Biological sex, gender identity, gender expression, and sexual orientation are four different things, and they vary on their own. Much of the confusion, and much of the public argument, comes from treating them as one. This piece defines the constructs and reports the evidence; the value and policy questions are real, and separate.
The most valuable move in this whole area is also the simplest: keep four different things apart. They are constantly merged in everyday talk, and most of the resulting confusion clears the moment they are separated.
Biological sex is the physical layer, chromosomes, hormones, gonads, and anatomy. For most people these line up as male or female, so sex is strongly bimodal. It is also made of parts that usually match but occasionally do not, which is what produces the natural variations grouped under intersex, or differences of sex development. Underneath, sex is built by a multi-step genetic cascade rather than thrown by a single switch, and those steps do not always run the same way, so chromosomes alone do not fully dictate the outcome (Arboleda, Sandberg and Vilain, 2014). The clean two-box picture is an idealisation of a process that carries real, biologically grounded variation.
Gender identity is something else entirely: a person's internal sense of being a man, a woman, both, or neither. For most people it matches their birth sex; for a minority it does not. The conceptual split between biological sex and this inner sense was drawn out in the clinical literature decades ago (Stoller, 1968).
Gender expression and roles are different again: how a person presents and behaves, the clothes, voice, manner, and the social expectations a culture attaches to "masculine" and "feminine". A useful early finding here is that masculine and feminine are not two ends of a single line. They are two largely independent dimensions, so a person can be high on both, low on both, or any mix (Bem, 1974). And what actually counts as masculine or feminine, the clothes, jobs, manners, and traits, varies markedly across cultures and across history, which is strong evidence that rigid gender roles are social conventions rather than biological mandates. Expression is also the layer most visible to others and the least reliable guide to the rest.
Sexual orientation is the pattern of who a person is attracted to, and it is worth separating even further: the orientation (the pattern itself) is distinct from the identity (the label someone uses, like "bisexual" or "straight") and from behaviour (what they actually do), which need not all line up. The first large attempt to map it found attraction spread along a continuum rather than sorting into two boxes (Kinsey et al., 1948).
The reason to hold these four apart is that they move independently. A person's expression does not reveal their identity; their identity does not reveal their orientation; none of the three follows from their biological sex. Treat them as one thing and you will misread people and talk past other people constantly.
Two well-supported findings cut against common assumptions. The first is about how different the sexes really are. The popular picture is of two psychologically distinct kinds of people. The data say otherwise: across a large body of meta-analysis, men and women come out more alike than different on most measures, with the typical difference small and the overlap large, though a handful of differences are bigger (Hyde, 2005). Real averages differ on some things; near-total separation is rare.
The second concerns where sexual orientation comes from, an area where the science has quietly settled a few points often still argued in public. Orientation is not a lifestyle choice: attraction typically dawns around age ten, well before any sexual activity, and people do not decide it (Bailey et al., 2016). It has biological roots, with moderate heritability, around a third in the largest analysis, and identifiable correlates. But it is not a single "gay gene": a genome-wide study of nearly half a million people found several genetic markers, each with a tiny effect, together explaining only part of the variation and unable to predict any individual (Ganna et al., 2019). So "born this way" is too simple, and "chosen" is wrong; the honest summary is biologically influenced, not fully determined, and not a decision. There is also real fluidity for some people, reported more often by women, where attractions shift over time without that being a choice either (Diamond, 2016).
Gender identity has a similar story taking shape, though the evidence is younger and thinner. Twin studies point to genuine heritable, biological contributions, with gender identity looking like a complex polygenic trait rather than a pure social construction or a simple choice (Polderman et al., 2018), and brain-imaging work finds that the brains of transgender people tend to show distinctive patterns that sit between the typical male and female ones rather than matching their birth sex (Guillamón, Junqué and Gómez-Gil, 2016). Two cautions keep this honest. The neural findings are suggestive, not settled: the samples are small, many participants were already taking hormones, and no single "brain-sex" switch has been shown to determine identity. And, more important, the dignity of a person's gender identity does not hang on any of this. Grounding respect in a contested biomarker is fragile, because if the finding later weakens, the respect should not, given that how we treat people is a values commitment rather than a verdict awaiting the next brain scan.
