Two mechanisms, two windows. Between gestational day 17 and 22, before most women have tested, alcohol switches on a cell-death signal in the cells building the face; where enough die you get the signature: smooth philtrum, thin upper lip, narrowed eyes ¹. The brain is a slower target, killed and mis-wired through both later trimesters ². No shared dose threshold, no single window, so the answer covers all three trimesters.

The low-dose question is the only one still open, and the honest answer isn't "safe." Miscarriage risk climbs from the very first drink, about 6% per drink per week, with no floor below which it goes flat ³. One or two glasses a week look harmless in the data only because light drinkers are wealthier and better-educated, and their kids score higher for reasons unrelated to the wine. Strip that confounding out with a genetic natural experiment and even modest first-trimester drinking shows a small, real drop in childhood IQ ⁴. Its apparent safety is absence of evidence, not evidence of absence ⁵.

The trigger is "could be pregnant," well before "confirmed pregnant." The face-building window opens roughly two weeks after a missed period, before most tests come back positive ⁶.

If you drank before you knew, tell your obstetrician plainly. Honest disclosure gets the child screened early, the single biggest protective factor there is ⁷.

The visible cases are the tip. The child with the small head and flattened face gets diagnosed at birth. The larger group reads normal until around age five, when a teacher starts using words like "regulation" and "impulse," and by second grade there's a diagnosis nobody flagged during the pregnancy. Among fetal-alcohol adults followed longest, most have a mental-health diagnosis, most don't live independently, and roughly one in eight has ever held a job ⁸.