The hierarchical relation

This work is licensed under the Creative Commons | © Michael Houseman. ISSN 2049-1115 (Online). DOI: http://dx.doi.org/10.14318/hau5.1.012


The hierarchical relation

A particular ideology or a general model?

Michael HOUSEMAN, École Practique des Hautes Études

Translated by Eléonore Rimbault

The hierarchical relationship introduced by Louis Dumont has yet to receive the attention it deserves. Drawing on Dumont’s own remarks, this text sets out to present the hierarchical model in a systematic fashion. After dealing with a number of definitional issues (the notions of encompassment, value, levels, ideology, etc.), it highlights the theoretical originality of this model and considers some of the questions it raises: What possible forms can it take? How does it relate to empirical facts? What mental processes does it presuppose?

Keywords: Louis Dumont, hierarchy, hierarchical relationship, encompassment, representations

I believe that hierarchy is not, essentially, a chain of super-imposed commands, nor even a chain of beings of decreasing dignity, nor yet a taxonomic tree, but a relation that can succinctly be called “the encompassing of the contrary.”

(Dumont 1980: 239)

The currently emerging debate on the idea of hierarchy encounters from the start an all too frequent problem in the social sciences, and in particular in anthropology, [252]version of the principle of hierarchynamely, that of propositions which, in spite of their theoretical implications, have not been subjected to rigorous scrutiny. My aim here is to address this problem.

The principle of hierarchical relationship was introduced by Louis Dumont in 1966 by an innovative analysis drawing both on Talcott Parsons’ ideas on structural affinities between subsystems of the social system, and on the organizational features of Indian castes. Yet this idea has not been met with the substantial debate one might have expected, and has encountered numerous resistances—among which the strongest has certainly been the tendency to reduce hierarchy to a particular form of social stratification. It is therefore difficult to assess the relevance of this principle without scrutinizing it further, so as to assess its advantages and limitations. Either the notion has little validity, which requires demonstration, or it possesses merits and virtues, in which case they should be put to use. Neither of these two possibilities have been pursued thus far.1

Some may argue that in the absence of a more precise and expanded formulation, this principle remains difficult to use. In this perspective, it is regrettable that, aside from a few pages of a Postface entirely devoted to it, the hierarchical relationship has yet to have been presented in a systematic fashion, and has thus remained the implicit generalization of a comparative sociology based on Indian material. The necessity of such a presentation as a stage prior to any discussion justifies the internal organization of the following pages. Drawing on Dumont’s remarks, I will first try to specify what, exactly, a hierarchical model might be. Then, by considering some of its aspects—and some of the questions it raises—more closely, I will foreground its originality.


The principle of hierarchy as put forward by Dumont essentially rests on the following proposition: most of the time, if not always, the terms of a given opposition are related in different ways to the whole they compose; it follows that any relationship that distinguishes between them is inseparable from a reference to the whole that orders them with respect to each other. It is nevertheless useful, for the sake of clarity, to discriminate between two versions of this principle. The first, which I will call the “restricted” version, specifies the nature of this difference: one of the terms is identical to the whole, and as a consequence, encompasses the other term, its contrary. This is a “hierarchical opposition” proper. The other version of the principle, which I will call the “general” version, and of which hierarchical opposition is but a special case, does not specify the nature of this difference; it simply calls attention to its existence and draws consequences from this, among which the most important is the taking into account of differences of value. Let us consider these two versions in order.

[253]The restricted version of the principle of hierarchy proposes that just as a distinctive opposition (positing a relationship of contrariety between two terms) is recognized, a hiearchical opposition (defined as an encompassing-encompassed relationship between these terms) should be recognized as well. As an example, Dumont invites us to consider sexual differentiation as presented in the Biblical narrative, that is, the creation of Eve from Adam’s rib. In this case, Adam is two things at once, and the relationship between him and Eve is consequently dual: on the one hand, as a prototype of the human species’ male individuals, Adam is opposed to Eve; on the other, as the original representative of this species from which Eve was extracted, he encompasses her. The English word “man,” by signifying both the category of male human beings and that of human beings in general, has a similar bivalence. In both of these examples, one mythical, the other linguistic, we can distinguish two levels: on a higher level, there is unity, on a lower level, distinction. These two relations taken together—i.e., the conjunction of identity and contrariety—constitute the hierarchical relationship.

