Exploring Levels of Explanation  Concepts 
Part I: Interactions Between Ontic and Epistemic Levels


 Psychology Department 
The King's College
 Edmonton, Alberta, Canada

From: Perspectives on Science and Christian Faith 41 (December 1989): 194-205.

This paper invites scholars interested in science-religion relationships to examine the notion of "levels of explanation" in more depth than is usual. Interactions among epistemological and ontological aspects of levels have been explored in writings about "hierarchy theory." Examples of expanded levels ideas are considered in three areas: evolutionary biology, molecular biology, and epistemology. These examples suggest that connections between ontological and epistemological levels are important to explore when the scope of a scientific theory is broad enough to be reflexive, when one looks closely at a boundary between levels, or when one considers connections between scientific and religious knowledge.

In recent years the pages of this journal have carried several discussions highlighting some key concepts for understanding science-religion dialogue, including reflexivity, complementarity, and levels of explanation (e.g., Cramer, 1985; Haas, 1983; Van Leeuwen, 1983). While these concepts are closely interrelated, emphasizing one of them for discussion is much more manageable than trying to sort them all out at once. My aim in this paper is to support this clarification process by looking further at the notion of "levels of explanation." In particular, I examine literature from an interdisciplinary area of study sometimes referred to as "hierarchy theory" because it is one valuable resource for clarification.1

Levels of explanation and related notions often hold pivotal positions in discussions of religion and science relationships, particularly among scholars seeking to demonstrate the compatibility of the two domains (e.g., Barbour, 1966; Bube, 1971; Capra, 1982; MacKay, 1979; Peacocke, 1986; Polkinghorne, 1986). Despite the popularity of levels notions in the exploration of science-religion interfaces, there are many unanswered questions regarding the viability of current formulations (e.g., Cramer, 1985; cf. Orlebeke, 1977). In fact, there also are unresolved questions in the broader literature on "hierarchy theory," so the open-endedness of levels notions in religion-science discussions is not surprising (Pattee, 1973; cf. Grene, 1972; Salthe, 1985). Moreover, since concepts of a hierarchy of levels have a wide-ranging history in both Western and Eastern intellectual traditions (e.g., Capra, 1982; Leake, 1969), it is probably unrealistic to expect coherent, consistent use of these ideas across diverse areas of study. In this paper, I explore a selected set of questions in hierarchy theory to highlight potential directions for fruitful development of hierarchy models. A next step, applying these issues to science-religion dialogues, is the task of a companion paper.

Background Terminology

Given the diversity and poorly coordinated nature of literature on hierarchy theory and levels of explanation, terminology is an important preliminary consideration. Although the term "hierarchy" is often used to describe authority relations in institutions or relations among taxa in classification systems, the focus for this paper is the use of "hierarchy" to describe an ordered sequence of levels more generally. Different kinds of hierarchies can be distinguished from the outset.2 Levels of "explanation," "analysis," "abstraction," or "description" emphasize hierarchical strucuture as reflected in knowledge systems. Epistemic concerns (i.e., concerns about the nature of knowledge) are frequently emphasized by "levels of explanation" terminology. Sometimes levels of explanation are expressed as a form of the traditional "hierarchy of disciplines" which features physics at the base, biology in the middle, and social science toward the top (cf. Beckner, 1974). By contrast, one can distinguish levels of "being," "reality," or "organization" as designations for basic ontic (reality) structures, usually ¨emphasizing nested composition patterns of elementary entities or "things." The "levels of organization" frequently used as a framework for biology curricula is a classic illustration of compositional levels: cell, organism, population, etc. Each level of the hierarchy is defined by things which are composed of entities from the level below: for example, organisms are made up of cells. Finally, one can also identify levels of "activity" or "modal aspects" which reflect functions or processes instead of the entities in compositional hierarchies (e.g., Barbour, 1966; Hart, 1984). For example, some authors argue that mind-brain questions reflect different modes of functioning (physiochemical and cognitive) of a single entity (a person) rather than reflecting different entities (brains and minds).3 It is helpful to note that both compositional and functional hierarchies describe reality (ontic patterns), though functions and entities are not usually considered the same kind of reality. Differentiation among hierarchies can easily be continued beyond these types, but for the present purposes distinguishing epistemic, compositional, and functional hierarchies of levels will suffice.

A central reason for clarifying and elaborating our understanding of hierarchies of levels is to contribute to dialogue between religion and science. Eventually, this aim will require developing insights from hierarchy theory beyond current consensus in the field. As a beginning, however, it is important to learn what we can from what is currently being developed. My strategy for getting started is to focus on interconnections between epistemic and ontic hierarchies, and to illustrate what happens when one does this. Admittedly, the material is at times abstract, but then hierarchy theory is similar to the interdisciplinary nature of religion-science dialogue in that it requires most of us to be out of our depth most of the time. In reading philosophers of science and scientists working on theory construction, I find their work intertwines to a large degree with this area, so no attempt is made to systematically separate various disciplinary strands in the following discussion. My hope is that the continuity of topic justifies any blurring of disciplinary boundaries.

To focus the discussion of interconnections between ontic and epistemic hierarchies, I will make brief excursions into theoretical biology and philosophy of science. In the recent history of science, biology has provided the richest context for development of hierarchy theory. In particular, biological ecology,4 molecular biology, and evolutionary theory have used levels concepts in systematic theorizing. It comes as no surprise, then, that two of the three selections overviewed in this article focus on biological topics. First, selected issues in evolutionary biology provide an illustration of the breadth of scope possible within hierarchical models. A second point focuses on molecular biology as a case where hierarchy considerations arise at a boundary between levels. Finally, a view of epistemology is discussed to raise issues surrounding connections of hierarchical models with social and spiritual reality. Many aspects of the positions presented here are controversial and nothing is without challenge, especially within the disciplines of the authors. Nevertheless, the work considered below may be able to aid attempts to elaborate our understanding of ontic and epistemic levels.

