Dating advice often places a citation beside a confident instruction and asks the reader to assume the bridge between them is sound. Sometimes the study is real but measured something narrower. Sometimes the effect is small. Sometimes a single laboratory produced the whole result. Sometimes the paper was later retracted.
You Can’t Read Her uses more than 80 sources, including peer-reviewed articles, meta-analyses, retraction records, author statements and historical reports. They are identified for what they are. Nothing becomes peer-reviewed because it appears in a bibliography.
The five evidence labels
| Label | What earns it | How to use it |
|---|---|---|
| Strong | A meta-analysis, large validated measure or finding replicated by independent research groups. | A reasonable planning assumption, with individual variation kept in view. |
| Moderate | Consistent evidence, often from one research program or with limited independent replication. | A good bet inside the conditions the research actually tested. |
| Emerging | Limited, context-dependent or contested evidence. | A hypothesis worth watching, not a rule about another person. |
| Speculative | Theory, thin sourcing, compromised data or a practical leap beyond the evidence. | Hold it lightly until stronger evidence arrives. |
| Dead | A retracted result, a claim that failed a large preregistered test or advice that never had supporting evidence. | Do not build a tactic on it. |
The labels are editorial judgments, not a published scientific grading standard. They make the author’s confidence visible so readers can disagree with it.
The source-checking workflow
- Trace the claim to the original source. A popular article or advice book may paraphrase a study that said something narrower.
- Name the measured outcome. Flirting, perceived sexual intent, self-reported desire, approach frequency and consent are different variables.
- Keep the sample attached. Undergraduates, speed dates, bar observations and nationally representative samples support different levels of generalization.
- Look for magnitude, not only significance. A small association can be publishable and still be useless for reading one individual.
- Check replication and later history. Meta-analyses, preregistered tests, expressions of concern and retractions can change what an old result deserves.
- Mark the practical bridge. If the study did not test the advice, the recommendation is identified as an inference.
Two numbers readers meet often
d describes a gap between group averages. A larger value means the averages are farther apart. It does not eliminate overlap, and it cannot tell you where one person falls.
r describes a linear association. It runs from −1 to +1. Squaring it gives the approximate share of variance accounted for. An association of r = .35 explains about 12% of the variance, which sounds different from the phrase “statistically significant.”
Effect sizes help keep a study’s practical importance in proportion. They do not rescue a poor design or turn a population pattern into an individual diagnosis.
One example: reported flirting
The Hall, Xing and Brooks study is often reduced to a viral accuracy statistic. The book keeps the denominator and the measurement visible. Partners agreed with 7 of 25 self-reports of flirting and 66 of 79 self-reports of not flirting in one sample of 104 single heterosexual college students.
The result earns a Moderate, context-limited label. It supports humility about first impressions. It does not measure attraction or consent, and it does not establish a universal percentage for men or women.
Retractions are named, not quietly removed
Five papers from one field-research corpus were retracted online on February 25, 2025. The affected topics included courtship tactics that had circulated widely in popular advice. The book identifies the retractions and separates them from other problems, such as small effects, weak sources or a failure to replicate.
A retraction removes that paper as support. It does not automatically prove the opposite of every broader claim. See The Graveyard for the claim-by-claim audit and the dated corrections record for wording that differs in the current PDF.
Limits of the book’s own method
- The author is not presenting himself as an academic, clinician or dating coach.
- The five labels involve judgment. Another careful reader may grade a study differently.
- Coverage is extensive for the book’s claims, not a systematic review of every dating study ever published.
- Many studies in this field rely on young, heterosexual or undergraduate samples.
- A practical method can be reasoned from research without having been tested as a complete intervention. Those steps are called inferences.
Selected sources
The complete Kindle edition contains the full source list. These papers anchor the public research guides on this site.
- Hall, J. A., Xing, C., & Brooks, S. (2015). Accurately detecting flirting.
- Joel, S., Eastwick, P. W., & Finkel, E. J. (2017). Is romantic desire predictable?
- Farris, C., Treat, T. A., Viken, R. J., & McFall, R. M. (2008). Perceptual mechanisms in decoding women’s sexual intent.
- Lee, A. J., et al. (2020). Sex differences in misperceptions of sexual interest.
- Eastwick, P. W., Finkel, E. J., Mochon, D., & Ariely, D. (2007). Selective versus unselective romantic desire.
- Lehmann, G. K., Elliot, A. J., & Calin-Jageman, R. J. (2018). Meta-analysis of the effect of red on perceived attractiveness.
- Muehlenhard, C. L., et al. (2016). The complexities of sexual consent among college students.
- Reis, H. T., et al. (2011). Familiarity does indeed promote attraction in live interaction.
- Teichmann, L., et al. (2026). How the timing of texting triggers romantic interest after the first date.
See the method in use
Read Part One free
The 38-page sample applies this method to flirting detection, romantic prediction and retracted dating research. Every grade appears beside its source.