A famous and influential collection of essays on the developmental plasticity of behaviour entitled ‘Constraints on Learning’ (Hinde, R.A. & Stevenson-Hinde, S. 1973. London; Academic Press) forcefully drew attention to the many departures and exceptions to classical psychological learning theory that had accumulated in the literature of the time. But in the light of subsequent developments, it can be argued that the term ‘constraints’ itself has connotations that may have limited further progress toward gaining a fuller understanding not just of learning, but of developmental plasticity in general. Either explicitly or implicitly it embodied a widespread view that, if it were not for constraints, learning would have a virtually unlimited potential for creating behavioural change, in any direction and to any degree one might imagine. The constraints presumably envisioned were imposed on a general ability to modify behaviour.

An alternative to the concept of constraints on general purpose developmental programs is the view that the functional requirements for developmental plasticity in different behavioural contexts vary so radically that selection has produced not one but a whole range of underlying developmental mechanisms. Thus, rather than a single variously-constrained mechanism of behavioural plasticity, we can visualise multiple mechanisms, each specialised for a particular set of tasks, and restricted to a particular behavioural context.

One consequence of this alternative view is that it adds further emphasis to the importance of genetic contributions to behavioural plasticity, because if there are multiple mechanisms, they require us to postulate distinct physiological and neural machinery to sustain each of them. But we must proceed carefully. The suggestion that genetic mechanisms are more involved in developmental plasticity than has been assumed may seem paradoxical, if we believe that genetic control implies stereotyped patterns of development and predetermined developmental endpoints. We will present a case that this is an outdated view. A genotype encodes instructions for sequences of organism/environment interactions designed to develop not just one phenotype, but a range of alternative, environmentally-cued phenotypes. This is the concept of phenotypic plasticity (West-Eberhard, M.J. 1989. Annual Review of Ecology and Systematics 20: 249-278; West-Eberhard 1998; Schlichting, C.D. & Pigliucci, M. 1998. Sunderland MA; Sinauer Associates). When we are dealing with complex behavioural traits, the number of alternative phenotypes to which a given genotype may give rise can be large, each with its own distinctive experience-related ontogenetic trajectory. Phenotypic plasticity is a fundamental attribute of behavioural systems, and its study may be especially valuable in resolving some of the controversies about the role of genetic mechanisms in behavioural development. We often regard the alternative phenotypes to which a given genotype gives rise as only reflecting the environmental component of the ontogenetic equation, and not the genetic component. But the way in which environmental variation causes changes in the course of development reflects genetic contributions as well. Changes in patterns of gene expression ultimately underlie the adaptive responses involved in developmental plasticity and must be understood if we are to comprehend how environmental events influence behavioural development. Some environmentally-induced developmental changes are selected against, and others are adaptively neutral. Many clearly increase fitness, however, including the consequences of learning and those that induce change in neuronal development and activity, physiological state, hormonal status, and patterns of growth.

One corrective to the nature versus nurture misconception has been the idea of modularity in the mechanisms that make up perception and learning, usually accompanied by distinct neuroanatomical substrates and domains of behavioural influence. Examples of such specialised multiple systems in avian memory are described in the chapter in this Symposium by David Sherry. Research on song learning, described in the chapter by Peter Marler and in several contributions to the symposium on The Neurobiology of Vocal Learning in Birds (Nottebohm & Konishi, this volume), has shown that the introduction into a phylogenetic line of novel developmental programs - programs that involve learning - requires new patterns of neural organisation, dependent in turn on changes in the genetic instructions for developing a brain. As Ken and Mary Able show in their chapter, there are many remarkable parallels in the development of navigation by birds, as pre-existing sensory predispositions guide them in exquisitely sensitive reactions to environmental cues that serve to refine and calibrate the ability to navigate accurately over extraordinary distances. The apparently paradoxical concept of ‘instincts to learn’ in fact captures the reality of many naturally-occurring learning processes in avian navigation.

We do not usually think of the developmental plasticity that circadian and circannual rhythms display as cases of learning in the normal sense. In fact, there are many parallels, as Ebo Gwinner and John Wingfield demonstrate in their chapters in this Symposium, providing us with paradigmatic cases of phenotypic plasticity. Within a framework furnished by basic endogenous rhythms, cues from the environment experienced at particular phases of an endogenous circadian or circannual cycle fine-tune the temporal patterns of activity and sleep, reproduction, migration and molt. They achieve this by eliciting highly organised reactions, engaging genetically prescribed patterns of hormonal activity, neural activity, and growth.