Foundations of Evolutionary Theory5

One set of ontic-epistemic interactions among hierarchies arises from acknowledging the reflexivity of an observer generating a theory about an ontic hierarchy in which the observer "fits." For example, a biologist studying evolution can develop a hierarchical model in which humans (including theoretical biologists) fit at an organism level. The broad scope of evolutionary theory requires reflexive theory because the theory addresses realities (e.g., process rates and levels of organization-see below) which are part of the context for theory building by humans. Note that this self-reference involves both descriptive levels of theory construction and the compositional levels of organization. So reflexivity implies interconnections among ontic and epistemic hierarchies. Stanley Salthe explores several issues dealing with self-reference and hierarchy as background for his proposal of a framework for evolutionary biology.6 To illustrate perspectives offered by hierarchy theory, I will discuss two topics addressed by Salthe: (a) differences in average rate of processes between levels, and (b) the interdependence of research interests and levels of organization.

Differences in the average rates of key processes are often observed in entities at different levels.7 This pattern has consequences for an observer situated at a given level. Considering processes of growth and decay, for example, the lifespan of organisms is often in a range from one to a few generations, while the "lifespan" or duration of biological populations extends over many generations. More generally, from the perspective of an observer's level, processes at a preceding, micro, or "lower" level (e.g., an atom level preceding a cell level) are "seen" as constants. That is, micro-level processes generate constraints reflecting the cumulation of multiple cycles because the rates of preceding level processes are much faster than processes at a given level of observation. For example, the temperature of objects is not directly perceived by humans as vibratory patterns of atomic structure, but as a molar, semistable characteristic. Interestingly, interactions of an observer's level with subsequent, macro, or "higher" levels are also often perceived as constant constraints, in this case because the process rate is so slow relative to the rates of observation processes. Consider our perceptions of glacial activity; we cannot directly perceive glacial motion.8

Since observers are "embedded" in a specific level and average process rates differ between levels, one can see why entities at different levels (e.g., atoms and ecosystems) can be difficult to perceive when compared to observations of entities at the observer's level. For micro-levels, the "thing-ness" of small entities is lost in the fabric of events because micro-processes cumulate to act as constraints at the observer's level. For macro-levels, entity activities are seen as constant patterns of constraint characterizing the environment because the rates of change are slow relative to observation level processes. Since the co-occurrence of multiple processes is a major principle for identification of entities, rate differentials render the identification of entities at other levels more difficult by inhibiting the perception of processes. Using the terms presented above, descriptive levels (for identification of entities and processes) are intertwined with levels of organization (as reflected in the average rates of processes) in a manner which shows up particularly in instances of cross-level observation.9

Whatever the source of the change in perspective, 
a new entity inhabits discussion which 
cannot be simply set along side of familiar things.

A closer look at how explanatory and compositional levels intertwine here requires an examination of dynamics in hierarchical systems in which an observer is embedded. An illustration may help clarify what Salthe is suggesting. Consider the process of an observer overcoming the difficulties of perceiving some target entity which is at a level other than the observer's: say, an ecosystem. This learning process can be broken down into phases for the sake of discussion (without implying any rigid categories or temporal ordering). At first, the observer does not distinguish a target ecosystem from among the variety of environmental features and the fabric of events shaping the world. In this first phase, the observer and target interact in such a way that the target is not clearly perceived as an entity. A forest or riverbank is seen as a place, an undifferentiated aspect of the environment. This "observation" process reflects the average rate of processes characterizing both the observer level and the target level. One might well see "seasons" in the forest and adjust actions like tapping the maples for sap only in the spring. But one might not, for example, notice the patterns which distinguish seasons in the forest from the seasons of the prairie or mountain top. They are just different places, not systems which actively shape the weather and climate.

In a second phase, some change in the observation and perception processes results in a new identification of coherence. Perhaps stories across generations about a place, a given forest, start to come together in such a way as to encourage seeing the forest as a "living thing." Or perhaps a field biologist identifies patterns across a number of contemporary ponds which fit historic data for a given pond, suggesting that ponds have "lifespans," that they emerge, "grow, develop, and die." Whatever the source of the change in perspective, a new entity inhabits discussion which cannot be simply set along side of familiar things. A bear and a riverbank are simply not the same kind of things. One develops a new level of explanation-or fits this new entity somewhere in one's already well developed descriptive hierarchy.

A third phase emerges when the modified levels of explanation function as a framework for action in addition to providing a perceptual frame. Perhaps the observer develops a new observation technology to take into account the new understanding of phenomena related to the target. Clearly one can identify rates of certain target processes without necessarily employing a sophisticated epistemic hierarchy as a framework. Knowing that spring is the best time to tap maples for sap does not require a modern ecosystems theory. Nevertheless, coordinating actions across mutiple domains or tracing complex interactions across wide systems often improve with the insights gained from more elaborate models (including hierarchy structures). The helpfulness of ecosystems concepts in tracing the deaths of maples in Quebec to acid rain serves as a case in point.

Finally, a fourth phase involves the impact of changed observer actions on physical reality. If a soil scientist advises a farmer, the resulting agricultural technology, embodying relevant levels of description, may significantly impact the interaction of the farmer with the ecosystems related to the farm. Or a scientist developing observational technology for marine ecosystems based on novel concepts may change the interactions of the observer with the target, an ontic outcome of epistemic processes. Fourth phase activity thus reflects an ontic change in the interlevel interaction between the observer and the target. If farmer-field ecosystem interaction changes as a result of the introduction of new herbicides or fertilizers, the ecosystem may simply adjust to the new relationship or it may change drastically if important stability mechanisms are disrupted. The main point here is that changes in levels of description may lead to ontic changes, perhaps even in levels of organization if an ecosystem transforms radically.10 Any change might, in turn, bring us back to Phase One if relevant entities are not perceived. Hierarchy models can help understand such processes as part of the dynamics of hierarchical systems.

To summarize the illustration, the transition from first to second phase described above reflects a familiar process to working scientists and "scientific realists." The intertwining of composition levels and description levels involves ontic patterns shaping epistemic patterns through observation. More simply, we "learn something about the world through observing." The later transition from phase three to phase four may be less intuitively familiar, however, in that we find our understanding actually forming reality.11 An observer's levels of explanation become a framework for action which results in ontic consequences, possibly shaping levels of organization.

