A structured web research synthesis, current to September 2026: a 56-row cross-module evidence shortlist plus two broader scholarly pools — 947 extracted rows on nonverbal and vocal channels and 873 rows separating romantic love from sexual desire. It converges on one conclusion: single body-language cues are weak evidence, repeated reciprocal investment is stronger evidence of relational interest, and explicit voluntary communication is the only safe basis for sexual escalation.
Each row is an observable behaviour cluster. Cell darkness shows how consistent that behaviour is with each state in the research — hover any row for its specificity, major confounds and the correct action.
There is no scientifically valid body-language decoder for a woman's inner state. Smiling, eye contact, laughter, touch, proximity, hair touching, lip behaviour, posture and voice all occur in friendliness, anxiety, impression management, cultural display and attraction. What separates signal from noise is her baseline, selectivity, change from baseline, repetition, reciprocity and context — never one cue. Misperception runs the other way too: observers detect the absence of attraction better than its presence, and men overperceive sexual intent more than women in study after study.
Compare her behaviour toward you against her behaviour toward other people in the same context, and against her own behaviour in other contexts. Ask five questions of any warm signal: is it selective (toward you specifically), repeated (across occasions), reciprocal (she initiates as well as responds), costly (effort, time, follow-through), and freely chosen (no role, script or pressure requires it)? Compare her to herself — never to stereotypes of women.
Science cannot certify inner indifference from behaviour, so this is graded by how strongly the behaviour controls what you may do. Any boundary controls behaviour regardless of its hidden motive.
Supported: attention and gaze track interest at group level; love-focused viewing dwells more on the face, desire-focused viewing more on the body (eye-tracking, group averages). Confounds: listening style, culture, anxiety. Cannot infer: her state from where she looks in one moment; “micro-expressions” are transient facial movements, not validated mind-reading.
Supported: open, expansive posture raises approachability ratings for both sexes. Confounds: comfort, seating, habit. Cannot infer: intent from stillness — freezing or passive compliance is not desire.
Moore's 52-behaviour courtship catalog was observational and context-bound. Preening, hair touching and lip behaviour are low-specificity: grooming, habit, dry lips. Not diagnostic for an individual.
Pupil size is strongly driven by luminance and general arousal; flushing and breathing by temperature, alcohol, exertion, anxiety. Nonspecific, largely involuntary, and never consent.
Supported: chosen closeness and initiated touch correlate with liking at group level. Confounds: loud venues force proximity; crowding and culture set touch norms; the same touch conveys friendliness and attraction. Never reciprocate with an unsolicited touch “test”.
Vocal-pitch findings depend on task, strategy and culture — a 22-country study found perceptions vary with relational mobility. Verbal content is the strongest channel: questions, disclosure, future talk, and above all explicit words. Being perceived as attractive is a different research question from her signalling attraction.
These are distinct motivational systems that often coexist and often do not: in Tennov's classic survey, 61% of women agreed they had been in love without any need for sex, and 53% had felt sexual attraction without a trace of love. Infer investment from a repeated pattern of voluntary time, selective attention, reciprocal disclosure, follow-through and social integration. Infer immediate desire only from explicit sexual communication plus active ongoing reciprocity. Romantic interest never implies sexual consent; sexual desire never implies commitment. Vulnerability or a social-circle introduction is data only inside a repeated pattern — never an invitation.
Warmth is role-required. Default reading: courtesy, unless there is clear, voluntary, off-role reciprocity she initiates.
Loudness forces proximity and shouting, low light dilates pupils, crowding changes touch and distance, alcohol impairs everyone's judgment — nonverbal specificity is lower here, not higher. Intoxication invalidates consent.
Quiet contexts make conversational investment easier to observe — still evidence of interest at most, never proof of desire. In friend groups, watch selectivity relative to how she treats the group.
The context itself signals openness to evaluation, not agreement to anything. Continue to read pattern and ask; even mid-date, only words escalate.
More informative: who initiates, substantive replies, question reciprocity, remembering details, repairing interrupted threads, making and keeping plans, moving the conversation forward. Less informative: raw response latency and emoji counts — schedules, attachment style, platform norms and communication style confound both. The emoji–dating association and the next-morning-text result are probabilistic group effects, not rules. Explicit sexting is real evidence of desire in that exchange — and grants no consent to meet, touch, record, share, or continue later.
