Subalpine forests in North America and Japan dominated by Abies spp. show an unusual pattern of regeneration in which recruitment is confined to a narrow window in time and space that just precedes the death of the largest trees. Previously, Silvertown suggested that selection in such forests should favour delayed reproduction. A graphical model supporting this prediction also suggests that trees in ‘normal’ forests should benefit from precocious reproduction when they are taller than their neighbours, while trees in wave populations should not. Here, we present a field test of the two hypotheses based upon a comparison of the life history of trees in adjacent wave and non–wave populations at Whiteface Mountain, New York. The results show that reproduction commences at a similar age in both kinds of forest, and that trees in the subalpine zone reproduce later and die earlier than conspecifics in lowland forests. The failure of the first hypothesis can be explained by modifying our original assumptions about how reproductive costs and benefits vary with age in the model. In our test of the second hypothesis, we find that the reproductive behaviour of individual trees in the two kinds of forest is different and consistent with our prediction. Phenotypic plasticity for age at first reproduction appears to be present only in the population where it is adaptive.