I have described this point in some detail because it serves to illustrate reciprocal interactions between epistemic and ontic levels, a point easily passed over in summary. My emphasis is on the necessity of distinguishing between levels of explanation and levels of organization in order to formulate claims like Salthe's. To the degree that Salthe's assertions (or disagreements with him) make use of epistemic-ontic relations among levels, my aim of illustrating the value of the distinction between different kinds of hierarchies is supported.

Changes in levels of description may lead to ontic changes, 
perhaps even in levels of organization if an ecosystem transforms radically.

Observer-embeddedness in nature implies that the guiding interests of observation interact with the different kinds of entities to generate different levels of organization, each reflecting the same hierarchy of nature.

At this point in the discussion, one might be tempted to simply identify levels of explanation directly with corresponding levels of organization, given their reciprocal interdependence. But a second aspect of Salthe's consideration of observer reflexivity suggests that any such correspondence cannot be simple. Consider his example of a geologist and biologist formulating the levels of organization in nature (especially pp. 168f). One easily finds a correspondence between their hierarchies at the molecular level even though the molecules involved are quite different. When one examines the levels above that, however, the phenomena are not comparable. Icicles and turbulence patterns are composed of molecules, as are organisms, but even though we are studying the "same" level of nature in an important sense, we find they are not the same "level of organization."12 The problem shows up when one starts to connect geological and biological levels of organization. They should reflect the same ontic levels because nature is unitary, not divided up into separate geological and biological worlds. But if we collapse the hierarchies by saying that icicles and organisms are on the same level, we run into problems. Since both hierarchies are compositional, we say that biological populations are made of organisms. If icicles are at the same level as organisms in the same compositional hierarchy, however, then icicles are also components of biological populations.13 For Salthe, this kind of confusion is an error of logical type, an indication of absurdity in the same way that asking about the mass of one's emotions is absurd.

So what are we left with? Biologists and geologists have a few lower levels which correspond, but we cannot collapse levels of organization studied in the two disciplines into a single, linearly ordered hierarchy of nature, because we generate logical absurdities by doing so. This does not imply, in Salthe's view, that there are multiple natural worlds which cannot be combined. He asserts clearly his assumption of a unitary hierarchy of nature behind all the studies of various sciences. For Salthe, what this inability to combine different formulations of the same ontic hierarchy means is that "[l]evels of organization can be seen to be the consequence of the observer's being embedded in "the hierarchy of nature (p. 167). That is, the compositional hierarchies of biologists and geologists, each reflecting the single hierarchy of nature from a disciplinary perspective, do not coincide completely because different kinds of entities (rocks and organisms) exist in nature and researchers' interests address this complexity. Observer-embeddedness in nature implies that the guiding interests of observation interact with the different kinds of entities to generate different levels of organization, each reflecting the same hierarchy of nature. In short, the patterns of reality (levels of organization) not only reflect the unity of nature's hierarchy but also the characteristics of observers (researcher interests, in this instance) interacting with the larger ontic context.

Nothing in this discussion of genetic codes indicates 
a total uniqueness of the questions about function and description.

To summarize, Salthe argues that observer-embeddedness in nature's hierarchy yields an interdependence between epistemic and ontic hierarchies for cross-level observation. Difficulties arise in the identification of entities at lower and higher levels partly because of average rate differences in key processes at each level. Likewise, one encounters logical difficulties in combining compositional hierarchies from different sciences even though they reflect the same hierarchy of creation. Independent of one's final evaluation of Salthe's proposals as evolutionary biology, his work on hierarchy theory addresses complications following the acknowledgment of an observer's place in a hierarchy. His views illustrate the value of distinguishing compositional, functional, and epistemic levels. Such distinctions become necessary for Salthe when formulating foundations for evolutionary theory because the scope of the theory is all life on earth, yielding self-reference. Evidently, the complexity which attends reflexivity can be managed to some degree with the tools of hierarchy theory.

Molecular Biology and Genetic Coding

Another important context for exploring relations between epistemic and ontic hierarchies has received considerable attention: molecular biology and study of the "genetic code" (e.g., Pattee, 1971, 1973, 1979). The amount of attention paid to this area generates the advantage of widespread familiarity with major results, but the attention does not necessarily entail an appreciation of foundational issues in the field. Along with other scholars, Pattee (1979) has been attempting to stimulate the molecular biology community to greater concern for fundamental elements of their work. Specifically, the empirical successes of molecular biology have, according to Pattee, failed to stimulate a resolution of certain conceptual debates in the literature which were present prior to elucidation of mechanisms. Instead, interest seems to have waned and attention is now focused on the "business as usual" of building on key discoveries while some conceptual complexities lie unexplored. For this paper, the issue of interest is that Pattee employs hierarchical models to address the conceptual problems to which he points.

Pattee suggests that we first of all need to generalize the notions of language and description in order to examine interconnections between ontic levels and descriptive levels in molecular biology. After discussing the relation between the chemistry of DNA and life, he states in summary:

Life is distinguished from inanimate matter by the co-ordination of its constraints. The fundamental function of this co-ordination is to allow alternative descriptions to be translated into alternative actions. The basic example of this function is the co-ordinated set of macromolecules which executes the genetic coding. It is useful to think of such co-ordinated constraints as generalized language structures that classify the detailed dynamical processes at one level of organization according to their importance for function at a higher level. In this sense, co-ordinated constraints, language structures, alternative descriptions and hierarchical controls are inseparably related concepts. (1971, p. 171)

Since the sequences of bases are physico-chemically "indeterminate," DNA strands and associated macromolecules bear information ("alternative descriptions"), thus permitting genetic patterns to serve as constraints (designs) for cellular functioning.14 The distinction here between physico-chemical processes and information is central to his claim. As Pattee states, "the relation between the structural [in my terms, compositional] and descriptive levels is the central problem that must be solved to have a theory of hierarchical control" (1973, p. 136, emphasis deleted).

Summarizing Pattee's views in the vocabulary employed here, the topic of interest is the interface between molecular and cellular levels of a biological composition hierarchy. Conceptualizing genetic codes and information as "generalized language structures" or descriptions ties in epistemic hierarchies at this same boundary between levels. That is, he presents biological information as one level of an epistemic (informational) hierarchy.15 In my elaboration of Pattee's discussion, hereditary mechanisms are examples of compositional and epistemic hierarchies entwining at an interlevel interface. So his critique of molecular biology clearly distinguishes ontic and informational levels while confronting basic conceptual issues. Without presenting a full analysis of genetic codes, one can see that Pattee's foundational questions in molecular biology are more easily formulated once one has distinguished ontic and epistemic levels. Clarifying, extending, or critiquing Pattee's analysis requires even further elaboration of levels notions.