Several genuine debates remain, and it helps to see them clearly rather than through the heat that surrounds them.
The cleanest is about a number. You may have seen the claim that roughly 1.7% of people are intersex, and also the claim that it is more like 0.018%. Both are defensible, because they count different things: the larger figure includes any deviation from an idealised male or female body, while the smaller counts only genuinely ambiguous cases (Sax, 2002). The gap is not a factual disagreement but a definitional one, and it is a tidy reminder that a statistic means nothing until you know what it is counting (L0).
A deeper scientific question is how best to model sex and gender at all. A prominent review argues that evidence from neuroscience, hormones, psychology, and child development undercuts the clean two-box model and favours treating sex and gender as multidimensional (Hyde et al., 2019). It is an evidence-informed case, not universally accepted, and how to model this is itself an open scientific argument, not a settled fact in either direction. Worth noting alongside it: many cultures have long recognised more than two gender categories, so the two-category system is not a human universal.
The causes of gender identity, why most people's matches their birth sex and a minority's does not, are not well understood. Biological correlates have been proposed; no settled mechanism exists, and confident claims in either direction outrun the evidence.
Finally, the part of this field that generates the most heat is not really about the definitions above. It is about clinical care, especially for adolescents experiencing gender dysphoria, and there the evidence base is genuinely and sharply contested, with major clinical reviews currently reaching different conclusions (World Professional Association for Transgender Health, 2022; Cass, 2024). That is a clinical and policy question, and an important one, but it is a different question from how the constructs are defined or what the orientation science shows. This piece stays with the latter, and on the former the honest position is that the evidence is unsettled and best read close to the studies rather than through political framing.
Much of this rests on self-report and self-identification, the categories are socially situated and vary across cultures and eras, and most samples are Western. The field is also moving fast and is politically charged in both directions, which puts real pressure on what gets studied and how findings are reported. None of that means there is nothing solid here; it means reading carefully and separating what is established from what is contested.
What shapes gender identity and sexual orientation during development? Is sex and gender better modelled as a binary with exceptions or as genuinely multidimensional? What drives fluidity, and for whom? And, on the clinical side, what does the best evidence actually support for adolescent care, a question still being fought over.
The usable core: keep the four constructs distinct, never read one off another, and separate the questions that evidence can answer from the values and policy questions that it cannot.
Most practical errors here are conflation. A form, a dataset, or a marketing segment that treats "gender" and "sexual orientation" as the same field will collect nonsense, because they are different variables. If you need the information at all (and you should really consider this carefully), ask for the specific construct you actually mean, use a two-step approach for sex and gender (sex recorded at birth, plus current gender identity), offer an orientation list that reflects how people actually describe themselves, and never infer someone's identity or orientation from how they present. The goal is accuracy and respect at once, and the two pull in the same direction: precise categories are both more correct and less alienating than either a forced binary or a box-ticking gesture.
A great deal of public argument in this area is two sides talking past each other, because they are mixing up not just the four constructs but four different kinds of question: what a term means (definitional), what the evidence shows (empirical), what medicine should do (clinical), and what is right or fair (values). Most heat comes from treating a values disagreement as if it were a factual one, or vice versa. The clarifying discipline is to name which question is actually on the table. And a standing caution: the science here defines and measures, it does not settle the value questions, which have to be argued on their own terms rather than smuggled in under a statistic. For official data, the same rule as for companies applies, collect the specific variable you mean, or the numbers will not mean what you think.
Three habits. First, keep the four apart, sex, gender identity, gender expression, sexual orientation, and resist reading one from another, because they genuinely vary on their own. Second, separate the kinds of question: a definitional dispute, an empirical one, a clinical one, and a values one need completely different kinds of argument, and confusing them is most of why these conversations go badly. Third, hold contested things as contested: where the evidence is genuinely unsettled, say so, rather than borrowing certainty from your politics.
.png)