Thus, hierarchical ordering does not derive directly from the intrinsic qualities of the entities concerned, but from the interrelation of the categories these entities represent. More specifically, it refers to a relation between a whole (or a set) and one element of this whole (or set) where:

the element [e.g.: Eve, “woman”] belongs to the set and is in this sense consubstantial or identical with it; at the same time, the element is distinct from the set or stands in opposition to it. This is what I mean by the expression “the encompassing of the contrary.” (Dumont 1980: 240)

Easily lending itself to formalization (e.g., Euler diagrams), the hierarchical relation not only constitutes a clarifying alternative to the simple binary opposition, but also provides a powerful and rigorous tool for intercultural comparison. In order to better appreciate these two benefits of hierarchy, one must consider the general version of this principle.

The general version of the principle of hierarchy proposes that most oppositions, if not all, have a hierarchical aspect. To the extent that the terms of an opposition do not stand in the same relationship with respect to the whole they comprise, their differentiation is inseparable from a reference to this whole that orders one in relation to the other. And it is this reference to a totality, which, by ordering these elements, establishes a difference of value between them. Hence, the relative value of a pair of terms,

determined by their relation to the whole of which they are a part, is inherent in their distinction; it … cannot be dissociated from their distinction as if there were, on the one side, an idea of simple polarity, and on the other, a value that would be superadded to it. (Dumont 1980: 243–44)

The classical problem of lateral symbolism (the preeminence of the right hand) may serve to illustrate the usefulness of this principle. As a rule, this question has been approached by assuming a symmetry between the two hands, such that one goes about accounting for the preeminence of one hand over the other by linking the left-right dichotomy to other oppositions that are deemed homologous (man/[254]woman, day/night, tribal moieties, etc.). The perspective offered here sheds a new light on this question. The terms left and right cannot be defined in and of themselves, but necessarily refer to a whole, the body, which defines and organizes these terms, and to which they relate differently. In any situation therefore, it is clear that hands cannot be equal: “The implicit reference to the whole of our body has as its necessary consequence the preeminence of one of the hands over the other” (Dumont 1980: 243).

Hence the preeminence of the right hand turns out to be a false problem, at least to a certain extent. It is a problem which arises when one forgets to refer to a totality, and when an unfounded separation between facts (such as a presumed symmetry) and values (a specified asymmetry) is introduced. This separation is illegitimate: hands, Dumont notes, are different in value and nature at the same time. The preeminence of one over the other is therefore not contingent but necessary. The same is most probably true of other oppositions raised in relation to the right-left couple.2

Recognizing the hierarchical aspect of oppositions does not, of course, resolve the problem of their explanation. For instance, one still needs to account for why the right hand is generally favored over the left, although not always. But it does give us a more realistic framework to work with: “by substituting an asymmetrical or ordered opposition for a symmetrical or equi-statutory opposition that does not exist, we are drawing closer to the thought we are studying” (Dumont 1979: 811). Considering values forces us to take into account specific situations, particular points of view, and it introduces a new rigor in the study of the relationship between cultural elements.

Indeed, a hierarchical account of oppositions requires us to distinguish between the situations and contexts in which these oppositions arise. Some more classical approaches, in particular those that seek to carry out binary classifications, may of course take such situations into account. However, most often, by introducing an artificial separation between the facts (or ideas) and the way they interact within a system of values, they tend to generalize the relations they establish without contextual reference. The construction of tables of equivalence that frequently accompanies such efforts exemplifies this mode of reasoning.

In short, the distinction of situations ceases to be considered as pertinent at the moment we pass from elements to the set as a whole, as though each situation were in itself independent of the “mentality” as a whole, though it should be evident that the very distinction of situations depends on the mentality in question. (Dumont 1979: 808)

By contrast, in a hierarchical perspective, given that the relative value of a couple of terms, which is seen as constitutive of their distinction, relates to specific situations, defining the context is crucial to assess both particular oppositions and the way they relate to each other.

