Selection Differential

The selection differential is the gap between the parents you kept and the population you took them from, measured in the trait's own units. It is the input the breeder controls most directly, and the one most often overstated in retrospect.

Its standardised cousin, selection intensity $i$, expresses the same gap in standard deviations and depends only on the proportion selected. Keeping the top 10% gives $i \approx 1.76$; the top 5% gives 2.06; the top 1% gives 2.67. Those numbers show the diminishing return plainly — a tenfold cut in the number of parents buys only about 50% more differential.

The differential is signed, and a negative one is often exactly what is wanted. Shorter straw, earlier maturity, lower disease score and reduced lodging are all traits selected downwards, and the response follows the same direction.

The mistake worth guarding against is measuring the differential against the wrong baseline. It must be the mean of the population the parents were actually drawn from, in the same trial, in the same year. Comparing selected parents against a historic variety mean, or against a population grown elsewhere, folds environmental and year effects into a figure that is supposed to be a within-population gap — and the breeder's equation then confidently predicts a response that never arrives.