The purpose of hierarchical models is to disclose some 
of the intricate pattern of epistemic-ontic interaction in a "participative ontology."

One can extract several lessons from Pattee's discussion of molecular genetics.16 First, the empirical grounding of the area helps focus important questions, but data alone do not resolve questions about hierarchical interrelations. Pattee's concern about scientists focusing on mechanisms to the exclusion of conceptual fundamentals makes this point. Although empirical and philosophical questions are intimately intertwined in this context, it does not seem that we have a case of "naturalization" of fundamental issues in genetics, popularized statements to the contrary notwithstanding.17 Thus, simply going back to the lab will not make these questions go away. Second, I suggest that formulations of these issues via hierarchical models is more fruitful than attempts to avoid levels concepts. By restricting attention to "information," for example, instead of dealing with Pattee's notion of "generalized language structure" or informational levels, one might claim that hierarchy is irrelevant to genetic codes or biological information. However, such a move merely buries the same issues in the definition of "information." We can still ask about the relationships between genetic and linguistic information, for example, by suggesting that sorting out that relationship is important for adequate definitions. As a final lesson, I suggest that nothing in this discussion of genetic codes indicates a total uniqueness of the questions about function and description. One might anticipate that elaborations of hierarchical models in molecular genetics could be generalized in some form to apply to other domains of inquiry. For example, there may be fruitful parallels between the interlevel interface of genetic codes and traditional condundrums surrounding mind-brain or individual-society relations.18 In short, Pattee's work illustrates contributions of hierarchy models at a specific interlevel interface.

Philosophy of Science & Epistemology

Salthe and Pattee's work on hierarchy theory addresses epistemology as necessary for advancing our understanding of levels. Arbib and Hesse (1986), on the other hand, find levels notions necessary for their project of developing an epistemology. Their contributions to the present exploration of ontic-epistemic relations in hierarchy theory center on questions about the ontic status of epistemic hierarchies. Also, they deal explicitly with knowledge of spiritual domains. So a brief excursion into their work can illustrate another way hierarchy theory might benefit dialogue between scientists and religionists.

Their project extends schema theory from roots in cognitive science and Piaget to the development of an epistemology which can address both scientific and religious knowledge.19 Moreover, in their elaboration of schema theory Arbib and Hesse maintain a consistent awareness of hierarchical patterns (although they do not present a systematic hierarchy theory).

We need a multilevel description of the human being. Schema theory, as it develops, is to provide an ever more appropriate mental vocabulary, while neural processes provide the mechanism for schema storage and dynamics. This is both more and less than reductionism. In some sense, everything in human behavior or society is mediated by neural firing and other physicochemical processes. And yet, there is no useful sense in which our analysis of human beings can be conducted exclusively at that level. We have many different levels of description, including neural, mental, and social, and we find ways of illuminating any particular level of discourse by placing it within a higher level context and by seeking lower level mechanisms. In this way, we see how to think coherently of the neural and the social levels as placing constraints on the schema level of analysis without claiming that any level is the one true level at which we should conduct all discussion. (pp. 14f)

Although their phrasing in this passage might lend itself to an interpretation restricted to epistemic levels, Arbib and Hesse are clear that patterns of reality are also at stake: "we in fact advocate a permissiveness with respect to ontology: there are all manner of levels of reality."20

Within the background of these general comments, Arbib and Hesse address the ontic status of epistemic hierarchies: In what sense is language or scientific theory real? They start by accepting spatio-temporal reality as unproblematic in that members of our culture readily acknowledge that the physical world is "real." On the other hand, the reality status of values, rational truths, symbol systems, ideologies, and God is questioned (pp. 2-5; cf., pp. 58-62, 84). (Interestingly in this context, questions regarding the ontic status of logic are explicity avoided, pp. 59f.) Nevertheless, in their view knowledge systems are real at least in the sense that our theories, for example, impact physical reality through human action. For these authors, the clearest case of this process is technology: scientific knowledge clearly impacts the physical world through application in technology, and knowledge has to be real in some sense to influence spatio-temporal reality.21 In short, the interaction of epistemic levels with physical reality (via technology) demonstrates something of the ontic status of epistemic phenomena (e.g., scientific theory).

For the present discussion, my main interest is their broader principles for studying the ways in which knowledge systems are "real." First, they assert that hierarchical patterns of description are necessary to elaborate an epistemology adequate to deal with both scientific and religious knowledge. Then, while assuming unspecified hierarchical patterns of reality, they also note that (a) language, scientific knowledge, and other knowledge systems (which have hierarchic form themselves) are real in a different manner than is physical reality, but that (b) knowledge systems can change physical reality (with technology being the main example of how this happens). Thus, a major principle in Arbib and Hesse's approach is that when epistemic systems impact physical reality, the impact gives us important clues about the ontic status (type of reality) of the epistemic systems.22

When examining the reality of spiritual knowledge systems, 
we can look, in various senses, 
to the consequences of spiritual knowledge for 
spiritual (and physical and social and cognitive) reality.

Arbib and Hesse's argument presents a clear analogy to Pattee's claim that the functional consequences of genetic information in the physical system of a cell reflect the impact of a "generalized language structure." In both cases, information (in genes or theory) strongly influences physical reality (via organism functioning or human technology) through known mechanisms (cellular decoding systems and technology). Similarly, Salthe's views on observer reflexivity portray observation processes as inextricably intertwined with levels of organization, observer actions, and observer interests. These facets of observer epistemology actually shape reality. In this sense, descriptive levels are not simply shortcomings of human capacity which limit the extent of human knowledge.23 Despite widely diverse foci of discussion, these authors all suggest a "participative ontology" of some kind where epistemic issues are intimately intertwined with ontic processes. In this kind of "constructivist" view, reality independent of an observer is, in principle, not identical to reality including observers.24 The purpose of hierarchical models in this context is to disclose some of the intricate pattern of epistemic-ontic interaction in a "participative ontology." Once again, whether or not we want to accept a constructivist viewpoint, I want to point out the value of levels notions in explicating what the concepts mean. In fact, it seems to me that hierarchical models are necessary either to support or to refute constructivist theories like those examined here.

