Most dating advice starts with the same promise: collect enough clues and you can know what a woman means before she tells you. The research is much less confident. People miss flirting. They also read sexual interest into friendliness. Large batteries of traits and preferences can predict who tends to want or be wanted, yet still fail to predict which two people will want each other.
That does not make dating random. It changes where useful information comes from. A private guess is weak evidence. A response to a clear, low-pressure offer is stronger.
1. Begin with uncertainty
In a study by Jeffrey Hall, Chong Xing and Seth Brooks, 52 pairs of single heterosexual college students spoke for 10 to 12 minutes. Afterwards, each person reported whether they had flirted and whether their partner had flirted. Partners agreed with only 7 of 25 self-reports of flirting. They agreed with 66 of 79 self-reports of not flirting.
Those numbers need their limits. The study measured agreement with one yes-or-no self-report. It did not measure attraction, sexual interest or consent. It does not mean every negative read is correct, or that anyone has a fixed “28% accuracy rate.” It supports a narrower conclusion: a first impression about flirting can be wrong, so treat it as a hypothesis.
2. Do not ask a profile to predict a pair
Samantha Joel, Paul Eastwick and Eli Finkel tested whether more than 100 pre-date traits and preferences could predict romantic desire in two undergraduate speed-dating samples. Their models predicted broad tendencies, such as who generally desired more people and who was generally desired more. A model trained on the first sample and tested on the second explained less than 0.1% of the pair-specific part of initial desire.
The study concerned four-minute, opposite-sex speed dates among young undergraduates. It did not test long-term compatibility. Still, it is a useful warning against false precision. A detailed profile can help you decide whom to meet. It cannot tell you how one particular meeting will feel.
In 2026, researchers asked a different prediction question: could a language model detect attraction after reading the date itself? The answer was a small, measurable signal with important limits. See what ChatGPT predicted from 964 speed-date transcripts.
3. Look for a response, not a tell
Isolated cues are easy to overread. A smile can mean warmth, politeness, nervousness or attraction. Fast texting can reflect enthusiasm, free time or habit. Eye contact can be mutual interest, simple attention or the way someone speaks with everyone.
A response becomes more informative when it answers a specific offer. You suggest coffee on Thursday. She accepts, proposes another time or keeps the exchange moving. That is different from interpreting the angle of her feet across a room. It still does not authorize a bigger assumption. It tells you what she chose in that moment.
4. Keep the answer usable
The quality of the information depends on whether the other person can answer freely. An invitation becomes harder to read when it carries pressure, repeated asks, public embarrassment, a demand for reasons or a consequence for saying no.
A useful invitation is specific enough to answer and light enough to decline. “I’ve enjoyed talking with you. Would you like to get coffee Thursday?” works. So does, “I’d like to keep this going, only if you would.” Ask once. If the answer is no, hesitant or persistently unclear, stop trying to turn it into yes.
Five popular dating claims checked
This is a selected evidence check, not a systematic review of every dating study. Each row names the measured result, later record or boundary that matters to the practical claim. The sources were reviewed for this page on August 10, 2026.
| Popular claim | Evidence status | What was actually measured | Where to check it |
|---|---|---|---|
| “You can learn the signs she likes you.” | Moderate, context-limited evidence that self-reported flirting is often missed. | In one sample of 104 single heterosexual students, partners agreed with 7 of 25 reports of flirting and 66 of 79 reports of not flirting. | Read the flirting analysis. |
| “A profile can predict whether two people will click.” | Not supported for pair-specific initial desire in the tested samples. | More than 100 pre-date measures predicted broad tendencies. A model tested on a second sample explained less than 0.1% of the pair-specific component. | Joel, Eastwick and Finkel. |
| “Wearing red reliably makes a man more attractive.” | Very small and uncertain estimate. | For women rating men, a 2018 meta-analysis estimated d = 0.13 and found evidence of publication bias. Its preregistered studies averaged d = −0.10. | See the claim audit. |
| “A light touch is a proven courtship tactic.” | The two papers used for that tactic were retracted. | Two touch-and-compliance articles were among five papers retracted online on February 25, 2025. The notice cited serious unresolved methodological, ethics and data concerns. | Read the retraction notice. |
| “The book’s invitation method is experimentally proven.” | No. It is the author’s practical inference. | The cited studies support premises about uncertainty, prediction and misperception. They did not test the complete invitation method as an intervention. | See the grading method and limits. |
The public Attraction Evidence Ledger extends this audit to 15 recurring claims and makes the evidence table downloadable as CSV or JSON.
What “evidence-based” should mean
A citation is a starting point, not a seal of approval. Before advice earns confidence, ask what the study measured, who participated, whether the result replicated, how large the effect was and whether the practical claim goes beyond the finding.
Find the original
Advice often cites a summary that cites a book that cites a study. Trace the claim back to the paper and check whether the recommendation appears there.
Keep the sample visible
A finding from undergraduates, speed dates or one laboratory can be useful. It should not be rewritten as a universal law of women.
Separate findings from inference
Research may support a premise without testing the advice built from it. Label that bridge as the author’s inference.
Check what happened later
Retractions, larger studies and failed replications can change what an older result deserves today.
The useful rule
You do not need certainty before you show interest. You need proportion. Let weak evidence support only a small next step. Keep that step easy to decline. Then update from the answer instead of defending the original guess.
That is less dramatic than a body-language decoder. It is also more honest, more respectful and more likely to tell you what you actually need to know.
Research cited
- Hall, J. A., Xing, C., & Brooks, S. (2015). “Accurately detecting flirting.” Communication Research, 42(7), 939–958.
- Joel, S., Eastwick, P. W., & Finkel, E. J. (2017). “Is romantic desire predictable?” Psychological Science, 28(10), 1478–1489.
- Farris, C., Treat, T. A., Viken, R. J., & McFall, R. M. (2008). “Perceptual mechanisms that characterize gender differences in decoding women’s sexual intent.” Psychological Science, 19(4), 348–354.
- Eastwick, P. W., Finkel, E. J., Mochon, D., & Ariely, D. (2007). “Selective versus unselective romantic desire.” Psychological Science, 18(4), 317–319.
- Lehmann, G. K., Elliot, A. J., & Calin-Jageman, R. J. (2018). “Meta-analysis of the effect of red on perceived attractiveness.” Evolutionary Psychology, 16(4).
- Sage Publishing. (2025). “Retraction notice” covering five articles from one field-research corpus. Perceptual and Motor Skills.
Study summaries above state the limits that matter to the practical claim. See how the book grades evidence.
From the research to the real conversation
Read the first three chapters before you buy
You Can’t Read Her checks familiar dating advice against more than 80 cited sources, labels uncertainty and gives you a practical method for acting without pretending to know.