, R., Conti, J., Nelson, K. M., Xuan, Z., Gordon, A. R., & Katz-Wise, S. L. (2025). One-in-ten adolescents and young adults report changes in their sexual orientation identity over 2-month follow-up: Results from a longitudinal cohort study. Psychology of Sexual Orientation and Gender Diversity. Advance online publication. https://doi.org/10.1037/sgd0000782
- Ranker, L. R., Katz-Wise, S. L., Gordon, A. R., Korkodilos, R., Xuan, Z., & Nelson, K. M. (2025). Sexual orientation identity and attraction changes among youth: Prospective versus retrospective measurement. Annals of Epidemiology, 112, 8–14. https://doi.org/10.1016/j.annepidem.2025.10.006
- Rosario, M., Schrimshaw, E. W., Hunter, J., & Braun, L. (2006). Sexual identity development among gay, lesbian, and bisexual youths: Consistency and change over time. The Journal of Sex Research, 43(1), 46–58. https://doi.org/10.1080/00224490609552298
- Rosario, M., Schrimshaw, E. W., & Hunter, J. (2008). Predicting different patterns of sexual identity development over time among lesbian, gay, and bisexual youths: A cluster analytic approach. American Journal of Community Psychology, 42(3–4), 266–282. https://doi.org/10.1007/s10464-008-9207-7
- Savin-Williams, R. C., & Ream, G. L. (2007). Prevalence and stability of sexual orientation components during adolescence and young adulthood. Archives of Sexual Behavior, 36(3), 385–394. https://doi.org/10.1007/s10508-006-9088-5
- Savin-Williams, R. C., Joyner, K., & Rieger, G. (2012). Prevalence and stability of self-reported sexual orientation identity during young adulthood. Archives of Sexual Behavior, 41(1), 103–110. https://doi.org/10.1007/s10508-012-9913-y