Chemical basis of pupal cannibalism in a caterpillar (Utetheisa ornatrix).

The moth Utetheisa ornatrix derives protection against predation from systemic pyrrolizidine alkaloids (PAs) that it sequesters as a larva from its foodplants (Leguminosae, Crotalaria spp.). We here show, in laboratory tests, that Utetheisa deficient in body PA can make up for the chemical shortfall by cannibalizing pupae. We present evidence indicating that cannibalism in larvae is elicited not by hunger, but possibly by PA deficiency itself, and that in making cannibalistic choices larvae prefer PA-containing over PA-free pupae. PAs themselves, either in crystalline form or as additives to food items, proved phagostimulatory to larvae. In nature Utetheisa tend to pupate away from their foodplant, essentially out of reach of larval attack. The threat of cannibalism may have contributed to the evolution of this pupation behavior.     

Overcompensation and population cycles in an ungulate.

Although theoretical studies show that overcompensatory density-dependent mechanisms can potentially generate regular or chaotic fluctuations in animal numbers, the majority of realistic single-species models of invertebrate populations are not overcompensatory enough to cause sustained population cycles. The possibility that overcompensation may generate cycles or chaos in vertebrate populations has seldom been considered. Here we show that highly overcompensatng density-dependent mortality can generate recurrent population crashes consistent with those observed in a naturally limited population of Soay sheep. The observed interval of three or more years between crashes points to sharp 'focusing' of mortality over a narrow range of population density.     

On the limitations of comparing mean square forecast errors

Linear models are invariant under non-singular, scale-preserving linear transformations, whereas mean square forecast errors (MSFEs) are not. Different rankings may result across models or methods from choosing alternative yet isomorphic representations of a process. One approach can dominate others for comparisons in levels, yet lose to another for differences, to a second for cointegrating vectors and to a third for combinations of variables. The potential for switches in ranking is related to criticisms of the inadequacy of MSFE against encompassing criteria, which are invariant under linear transforms and entail MSFE dominance. An invariant evaluation criterion which avoids misleading outcomes is examined in a Monte Carlo study of forecasting methods.