The step-closer / light-touch “test” is corrected here: never use touch or proximity as an unsolicited diagnostic probe. Use a reversible, low-pressure verbal bid — “Would you like to get coffee sometime?”, “Can I sit closer?”, “May I kiss you?” — and read three states:
Proceed with that specific step, keep checking in.
Pause and ask plainly; accept the answer without pressure.
Stop warmly, no persuasion, no repeat attempts.
Consent must be voluntary, specific, informed, ongoing and revocable. Silence, freezing, intoxication, previous consent or passive nonresistance is not a yes.
Evolutionary note: Sexual Strategies Theory (Buss & Schmitt) organises hypotheses about long- and short-term tradeoffs and flexible strategies. It is a framework over population averages — it licenses no deterministic claim about an individual woman, and hypotheses, self-reported preferences, observed behaviour and consent are four separate things.
| Behaviour cluster | Specificity | Major confounds | Correct action |
|---|---|---|---|
| Smiling, warmth, laughter | Very low | Friendliness, politeness, nervousness, role requirements; used for both platonic warmth and attraction (Muehlenhard et al.) | Enjoy the conversation; infer nothing yet |
| Sustained eye contact / gaze | Low | Cultural norms, listening style; observers detect absence of attraction better than presence (Hall et al. 2015) | Note it only as part of a repeated cluster |
| Proximity, leaning in | Low | Loud venues force closeness; crowding and seating change distance | Check whether it survives a quiet context |
| Light social touch | Low | Conveys both platonic friendliness and attraction; norms vary widely | Never mirror with unsolicited touch — use a verbal bid |
| Preening, hair, lip behaviour | Very low | Grooming habit, anxiety, dry lips; courtship catalogs are observational, not diagnostic | Discount as evidence |
| Voice change (pitch, softness) | Low | Task- and strategy-dependent; pitch perceptions vary across 22 cultures (Aung et al. 2024) | Do not read the voice as a verdict |
| Pupils, flushing, breathing | Nonspecific | Luminance, alcohol, caffeine, general arousal, fear; largely private and involuntary | Never treat physiology as interest — and never as consent |
| Substantive questions, self-disclosure | Moderate | Extraversion, professional rapport; must be selective toward you vs. everyone | Compare against her baseline with others |
| Repeated voluntary time, plans kept, future inclusion | Higher | Close friendship can look identical for a while; needs repetition and reciprocity | Strongest ground for a low-pressure verbal invitation |
| Digital reciprocity: initiating, remembering, plan-making | Higher | Schedules, attachment style and platform norms confound latency and emoji | Weigh substance and continuity, not response time |
| Explicit romantic words | High | Little; words can be tested gently for seriousness over time | Believe her; reciprocate honestly |
| Explicit sexual language + active reciprocal participation | Highest | Intoxication invalidates it; consent is specific, ongoing and revocable | Ask before each step; keep checking in |
| Short replies, looking away, one cancelled plan | Ambiguous | Workload, anxiety, safety habits, neurodivergence, culture | Do not push; one gentle invitation, then accept the pattern |
| Declined invitations, no counteroffer, moving away, freezing | Strong pattern | Freezing can be fear, not indifference — it still means stop | Stop warmly; do not test again |
| “No”, “not interested”, ending contact, blocking | Definitive | None that matter — a boundary controls behaviour regardless of motive | Stop completely; no follow-up |
Labels mark study type. Excluded: the retracted Moore/Butler “Courtship compliance” result, low-quality commercial dating pages, and Wikipedia rows for substantive romantic-versus-sexual claims.
Method: structured web research synthesis current to 2026-09-02, not a systematic review or meta-analysis. Sources: a 56-row cross-module evidence shortlist (baseline/interest, romantic vs sexual, digital, consent), a 947-row extracted pool on nonverbal and vocal channels, and an 873-row pool separating romantic love from sexual desire; six broad academic search tracks surfaced thousands of candidate pages. Matrix cell values are editorial calibrations of the cited evidence, not measured frequencies. Many underlying studies use Western, heterosexual, college, speed-dating, vignette or laboratory samples; women are heterogeneous, and culture, orientation, age, personality, trauma history and neurodivergence shape signalling. Retracted, commercial and low-quality popular rows were cut.