The importance of (experimental) design
02-25-2014 - One of the hottest debates in evolutionary biology concerns the origin of behaviour: is it genetically encoded or do animals and birds copy their parents or other individuals? A classic experiment published in 2000 seemed to provide overwhelming evidence that a particular behavioural choice (whether individuals of a species of swallow breed in a small colony or a large one) is largely genetically determined. Together with colleagues in France, Richard Wagner of the University of Veterinary Medicine, Vienna has re-examined the data and shown that the findings could be explained by random choice. The design of the original experiment – which represents a blueprint for a vast range of studies of heritability of behaviour – contains two pitfalls that combine to undermine the conclusions. The results are published in the journal Scientific Reports.
There are clear advantages to living in cities: safety, ready availability of infrastructure, plenty of company etc. Nevertheless, a large number of people eschew them for the benefits of country life, such as clean air and lots of space. Many species of animals, and particularly birds, face the same choice between living in large groups or remaining in smaller ones, thereby avoiding disadvantages of larger colonies such as the increased risk of disease and increased aggression from neighbours. What causes different individuals of a particular species to take the decisions they do?
It's all in the genes
One possible explanation is that animals and birds might be genetically influenced to breed in smaller or larger colonies. This idea gained widespread acceptance in 2000, when Charles and Mary Brown reported the results of a field experiment of unprecedented scope. The Browns worked on Cliff Swallows, a species of American swallow that breeds in colonies of variable size. Some individuals appear to be more resistant to nest parasites so can breed in larger, more parasite-infested colonies where they can achieve higher foraging efficiencies by following their many neighbours to the insect swarms on which they feed. The scientists cross-fostered an astounding 2,000 nestlings from nests in small colonies to nests in large colonies and vice versa and recaptured over 700 of the birds when they returned to breed in the following years. Analysis of the data showed convincingly that swallows hatched in larger colonies (but reared in smaller ones) chose to breed in large colonies, whereas the reverse was true for swallows hatched in small colonies. In other words, the birds’ choice is determined by genetic factors.
Or is it?
Appealing though the conclusion might be, something does not sound quite right. How would a genetic component of the choice of group size be selected and maintained throughout evolution? Étienne Danchin of the University of Toulouse and Richard Wagner of the Vetmeduni's Konrad Lorenz Institute of Ethology wondered whether the results might not be influenced too much by the experimental design. Together with Éric Wajnberg, a modelling specialist at the INRA in Sophia Antipolis, they simulated the data in the original paper and found that the results could be generated by chance.
The problem stems largely from the fallacy known as “regression to the mean” (RTM), which was first identified by Sir Francis Galton, a cousin of Charles Darwin, in the 19th century. It results from the fact that uncommonly large or small measurements are generally followed by more normal measurements simply because normal values are by definition far more common than extreme ones. In the Browns' experiment, when individuals are fostered from small to large colonies they will on average recruit to colonies that are statistically smaller than their foster colony because these are closer to the mean colony size and vice versa.
Danchin and Wagner actually found a second problem with the original data analysis. The so-called “spatial fallacy” was discovered by Arie van Noorwijk in 1984 and occurs when the set of potential dispersal sites differs among individuals according to where they were born. Together the two fallacies can account for the experiment’s results, even in the absence of an inherited component of the selection of colony size.
A question of design
The latest calculations are important for two reasons. First, they call the original conclusions into question and should thus encourage researchers to reconsider the role of genetic factors in certain behavioural decisions. Secondly, they sound a clear warning to all those working in behavioural science. As Wagner says, “The biggest shock is that the very experimental design that is widely used to control for all extraneous effects actually creates the RTM fallacy. Even the most carefully designed studies can suffer from the problem, making their results invalid.” Fortunately the scientists have proposed a way to avoid both pitfalls. Their simulations showed that restricting comparisons to individuals with the same set of options (in the present case by fostering all offspring to a single colony) can give rise to data that are immune to the pitfalls.
The Nature Publishing Group has just published the paper “Avoiding pitfalls in estimating heritability with the common options approach” by Étienne Danchin, Éric Wajnberg and Richard H. Wagner in its online journal Scientific Reports, where it is freely available. [Link 1](Scientific Reports Volume 4, doi:10.1038/srep03974)