One aspect of Arbib and Hesse's work covers topics not addressed by the authors focusing on biological topics: the interconnections of social and spiritual realities with a hierarchy of nature. The full range of biological, mental, and social levels are considered minimal context for an epistemic framework adequate for addressing spiritual reality. How, then, can we apply their principle for the "reality" of epistemic systems noted above? When examining the reality of spiritual knowledge systems, we can look, in various senses, to the consequences of spiritual knowledge for spiritual (and physical and social and cognitive) reality.22 Consider also how the broad scope of their model raises questions about reflexive features of Arbib and Hesse's program (paralleling Salthe's model of observation). Once one addresses spiritual activity in addition to cognitive and social activity, the question comes up regarding the relations of spiritual knowledge to the scholarly enterprise. Unless one posits a complete autonomy of scholarship from spirituality, another loop of self-reference arises when considering an epistemology of spiritual reality.25 In light of these points and those in previous paragraphs, it seems clear to me that Arbib and Hesse's levels conceptualizations will continue to distinguish and interrelate epistemic and ontic hierarchies (as well as challenging us with important issues). Once again, the distinctions highlighted here can prove beneficial for topics related to science-religion dialogue.

In our age of technological science, 
we are well accustomed to a variety of tools which extend the ranges of our senses...
there is no debate about the "reality" of invisible wavelengths of electromagnetic radiation.

In short, Arbib and Hesse do not present a systematic hierarchy theory, but their theory of knowledge does depend explicitly on levels ideas to conceptually ground their approach. This epistemology project stimulates important questions by going beyond scientific knowledge to include other forms of social and religious knowledge. Overall, careful examination of Arbib and Hesse's work hints at an intricate interweaving of hierarchy notions with the complexities of an epistemology of spiritual and scientific realities.


I started this paper by pointing out that "levels of explanation" and related ideas play important roles in many discussions of relationships between science and religion. That importance alone justifies attempts to work out more detailed understanding of these notions. By reviewing selections from literature on hierarchy theory and epistemology, I attempted to show that there are available resources available for clarifying levels concepts. One basic contribution was offered in the distinction between epistemic, compositional, and functional levels. Complexities associated with observer reflexivity, interlevel boundaries, and spiritual knowledge were clarified by distinguishing these different kinds of hierarchies. Rather than providing complete resolutions to the complexities encountered, this paper invites scholars interested in science-religion dialogue to continue the elaboration of available hierarchical models. I believe the work required will continue to be rewarded, and that available literature demonstrates the promise of levels notions for further grounding of religion and science dialogue.

The literature on hierarchy theory is certainly not the only resource we can draw upon in expanding our understanding of levels concepts. Additional resources in philosophy and systems theory, for example, are also well worth exploring. This paper does not attempt to review all valuable literature. There is, however, at least one major point which has not received adequate attention. I claim that elaborating our levels concepts will contribute to the dialogue between science and religion. With the introduction provided in this paper, I can turn more directly to the task of further illustrating these contributions in "Levels of Explanation: Part II.(1991)"



This paper is a revision of a portion of a paper presented at the 1988 ASA annual convention in Malibu, California. I want to acknowledge the contributions of colleagues to this paper. Vaden House provided critique and conversation which significantly strengthened both my thinking and my writing. I can no longer distinguish at this point between lessons learned in conversation with him from learning gained elsewhere. For repeated encouragement and material support, I thank Harry Cook and also Hank Bestman. Comments by several people at the 1988 conference and by several reviewers were stimulating as well as encouraging.


1The designation "hierarchy theory" has not acheived consensual acceptance, but it is a convenient designation for a multidisciplinary concerns tied together by questions involving hierarchies of levels (e.g., Pattee, 1973). For definitional issues, see note 2.

 There are many other resources besides levels ideas to draw upon in sorting out complex systems. For a popularized overview of various facets of complexity in science, see Davies, 1988.

2See, for example, Ayala, 1974; Barbour, 1966; Bunge, 1956; Grene, 1967; Pattee, 1973. No complete definition of levels or hierarchy is attempted because there is no widely accepted formulation to date despite substantive work in the area (e.g., Bunge, 1977). Generally, definitions of levels involve specifying units of analysis which characterize each level and definitions of hierarchies require an ordering among levels. For the present discussion, the common practice of accepting implicit definitions by use of example is adopted rather than the common alternative of stating precise but inadequate working definitions. Substantiating selected definitions is a worthwhile project which is beyond the scope of this paper.

The epistemic-ontic distinction highlighted here does not imply that one can somehow separate or isolate reality from knowing. Rather, one inherently refers to inseparable aspects of complex processes. Likewise, inseparability does not imply that the distinction is meaningless.

3See, for example, MacKay, 1982. I am not claiming here that MacKay distinguishes between ontic and epistemic levels. The nature of functions in comparison to entities is the question at issue, and some thinkers want to be able to assert the reality of functions as different from the reality of entities (e.g., Barbour, 1966; Hart, 1984). Making the distinction does not beg the question, but it does aid in formulating both the question and the associated debate.

4See, for example, Allen & Starr, 1982; O'Neill, DeAngelis, Waide, & Allen, 1986; and Webster, 1979.

5For this paper, I address evolutionary theory per se and do not consider evolutionism; i.e., issues following from such theory used as a basis for a myth of origins. Levels notions do not arbitrate directly between viewpoints on creation and evolution. Salthe's work is evolutionary while van der Meer (1989) is developing a hierarchical creationist model. It seems to me that levels notions are fundamental enough to fit with a variety of world views (though not all).

6Salthe, 1985. Although Salthe's approach is explicitly non-reductionist in some senses, it is not merely biologists who prefer "reductionist" approaches who might disagree with Salthe's proposals (cf., Williams, 1985). Marjorie Grene (1972), for example, argues for hierarchical models but takes issue with the general form of theoretical synthesis of evolution rooted in the hierarchies of a systems theory. My point in examining Salthe's work is to explore his extension of hierarchy theory to the interactions of epistemic and ontic hierarchies. Debating specific positions on reduction-emergence, evolutionary biology, or the nature of genetic information are related to these aims, but these debates also beyond the scope of this paper.

