This issue of the journal is focused on ‘top-down (downward) causation'. The words in this title, however, already raise or beg many questions. Causation can be of many kinds. They form our ways of ordering our scientific understanding of the world, all the way from the reductive concept of cause as elementary objects exerting forces on each other, through to the more holistic concept of attractors towards which whole systems move, and to adaptive selection taking place in the context of an ecosystem. As for ‘top’ and ‘down’, in the present scientific context, these are clearly metaphorical, as some of the articles in this issue of the journal make clear. Do we therefore know what we are talking about? The meeting at the Royal Society on which this set of papers is based included philosophers as well as scientists, and some of those (Jeremy Butterfield, Barry Loewer, Alan Love, Samir Okasha and Eric Scerri) have contributed articles to this issue. We would like also to thank those (Claus Kiefer, Peter Menzies, Jerome Feldman and David Papineau) who contributed only to the discussion meeting. Their contributions were also valuable, both at the meeting and by influencing the articles that have been written by others. We include a glossary with this introduction, composed by one of us (O'Connor). The clarification of the use of words and their semantic frames is an important role of philosophy, and this was evident in the discussions at the meeting and is now evident in many of the articles published here. Moreover, philosophical analysis is not limited to the papers by the professional philosophers. The idea of top-down causation is intimately related to concepts of emergence; indeed, it is a key factor in strong theories of emergence.
1. Does emergence occur? is it real, or just an epiphenomenon?
There are differing views in science, as well as in philosophy, about the reality of top-down causation. The aim of these articles is to provide answers convincing to both scientists and philosophers. The need is for an interdisciplinary dialogue, which we hope this Theme Issue provides. The issue arises in physics (Bishop ; Loewer ), chemistry (Scerri ), microbiology (Jaeger & Calkins ), epigenetics (Davies ), evolutionary biology (Okasha ), physiology (Noble ), neuroscience/psychology/cognitive science (Berntson et al. ; Atmanspacher ), social science (Elder-Vass ) and computer science (Booch ). Many examples provide good evidence for top-down causation. Thus, in physiology, it is taken for granted, and it is crucial in cognitive science. Getting clearer on the nature and varieties of causation is a key element in assessing such evidence (Butterfield ; Love).
2. Can it be an integrating theme within and across sciences?
This is a question that arises, on the one hand, within sciences, e.g. particle physics to nuclear physics, nuclear physics to atomic physics and so on, and across sciences on the other: e.g. physics to chemistry, chemistry to microbiology, psychology to sociology and so on. This is where the question arises whether there exists more than one kind of top-down causation (Ellis ). If so, the issue is not to show that one specific class of top-down causation is ubiquitous, but to consider whether it (as well as bottom-up causation) is ubiquitous in the real world, but in a number of different forms or kinds.
3. A ‘roadmap’ for further research in this area
The articles published here show that there is solid evidence that top-down causation does indeed occur in the domains of all the sciences, ranging from physics [1,2] to microbiology [4,5] to physiology  to sociology [8,10], in each of which the concept is deployed with considerable explanatory power. However, its status remains the subject of some controversy, as is shown for example by Scerri's paper on differing views in the context of chemistry. The idea of top-down causation is intimately tied in to the concept of the contextual emergence of complexity [1,9] and associated discussions of reductionism, with the claim being made by one of us  that true complexity cannot emerge in a bottom-up way alone. This is supported particularly by the way top-down effects can be seen at work in physiology  and in the functioning of the brain  and society . However, causation is multifaceted and complex . Like bottom-up causation, top-down causation should be seen as an interlevel phenomenon between neighzbouring levels in the hierarchy. Just as bottom-up causation does not (clearly) imply the existence of a clearly identifiable bottom level, top-down causation does not necessarily imply existence of a clearly identifiable topmost level .
Further work to take the discussion forward is needed both at a philosophical level, engaging with the general concept of top-down causation and its generic character, and the way it is realized in specific sciences. The papers by Butterfield, Bishop, Loewer and Love provide a good basis for taking the philosophical aspects of the discussion further, particularly through looking at the way higher and lower level actions mesh , and developing further the relation between existence of equivalence classes [4,14] and top-down causal effects. Their existence is an outcome of multiple realizability, which underlies the concept of entropy of macro states: that relation needs further development in generic cases (it is of course extremely well developed in the case of statistical physics). Particularly crucial is the issue of when higher level variables can and cannot be obtained by coarse graining of lower level variables : development of that discussion in different contexts, ranging from physics to social sciences, will be very useful. An issue that is not tackled in the present series of papers is how to identify levels of causation and emergence in truly complex structures such as interaction networks in general and the brain in particular; there is an ongoing technical discussion in this area. The further aspect that obviously needs much further discussion is the relation between top-down causation, contextual emergence of complexity and reductionism [8,9]. In philosophical discussion, this relation is considered in the context of the sometimes-supposed ‘causal completeness’ of fundamental physics [15,16].
Then there is a need to take the discussion further in each specific area, chemistry  and evolutionary theory  being particularly controversial examples where more clarity is needed (indeed even the concept of the chemical bond is in question: see Ball ). The idea of top-down causation is prevalent in almost all holistic discussions of brain function  and social interactions [8,10], but perhaps not often identified as such; making that identification specific will be useful. Important here is determining what different kinds of top-down causation should be recognized [13,14]; this needs further discussion.
