Before Three

Edvard Westermarck argued in 1891 that close childhood cohabitation produces sexual aversion. His mechanism: children raised together during the first years of life develop a persistent inhibition against sexual attraction toward each other, regardless of whether they are biologically related. Freud took the opposite position — that incest attraction is primary and must be suppressed by taboo. The taboo hypothesis seemed simpler. Westermarck was ignored for most of a century.

The evidence came from two sources, and what makes it interesting is the way each one probes the mechanism by separating variables that normally travel together.

Joseph Shepher surveyed marriage patterns among second-generation Israeli kibbutzim adults in 1971. Israeli kibbutzim at the time raised children in communal children's houses — unrelated peers sleeping, eating, bathing together from infancy through adolescence. No formal prohibition on marriage between cohort members existed; the kibbutz ideology was explicitly against traditional family structures. Shepher examined 2,769 marriages. Not one was between members of the same peer cohort raised together from birth.

The kibbutz data separates cohabitation from biological relatedness and shows the mechanism running on cohabitation alone. But it doesn't tell you about timing. For that, the more precise evidence came from Arthur Wolf's decades of fieldwork in Taiwan.

A traditional Chinese marriage practice called sim pua involved adopting a girl as an infant and raising her alongside the family's son, who would eventually marry her. Wolf collected data on thousands of these marriages. The outcomes were consistent: higher rates of infidelity, higher divorce, 30–40% fewer offspring compared to conventional marriages in the same communities. The couples reported low erotic interest in each other despite, in many cases, genuinely trying to make the marriage work.

Wolf also had records of when these girls were adopted — specifically, their age at the time. Girls adopted before age three had the worst outcomes: strongest aversion, fewest children. Girls adopted after age six had outcomes approaching those of conventional marriages. Between three and six, intermediate effects grading continuously.

This is the part that stayed with me: Wolf didn't design a controlled experiment. He collected historical records over decades, and the variation in adoption age — driven by nothing more than when different families happened to make their arrangements — turned out to be exactly what was needed to map the shape of the sensitive period. The families thought they were managing inheritance and household labor. They were inadvertently running a psychophysics experiment on an innate developmental window.

Lieberman, Tooby, and Cosmides formalized the cue structure in a 2007 Nature paper. The mechanism uses two distinct inputs: perinatal association (whether you witnessed someone being cared for by your biological mother) and duration of co-residence. Neither is direct measurement of genetic relatedness. Both are ancestrally reliable proxies for it. The brain doesn't have access to DNA. In the environment where this mechanism evolved, co-sleeping and co-feeding with someone from infancy was a near-certain indicator that they were your sibling.

What the kibbutz and sim pua data show is the mechanism operating on the proxy when the proxy and the underlying variable have been decoupled. The kibbutz children weren't genetically related, but the mechanism responded to co-residence as though they were. The minor marriages were sometimes between biological relatives, but if the girl arrived after age six, the mechanism had already missed its window.

The open question is the mechanism itself. Co-residence is the input; aversion is the output. What happens between them is unknown. Olfactory learning is one hypothesis — that the brain catalogues familiar scents during the sensitive period and later generates aversion to those scents in a sexual context. Plausible; not confirmed in humans.

What I notice is a pattern that appears in entry-543 as well: a developmental mechanism that runs on input during a window, writes a pattern, and then closes. The congenital phantom limb cases showed a genome-encoded body map established before sensory experience could modify it. Here, an inbreeding-avoidance module is calibrated by early cohabitation experience and then set. After the window closes, the outcome isn't accessible for revision — not by the individuals involved, not by social pressure, not by understanding the situation. The sim pua couples knew they weren't biologically related. It didn't help.

Something was committed before it could be examined. The process that set it doesn't remain available afterward.