It is also helpful to point out that Salthe's attempts to deal with reflexivity do not imply a disdain for non-reflexive theorizing (see especially chap. 6). As I read him, he sees his effort as a generalization attempt which addresses a complexity that was ignored for the sake of (necessary) simplification in the early stages of theory construction. In that sense, perhaps, the development of reflexive evolutionary theory can be seen as analogous to mechanics which takes friction into account. For more general background on reflexivity issues, see Bartlett & Suber (1987).

7It is helpful to note that rate differentials do not need to be exhaustive for his point to carry weight, they simply need to dominate the major processes of respective levels. See Allen & Starr (1982) for a discussion of similar points in biological ecology theory. In discussing Salthe's views, the distinction between entities and processes need not be strong because he sees the two languages as interchangeable to a large degree.

Discussions of process rates are pursued by Salthe to address the more central notion of transitivity of effects across multiple levels in a hierarchy. His broader concerns, although of interest, are not addressed here due to space limitations.

8I emphasize direct perception to make the point about rate differentials, not to deny that technological developments expand the range of human observation. In fact, the technology of observation is of interest in the present context and is mentioned below.

9It bears repeating that Salthe's point emphasizes the average rate differential between levels. Thus instances of direct interaction between entities of widely divergent levels and process rates are not counter-instances. The sun interacts directly with many organisms in photosynthesis and sunburn, but we still perceive our star as a constant part of the environment and its light as a constant flux of energy instead of as a waveform or a stream of particles (despite the theoretical perspectives gained by twentieth-century physics).

Also, there is nothing especially significant solely in the fact that rate differentials render the perception of entities and processes at different levels more difficult. In our age of technological science, we are well accustomed to a variety of tools which extend the ranges of our senses. A simple case is illustrated by the fact that, at this stage of scientific understanding, there is no debate about the "reality" of invisible wavelengths of electromagnetic radiation. Salthe's emphasis on the interaction of epistemic and ontic levels in rate differentials is better illustrated by debates about theoretical notions which are sometimes interpreted instrumentally and sometimes realistically. For example, the kinds of reality represented by species or ecosystems is not currently a matter of consensus among scholars. In debates about whether certain scientific constructs are about "real" objects or processes, reflexive interactions of human observers and other levels involve epistemic hierarchies and compositional hierarchies in mutual influence. In debates of this kind, definitional and methodological issues highlight an interdependence between epistemic and ontic hierarchies. Thus, as I read Salthe, his views amount to claiming that when we debate basic definitions, levels of explanation and levels of organization are equally basic in science.

This point can be illustrated by two questions, the first emphasizing epistemology and the second ontology, but each one depending equally on the other. (1) What perspectives are appropriate to study X (since Xs are this kind of thing)? (2) What kind of thing is X really, given our data and observations of it and related phenomena? In other words, once we are clear on the kind of reality a certain phenomenon reflects, we know best how to study it, and when we are clear on the ways to study a phenomenon, we can best determine what kind of reality the phenomenon reflects. This interdependence is particularly clear when methodological and definitional debates surface.

Just as self-proclaimed reductionists or materialists 
employ hierarchy theory as effectively as do "emergentists"
...so the use of hierarchical models will probably not, 
in and of itself, prejudge conclusions about other, related questions.

10Salthe discusses the emergence of new levels between old ones and higher levels above old ones. Pursuing those concerns would take this discussion too far afield, however interesting and important it is.

11See Arbib and Hesse, 1986, on technology as one mechanism for knowledge generating ontic consequences. Their views on this point are briefly noted below.

Most coupling processes do not reflect solely the differential rates of processes, though in Salthe's work rates are emphasized. Coupling and decoupling processes involve many facets. For the present discussion process rates serve well as a feature of activity in our world which concretizes some aspects of interlevel interactions.

12In Salthe's terms, these discussions are about a "rank" of nature as a class, and about the relations between levels of generality, functional levels, and cognitive processes like labeling and attention. I have simplified the discussion here because of space constraints. For more detail, see chapter 6 of Salthe in particular.

13This specific example can be addressed by changing the structure of hierarchies from simple linear orderings of levels to partial orderings of levels, yielding a "branched" hierarchy with a common trunk and geological and biological "forks" (cf., Salthe chaps. 3 & 6). This form of complexity fits with Salthe's general thesis, but it does not address other cases where there are no levels in common between two related hierarchies. The more complicated examples were not used in this discussion due to space limitations. For more detail on these issues in Salthe, see his discussions of relationships between geneological and ecological hierarchies, especially his chapter 7.

14Even more broadly, the set of related notions necessary for coherent and comprehensive study of genetic codes includes information, function, meaning, and structure (cf. Grene, 1967; MacKay, 1969; Pattee, 1979).

15In his 1979 paper, Pattee elaborates his views on relations between biological information, structure, and language, specifically arguing for complementary relations reflecting (generalized) measurement processes. The various complementarities are not synthesized into an overarching hierarchy as far as I can tell. He does, however, emphasize the epistemological nature of information-structure complementarity and he does distinguish, in passing, levels of complexity of information (e.g., pp. 218f). While his work in these and other papers provides a large amount of material to help ground and develop hierarchy models, I avoid greater detail here to conserve space. For discussion of the related complex of issues, see Bennett's comment on Pattee's 1979 paper, Pattee, 1978, and MacKay, 1969. Consideration of the relation between information and biological function is a further direction for elaborating here.

16Of course, many important questions are left unmentioned here. Many readers of this journal are probably aware, for example, that sources of information in genetic systems were explored in a recent conference sponsored by ASA's Committee for Integrity in Science Education (see Walter Hearn's description in the August/September 1988 issue of the Newsletter of the ASA/CSCA, 30(4), pp. 1-2). Since the papers from this conference are not being circulated pending publication, we will have to wait to benefit from the struggles there for additional detail on relationships between information, function, and biochemistry. For the present discussion, however, it is worth noting that disagreements regarding "physical discontinuity" or "a seamless cause-and-effect continuum" (see Walt Hearn's description) can most likely be argued on both sides with the support of hierarchy theory. Just as self-proclaimed reductionists or materialists employ hierarchy theory as effectively as do "emergentists" (e.g., Bunge, 1977; Campbell, 1974; Glenn, 1988; Wimsatt, 1986), so the use of hierarchical models will probably not, in and of itself, prejudge conclusions about other related questions.

