doi:10.1016/j.anbehav.2005.12.002

Hormones and Animal Social Behavior. By ELIZABETH ADKINS-REGAN. Princeton, New Jersey: Princeton University Press(2005), Pp. xviþ411. Price £29.95 paperback.

‘Mice live fast and die young, producing large litters inrapid succession if they survive at all. Elephants live slowand die old, producing a single offspring at long intervals.A male anglerfish spends his entire adult life attachedparasite-style to a female whose size dwarfs his, andneither sex reaches maturity until it finds the other .’In a world where genomes are ‘tickled’, siblings ‘bedamned’, and the ‘irony of nature’ provides a vertebratewith changeable sex type, one cannot help but delvedeeper into Adkins-Regan’s review of hormones at work.

With this book Adkins-Regan takes us on a tour throughcurrent concepts, theories and hypotheses in behaviouralendocrinology. The focus is on naturally occurring socialbehaviour, and the network idea of behaviour, thatincludes the interaction of an actor with other individuals.Hormones (predominantly steroids) are presented ascoordinators of reproduction, of suites of physiologicaland behavioural components, of different parts of thebrain, and of brain with body. The book progressivelybuilds up from the internal hormonal mechanisms in

ANIMAL BEHAVIOUR, 71, 3740

doi:10.1016/j.anbehav.2005.12.001

Neural Networks & Animal Behavior. By MAGNUS ENQUIST &STEFANO GHIRLANDA. Princeton, New Jersey: PrincetonUniversity Press (2005), Pp. ixþ253. Price £26.95paperback.

There are many who view neural networks simply asabstract models to investigate properties of neural systemsor tools for statistical data analysis. Although other areas ofscience and technology have embraced neural networks,our field has been reluctant to adopt such tools. Readers ofthis book, however, will be left in little doubt that neuralnetworks could become an important tool for studyinganimal behaviour, and its application must be encouraged.Neural networks will not appeal to everyone, nor will theyhave useful application in all areas of research. Indeed, bythe authors’ own admission, simple neural networks mayprove to have restricted utility, but the small body of workdone to date shows promising signs and suggests thefeasibility and usefulness of neural networks is far fromdecided. Enquist and Ghirlanda have presented neuralnetwork modelling in an easy to follow manner, makingit accessible to a broad spectrum of researchers. Theydescribe how such models have been used, and explainhow they could be developed to replicate experimentalfindings. In so doing, they have presented a clear case forthe use of neural networks in understanding behaviour.

The book is divided into six chapters, commencing withintroductions to the study of animal behaviour and neuralnetworks, and then describes the application of neuralnetwork models to a wide variety of topics within animalbehaviour. Their descriptions are intuitive, and likely to beaccessible to advanced level undergraduate and postgrad-uate students as well as more established researchers. Eachchapter concludes with a summary of the key pointspresented within the chapter as well as a reference list forrelevant further reading. Readers will also find useful thecomprehensive bibliography and subject index providedat the end of the book.

The first chapter provides background to studyinganimal behaviour and particularly the conceptual frame-work upon which all models are described, the ‘behaviourmap’. This is followed by an overview of the neuralnetwork methodology. For many who flip through thepages of this chapter, the prospect of dealing with morethan 50 mathematical equations will encourage them toheed the authors’ advice and skip most of this chapter(see Preface). However, the real value of this book willbe realized if readers make an effort to understand thetechnical side of this research tool. That being said, thepresentation of mathematical equations is done in sucha way that readers are eased into the more complicatedalgorithms. This is one of the more straightforwardintroductions to modelling that I have read.

The central chapters of this book describe the applica-tion of neural networks to three familiar subject areaswithin animal behaviour: mechanisms, learning andontogeny, and evolution. The first of these chaptersfocuses on the properties of a behaviour map and include

stimulus control, sensory processing, decision making andmotor control. The next two chapters deal with changesto the behaviour map. Unlike the preceding chapterwhich deals with input–output relationships, the nexttwo chapters examine the way in which these relation-ships are modified through learning and development, aswell as the process of evolution. Both add complexity tothe modelling process. As a consequence, the modelspresented are not well developed and only begin toreplicate the many different experimental findings andknowledge; a point that the authors make from the outset.The chapter summary for ‘learning and ontogeny’ providesan overview of what the authors feel is adequately repre-sented in neural network models to date, but also whatremains to be done. Similarly, the evolution chapter isdescribed as speculative. Enquist and Ghirlanda admit thatthere is still no clear picture of the role that these modelswill play in understanding the evolution of behaviour.

Notwithstanding the limitations of some of the workdescribed, this is a timely contribution to the field thatshould mark a turning point in the use of neural networksin animal behaviour research. It is likely to motivateanimal behaviour researchers to explore further neuralnetwork applications in their own research. Importantly,it is also likely to generate interest among those wellversed in neural network methodology, presenting a newforum for them to apply their knowledge and skills.

RICHARD PETERSCentre for Visual Sciences,Research School of Biological Sciences,Australian National University,Canberra ACT 0200, Australia