Two particular issues are probably key in making the idea more acceptable in the wider scientific community. Firstly, while it is generically taken for granted at higher levels such as physiology, it has made very little headway particularly in the physics community, where ideas of bottom-up causation are prevalent because of a very strong sense of physical determinism, and in sections of the molecular biology and neuroscience community, where reductionist mechanistic explanations are dominant . They are prevalent precisely because they are so successful; but the contention can be made nevertheless that they are only part of the causal nexus at work (that is the burden of many of the papers here). To make progress here, a key need is to show that top-down causation is also prevalent in physics itself. Laughlin  made this case strongly some time ago, but his paper made little impact. Bishop's paper is a welcome contribution in this regard; recent papers by one of us strongly make this case as regards quantum physics  and the arrow of time —both crucial foundational aspects of physics.
Secondly, the papers in this Theme Issue largely are explanatory of known aspects of science—that is, they are post hoc explanations. What is crucially needed is predictions of new tests that will confirm the hypothesis that top-down causation is real, and not just an epiphenomenon. The papers by Jaeger (microbiology) and Noble (physiology) are strong steps forward in this regard. This is perhaps the area where most needs to be done, across the board in all domains. Areas where striking progress is being made in this regard are epigenetics  and social neuroscience . These papers are in fact dealing with top-down causation: the way they provide experimental confirmation of this concept needs to be made more explicit.
- a collection of individuals in which the intrinsic properties of the collection and their joint results are strictly deducible, in principle, from the properties of, and spatio-temporal relations among, the composing individuals. (Contrast with system.)
- Causation As production
- a productive cause of an effect is the set of factors that produce, bring about or make happen the effect. It is controversial whether causation in this basic, metaphysical sense ever obtains outside the realm of fundamental physics. (Causal closure-denying emergentism maintains that it does so obtain.) And on some philosophical views, there are no productive causes in this sense, only patterns of counterfactual dependence among types of events.
- As counterfactual dependence
- B counterfactually depends on A just in case B would not have occurred had A not occurred. Patterns of covariant dependence clearly obtain at many levels of nature, not just in physics. Plausibly, it is this minimalist notion of causation that guides methods for drawing causal conclusions from statistical data.
- higher level or systemic features that shape dynamical processes at lower levels. One variety of top-down causation involves constraints that preclude many forms of physical possibility present in unorganized matter, possibilities that would lead to systemic dissolution. A stronger, more controversial variety would involve higher level features exerting an irreducible, productive causal influence upon lower level processes. (See emergence: productive causal.)
- Emergence Predictive
- systemic features of complex systems that could not be predicted (practically speaking; or for any finite knower; or for even an ideal knower) from the standpoint of a pre-emergent stage, despite a thorough knowledge of the features of, and laws governing, the systems' parts. (Diachronic.)
- Explanatorily irreducible
- systemic properties and dynamical patterns of complex systems that cannot be derived from laws governing more basic systems. (Synchronic.)
- Productive causal
- systemic properties that exert a non-redundant, productive causal influence on the behaviour of the system's more fundamental parts. Implies that fundamental physics is not ‘causally closed’ in the sense of there being, for any fundamental physical state, a complete set of determinants that is entirely composed of same-level fundamental physical states. (See causation: top-down.)
- the thesis that certain high-level properties (e.g. of mental states) exert no causal influence over other states. These properties would have causes but would not be causes in turn.
- the property of many mental states of being directed at, or about, entities of extra-mental phenomena. This is closely related to the notion of representation.
- Multiple realizability
- systemic properties that can be realized by a diverse range of specific arrangements of lower level entities.
- the thesis according to which all higher level properties are realized by arrangements of lower level properties. On this view, there are no metaphysically irreducible properties of conscious or other mental states.
- the intrinsic qualities of conscious experiences of which a subject is or can be non-inferentially aware (e.g. the varying ‘looks’ of the same blue surface under different lighting conditions, distance from the observer, etc.).
- the antithesis of emergentism. Hence, it comes in at least three varieties:
- systemic features of every kind of complex system could be predicted (at least for an ideal knower, not subject to computational and other limitations) prior to the appearance of such systems, based solely upon comprehensive information about the world's most fundamental constituents and their dynamics.
- systemic properties and dynamical patterns of complex systems exhibited over time interval t1–t2 could be derived, in principle, from information concerning the systems' components and their arrangement over t1–t2, together with the laws governing their evolution.
- Productive causal
- systemic properties never exert a non-redundant, productive causal influence on the behaviour of the system's more fundamental parts. Fundamental physics is ‘causally closed’ in the sense of there being, for any fundamental physical state, a complete set of determinants that is entirely composed of same-level fundamental physical states.
- a collection of individuals in which some of the intrinsic properties of the collection and their joint results are not strictly deducible, in principle, from the properties of, and spatio-temporal relations among, the composing individuals. Some feature or features of the interrelations among the parts give rise to novel systemic properties and forms of causal efficacy. (Contrast with aggregate.)
- Self-organizing system
- a system in which there is no central locus of information or control, despite the illusion of goal-directed coordination in the behaviour of the parts. (An example is an ant colony.)
- the thesis that every high-level property and phenomena is ‘fixed’ or strictly determined by the global distribution of low-level properties and relations. ‘No high-level difference without a low-level difference’.
One contribution of 15 to a Theme Issue ‘Top-down causation’.
- Received November 2, 2011.
- Accepted November 2, 2011.
- This journal is © 2011 The Royal Society