Instead, my expectation is that the continuity-discontinuity disagreement may in fact be made more constructive through clearer formulation of relevant hierarchies. My point about the fertility of hierarchy theory and the importance of informational hierarchies does not presume a resolution to continuity-discontinuity debates, though implications of my stance may the raise stakes of the discussion through a transformation of the questions. Better understanding of sources of information content in the genetic code cannot but benefit the discussion (by displacing controversy about speculative models, for example), but it will probably not resolve longstanding controversies.

17Naturalizing philosophical questions takes place when intellectual debate over questions is replaced by, or made moot by, empirical findings. From a naive point of view, an assumption of exhaustive naturalization of philosophy sometimes seems to be made by those who see philosophy as irrelevant to science.

18See Pattee, 1978; cf. Davies, 1988. The central strategy of hierarchy models is twofold: (a) to distinguish phenomena at different levels, and (b) to synthesize a set of such distinctions into a series of levels, a hierarchy. Both of these principles serve the central purpose of levels models: to adequately recognize and understand complexity. So the interest of Pattee and others in relating physical, biological, and social phenomena in hierarchy models clearly reflects their understanding of the cosmos. It is not merely an afterthought of grandiose theorising.

19Arbib and Hesse define schema as a "`unit of representation' of a person's world," including both synchronic and diachronic patterns of change (pp. 13f; cf. p. 61).

20Arbib & Hesse, 1986, p. 65. The details of relations between ontic levels and levels of description are purposely not specified in their presentation, however. Arbib & Hesse do describe their stance in general terms as a constructivism or perspectivism, distinguishing their view from "strong versions" of scientific realism (cf. pp. 10 & 182, passim). They also indicate that they consider their epistemology is compatible with many, but not all, ontologies. For the sake of clarity, these authors also note that the emphasis on levels originated with Hesse.

21The impact of scientific theory on our world via technology is only one form of epistemic influence. To prevent confusion, one would eventually need to distinguish this kind of influence (knowledge of physical reality impacting physical reality through human action) from, for example, self-fulfilling prophecy (social knowledge impacting on social reality; see, e.g., Watzlawick, 1984) or the effects of prayer (human action rooted in spiritual knowledge?) mediated through God's actions (spiritual reality?; see also note 22).

22A central point to sort out while exploring the ontic consequences of epistemic processes is the relationships among various levels and domains. For example, consider the "prayer-test controversy" which revolved around whether the prayers of Christians ought to result in empirically demonstrable consequences in physical domains (see Myers, 1978). Before one could test the efficacy of prayer, a well-developed theory (theology) would be required to predict what kinds of effects would be anticipated in various domains and at various levels: physical, biological, psychosocial, spiritual, etc .Likewise, an auxiliary theory of measurement would be necessary to indicate what observation processes would be required. Perhaps the prime question for scientists is whether any scientific approach can mesh with any subset of these questions. In effect, asking such questions is a natural consequence of a unified epistemology able to bridge spiritual and physical reality. Although we may not have, at present, fully formulated responses to these questions, my classroom experience attests to the clear relevance of these issues to the lives of many students (not to mention everybody else)! In my mind, the fruitfulness of Arbib and Hesse's goal is clear. Thus the tools we have available to move toward that goal, such as hierarchy theory, are of value.

23One strategy for accounting for empirical levels patterns is an epistemic pluralism (multiple levels of concepts) combined with a unileveled reality which, in a strong form of the thesis, sees levels as distortion due to the limitations of human cognition. This kind of approach can support either an instrumentalism which avoids ontic questions or perhaps a complex view of human fallibility. Although a realist approach certainly would need to develop a theory of error in human knowledge of hierarchies, realists would also require some form of validity in such knowledge. All I want to assert at this point is that I find it highly implausible to attribute all forms of hierarchy in the world to cognitive distortion.

I also assume in this discussion that the ontic impacts of mistaken understanding and valid understanding are distinguishable in some sense. From this point arises my reservations about instrumentalism.

24I make no attempt here to distinguish among various schools of thought using the label, nor to differentiate between "constructionism" and "constructivism." Arbib and Hesse use the term for their views and important parallels drawn with the other authors discussed here reflect the same general stance in my view. For overviews of a few forms of constructionism, see Watzlawick, 1984.

25I am the first to admit complexities associated with the relations between scholarship and faith. I am not trying to be glib, only to point out important concerns. It does not seem reasonable to anticipate easy answers for or against patterns of spiritual self-reference in scholarship. Arbib and Hesse's position stimulates such discussions as does the work of many other authors (e.g., Woltersdorff, 1984; MacKay, 1979; Mavrodes, 1977; Heie and Wolfe, 1987). My central point at this time is simply that elaboration of our levels notions is important to do justice to this issue, as well as others.

In the examples examined in this paper, obvious connections exist between evolution, molecular biology, and our understandings of creation. Rational and spiritual self-reference emerge behind every corner in such discussions. What Arbib and Hesse's epistemology offers is a framework within which to explore interrelations among specific theories of science and theology. Although the general case may permit the luxury of avoiding some details of levels structures, my clear sense is that elaborate hierarchy models are required for adequately addressing specific questions.


Allen, T.F.H. & Starr, T.B. (1982). Hierarchy: Perspectives for Ecological Complexity. Chicago: University of Chicago Press.

Arbib, M.A. & Hesse, M.B. (1986). The Construction of Reality. Cambridge: Cambridge University Press.

Ayala, F.J. (1974). Introduction to F.J. Ayala & T. Dobzhansky (eds.), Studies in the Philosophy of Biology: Reduction and Related Problems. Berkeley, CA: University of California Press.

