Choosing a baby name with a partner is a two-person decision with an unusual requirement: the outcome has to be something both people independently like, not something one person proposed and the other accepted. The way most couples actually run the process — one person reads candidate names aloud, the other reacts — works against that requirement in a couple of well-understood ways. This post lays out a process that avoids those problems, why it works, and how we built an app around it.
Why reading names aloud produces bad data
Two effects are doing the damage. The first is anchoring: your evaluation of a new option is pulled toward whatever signal arrives with it — first catalogued by Tversky and Kahneman (1974) in numeric estimation tasks, and one of the most consistently replicated findings in judgment research. Related conformity experiments (Asch, 1956) found that people will revise even unambiguous perceptual judgments after hearing a unanimous group disagree. If a name reaches you already wrapped in your partner's enthusiasm (or their pause, or their raised eyebrow), the reaction you form is a reaction to the name plus the signal— and you can't easily subtract the signal afterward. Couples describe the result as "I can't tell anymore whether I like this name or whether you do." That description is more accurate than it sounds.
The second is socially motivated softening. When your partner floats a name they clearly like, an honest "no" has an interpersonal cost, so it tends to come out as "maybe." The general pattern is well documented: people reliably withhold or blunt unwelcome messages to spare the recipient — the "MUM effect" (Rosen & Tesser, 1970) — and politeness theory treats refusals as exactly the kind of act speakers instinctively hedge (Brown & Levinson, 1987). Each softened answer is a small kindness and a small data error. Over a few weeks, the shortlist fills with names that survived on politeness rather than genuine preference, and it becomes hard to know which entries are real.
Both effects have the same structural fix: collect each person's reactions privately, before any signal from the other person exists, and only then compare. Agreement found this way has a useful property — when a name appears on both private lists, you know the preference is genuine on both sides, because neither reaction could have influenced the other. The value of protecting independence isn't hypothetical: in group-estimation experiments, even mild exposure to others' judgments collapsed the diversity of a group's answers and raised individual confidence without improving accuracy (Lorenz, Rauhut, Schweitzer & Helbing, 2011). Once independent judgment is contaminated, it doesn't come back.
The process
- Agree on constraints first.Before evaluating any specific name, settle the boundary questions: popularity tolerance, family names that must be considered, initials or letters to avoid, whether surname-style first names are in scope. This conversation is low-stakes precisely because no one's favorite is on the table yet.
- Evaluate a large pool separately.Each person sorts names into roughly love / like / no, alone, at their own pace. Volume matters: stated preferences ("I like classic names") are unreliable predictors of reactions to actual names, so the useful signal comes from reacting to hundreds of candidates rather than brainstorming a handful.
- Allow unexplained vetoes. Some names are unusable for reasons that are private — a former relationship, a difficult colleague. In a live conversation, striking such a name invites a question; in a private process, a veto needs no justification. This keeps the veto honest and cheap to use.
- Compare only the intersection.Names on both "love" lists are the shortlist. Names that are a love for one person and a like for the other are worth a conversation each. Names only one person liked can be dropped without discussion — and because the comparison only surfaces overlap, no one has to watch their favorites get rejected in real time.
- Revisit over time. Preferences drift across a pregnancy. A process that lets you re-sort earlier judgments (promote a like, demote a love) tracks that drift instead of freezing your February taste in place.
None of this requires software. Two notebooks and an evening of comparing lists implement steps 1–4 faithfully. The limits of the notebook version are practical: sourcing enough candidate names matched to your constraints, keeping the sorting effortless enough that you actually do it, and noticing the moment two private lists overlap.
How Pearfect Name maps onto the process
Pearfect Name is our attempt to build this process directly into software. The design constraint throughout is the one from the research above: no partner signal should reach you before your own judgment is recorded. Concretely:
- Independent queues and preferences.Partners pair via QR code (no accounts required), then each set their own filters — gender, style categories, origins, popularity range — and calibrate the app with a short elimination-style training round. Each person's swipe queue is generated from their own preferences, so the pools can differ.
- Swipes are private until they intersect. A partner never sees your dislikes, vetoes, or unmatched likes. The app reports only mutual positives — the intersection from step 4. This is the anchoring fix implemented as an information barrier rather than as self-discipline.
- Four response levels instead of two.Like, dislike, super-like, and hard-no. The hard-no is the unexplained veto from step 3: it removes the name permanently, and no one is told. Super-like distinguishes "genuinely love" from "fine," which makes the eventual intersection more informative. Any past swipe can be reclassified later (step 5) via a long-press.
- Constraint tools.Step 1 lives in settings: alongside category and popularity filters, spelling patterns can be boosted, suppressed, or excluded outright ("ends in -son", "starts with K"). The catalog is 11,600+ names, each tagged with origins and style categories — it's browsable free on this site if you want to inspect the data.
- Deliberate, opt-in influence — after independence has done its job.Once your own preferences are on record, seeing your partner's can be productive rather than contaminating. Pare Squared (premium) restricts your queue to names your partner already liked — useful for a focused shortlist session. Peer Review groups categories by agreement level ("We're Obsessed," "United Front," "Let's Talk About"), which is effectively the step-4 conversation agenda computed for you. The distinction we care about is timing: influence as a stage you choose to enter, not a default you start in.
- Support tooling.AI recommendations (premium) model both partners' swipe histories to surface candidates outside either person's search habits. A full-name preview checks first-middle-last combinations and flags initials that spell something unfortunate. Nerd Stats reports agreement rate and swipe pace. Swiping works offline and syncs when connectivity returns.
The free tier includes the full core process — unlimited swiping, matching, filters, pairing. Premium ($4.99/month with a one-month trial, or a one-time $19.99 lifetime unlock) adds the AI recommendations, Pare Squared, full Peer Review, and the themed name packs.
What the process buys you
A name both people privately loved before either knew the other did is a different object from a name one person proposed and the other accepted. The decision is more robust — there's no suppressed reservation waiting to resurface in month eight — and the moment of discovering the overlap is, reliably, one of the nicer moments of expecting. The process is designed to protect exactly that.
If you're choosing a tool, we've written an honest comparison of the options — the best baby name apps for couples in 2026 — including the apps we'd point Android households to, since ours is iPhone-only.
References
A note on scope: the studies below document these mechanisms in general settings — numeric estimation, perceptual judgment, everyday conversation — not baby naming. As far as we can find, no one has run the experiment on couples choosing names; applying these findings to naming is our inference, and we've tried to keep the claims proportionate. (The closest thing to a naming literature is sociological: Rossi's 1965 survey of family naming practices remains the classic.) Conformity effects in particular vary in magnitude across cultures and decades, so treat the Asch numbers as illustrative rather than universal.
- Asch, S. E. (1956). Studies of independence and conformity: I. A minority of one against a unanimous majority. Psychological Monographs: General and Applied, 70(9), 1–70. doi:10.1037/h0093718
- Brown, P., & Levinson, S. C. (1987). Politeness: Some universals in language usage. Cambridge University Press.
- Lorenz, J., Rauhut, H., Schweitzer, F., & Helbing, D. (2011). How social influence can undermine the wisdom of crowd effect. Proceedings of the National Academy of Sciences, 108(22), 9020–9025. doi:10.1073/pnas.1008636108
- Rosen, S., & Tesser, A. (1970). On reluctance to communicate undesirable information: The MUM effect. Sociometry, 33(3), 253–263. doi:10.2307/2786156
- Rossi, A. S. (1965). Naming children in middle-class families. American Sociological Review, 30(4). doi:10.2307/2091340
- Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. Science, 185(4157), 1124–1131. doi:10.1126/science.185.4157.1124