Barbour, I.G. (1966/1971). Issues in Science and Religion. New York: Harper & Row. (originally published in 1966 by Prentice-Hall)

Bartlett, S.J. & Suber, P. (eds.) (1987). Self-reference: Reflections on Reflexivity. Dordrecht, Holland: Martinus Nijhoff.

Beckner, M. (1974). Reduction, hierarchies, and organicism. In F.J. Ayala & T. Dobzhansky (eds.), Studies in the Philosophy of Biology: Reduction and Related Problems. Berkeley, CA: University of California Press.

Bube, R.H. (1971). The Human Quest: A New Look at Science and the Christian Faith. Waco, TX: Word.

Bunge, M. (1956). Do the levels of science reflect the levels of being? In M. Bunge, Metascientific Queries. Springfield, IL: Charles C. Thomas.

Bunge, M. (1977). Levels and reduction. American Journal of Physiology, 233(3), R75-R82.

Campbell, D.T. (1974). `Downward causation' in hierarchically organised biological systems. In F.J. Ayala & T. Dobzhansky (eds.), Studies in the Philosophy of Biology: Reduction and Related Problems. Berkeley, CA: University of California Press.

Capra, F. (1982). The Turning Point: Science, Society, and the Rising of Culture. New York: Simon & Schuster.

Cramer, J.A. (1985). Science, scientism and Christianity: The ideas of D. M. MacKay. Journal of the American Scientific Affiliation, 37(3), 142-148.

Davies, P.C.W. (1988). The Cosmic Blueprint: New Discoveries in Nature's Creative Abilities To Order the Universe. New York: Simon & Schuster.

Glenn, S.S. (1988). Contingencies and Metacontingencies: Toward a synthesis of behavior analysis and cultural materialism. The Behaviour Analyst, 11, 161-180.

Grene, M. (1967/1974). Biology and the problem of levels of reality. In M. Grene, The Understanding of Nature: Essays in the Philosophy of Biology (vol. 23 of the Boston Studies in the Philosophy of Science; pp. 35-52). Dordrecht, Holland: D. Reidel. (reprinted from The New Scholasticism, 41, 94-123).

Grene, M. (1972/1974). Aristotle and modern biology. In M. Grene, The Understanding of Nature: Essays in the Philosophy of Biology.

Haas, J.W., Jr. (1983). Complementarity and Christian thought-An assessment: 2. Logical complementarity. Journal of the American Scientific Affiliation, 35, 203-209.

Hart, H. (1984). Understanding Our World: An Integral Ontology. Lanham, MD: University Press of America.

Heie, H. & Wolfe, D.L. (eds.) (1987). The Reality of Christian Learning: Strategies for Faith-Discipline Integration. Grand Rapids, MI: Eerdmans.

Leake, C.D. (1969). Historical aspects of the concept of organization levels of living material. In L.L. Whyte, A.G. Wilson, & D. Wilson (eds.), Hierarchical Structures. New York: American Elsevier.

MacKay, D.M. (1969). Information, Mechanism and Meaning. Cambridge, MA: M.I.T. Press.

MacKay, D.M. (1979). Human Science and Human Dignity. Downers Grove, IL: InterVarsity Press.

MacKay, D.M. (1982). Ourselves and our brains: Duality without dualism. Psychoneuroendocrinology, 7, 285-294.

Mavrodes, G.I. (1977). On Christian scholarship (letter). The Reformed Journal, 27(July), 4-5.

Myers, D.G. (1978). The Human Puzzle: Psychological Research & Christian Belief. New York: Harper & Row.

O'Neill, R.V., DeAngelis, D.L., Waide, J.B., & Allen, T.F.H. (1986). A Hierarchical Concept of Ecosystems. Princeton, NJ: Princeton University Press.

Orlebeke, C.J. (1977). Donald MacKay's philosophy of science. Christian Scholar's Review, 7, 51-63.

Pattee, H.H. (1971/1976). Physical theories of biological co-ordination. In M. Grene & E. Mendelsohn (eds.), Topics in the Philosophy of Biology (vol. 27 of the Boston Studies in the Philosophy of Science, pp. 153-173). Dordrecht, Holland: D. Reidel. (reprinted from Quarterly Review of Biophysics, 4, 2-3).

Pattee, H.H. (1973). Postscript: Unsolved problems and potential applications of hierarchy theory. In H.H. Pattee (ed.), Hierarchy Theory: The Challenge of Complex Systems. New York: George Braziller.

Pattee, H.H. (1978). The complementarity principle in biological & social structures. Journal of Social & Biological Structures, 1, 191-200.

Pattee, H.H. (1979). The complementarity principle and the origin of macromolecular information. BioSystems, 11, 217-226.

Peacocke, A. (1986). God and the New Biology. San Francisco: Harper & Row.

Polkinghorne, J. (1986). One World: The Interaction of Science and Theology. Princeton, NJ: Princeton University Press.

Salthe, S.N. (1985). Evolving Hierarchical Systems: Their Structure and Representation. New York: Columbia University Press.

Van der Meer, J. (1989). Hierarchy: Towards a framework for the coherence of faith and science. Pro Rege, 17(3/4), 19-33.

Van Leeuwen, M. Stewart (1983). Reflexivity in North American psychology: Historical reflections on one aspect of a changing paradigm. Journal of the American Scientific Affiliation, 35, 162-167.

Watzlawick, P. (ed.) (1984). The Invented Reality. New York: Norton.

Webster, J.R. (1979). Hierarchical organization of ecosystems. In E. Halfon (ed.), Theoretical Systems Ecology: Advances and Case Studies. New York: Academic Press.

Williams, G.C. (1985). A defense of reductionism in evolutionary biology. Oxford Surveys in Evolutionary Biology, 2, 1-27.

Wimsatt, W. (1986). Forms of aggregativity. In A. Donagan, A. N. Perovich, Jr., & M.V. Wedin (eds.), Human Nature and Natural Knowledge: Essays Presented to Marjorie Grene on the Occasion of Her Seventy-fifth Birthday (vol. 89 of the Boston Studies in the Philosophy of Science; pp. 259-291). Dordrect, Holland: D. Reidel.

Woltersdorff, N. (1984). Reason Within the Bounds of Religion (2nd ed.). Grand Rapids, MI: Eerdmans.

Part II