In this manner, a hierarchical perspective forces us to differentiate situations both with regard to their nature and to their relative importance in relation to the [255]system of cultural representations of which they are a part. What should one understand, then, by situation? The value of a series of terms, it has been previously noted, derives from their relation to the totality they constitute. If we further assert that distinguishing between situations requires an attention to values as well, it is because doing so involves distinguishing between different totalities. In this way, paying systematic attention to situations or contexts entails making a distinction between the different ordered sub-sets, or, more exactly, the sub-systems that compose a system of cultural representations. Dumont calls these sub-systems “levels,” and he calls the sum of levels or sub-systems that comprise a given system of cultural ideas and values (he speaks of “value-ideas” [idées-valeurs]) an “ideology” (Dumont 1979: 808).

Here, a new problem arises: how can such sub-systems (or levels) be delineated? The principle of hierarchy itself provides an answer: by changes, or, more precisely, reversals of value.

By definition, a symmetrical opposition may be reversed at will: its reversal produces nothing. On the contrary, the reversal of an asymmetrical opposition is significant, for the reversed opposition is not the same as the initial opposition. If the reversed opposition is encountered in the same whole in which the direct opposition was present, it is evidence of a change of level. (Dumont 1979: 811)

Value reversals—for instance when the left is favored over the right, when “woman” is presented as encompassing “man” rather than the opposite, etc.—occur in situations or contexts associated with identifiable levels or sub-systems within an ideology. Taking the existence of such sub-systems into account as an essential feature of an ideology allows us to give a simple structural explanation for these inversions that might otherwise—considering the trouble they introduce in interpretive attempts based on binary classification—be brushed aside.

Nevertheless, we have yet to determine how these sub-systems or levels are organized among themselves. Dumont’s hypothesis is implicit in his two-fold use of the word “level,” which he adopts to designate both the different dimensions of a given hierarchical opposition (identity and contrariety), and the different sub-systems in which this opposition and its reversal arise. The unitary integration of these subsystems, as well as their distinction, follow a hierarchical ordering: one level is contained in another according to the principle of encompassment of the contrary (hierarchy in the restricted sense). Thus, the elements of a sub-system are distinguished in terms of value with respect to the totality this sub-system represents, and furthermore, the sub-systems themselves are also distinguished in value with respect to the totality of which they form a part, that is, ultimately, the ideology as a whole.

When dealing with an opposition between a pair of terms, it is thus important, while taking into account a relation of superiority/inferiority between them, to also specify the ideological level at which this hierarchical relation itself is situated. One tribal moiety, for instance, may be represented as superior to the other in some situations, and inferior in others; this is because their interrelationship pertains to two different levels or sub-systems, themselves hierarchically ordered within a system of representation according to the principle of encompassment and not just by virtue of being attributed different ranks.

[256]Note, finally, that in the light of this hypothesis, the presence of a plurality of sub-systems or levels (indexed by value reversals) is a quality inherent to the organization of ideological systems. Indeed, a given hierarchical opposition, a relation of superiority/inferiority between two terms, “cannot be true from one end of experience to the other (only artificial hierarchies make this claim), for this would be to deny the hierarchical dimension itself, which requires situations to be distinguished by value” (Dumont 1980: 244). Value differences sustained by the hierarchical principle are thus not unilateral but “bidimensional”: the occurrence of a hierarchical relation between a pair of terms in itself implies its reversal—i.e., the realization of an ordering that is both “contrary” to the first and encompassed by it (and therefore subordinated to it). In this way,

The reversal is built-in: the moment the second function [that of the inferior term of the couple] is defined, it entails the reversal for the situations belonging to it. That is to say, hierarchy is bidimensional, it bears not only the entities considered but also on the corresponding situation, and this bidimensionality entails the reversal (ibid.: 225)

We can now see that underlying this first overview of the “restricted” and “generalized” versions of the principle of hierarchy is a particular overall understanding of ideological systems which includes the encompassing-encompassed relationship, a taking into account of value differences, the bidimensional nature of superior/ inferior relations, and so forth. This view nonetheless remains implicit to a great extent. To make it more explicit, let us look further into the nature of a general model based on the hierarchical principle.


Hierarchy is considered here to be a general principle governing the organization of ideological systems. The latter relate not to the world, but to representations of the world. It seems unlikely, however, that the global structure of these systems is present as such, as a kind of road map to the ideology concerned, in the minds of each of its actors. Thus, in the perspective considered here, systems of representation founded on the principle of hierarchy do not reproduce the thinking of particular participants. Rather, they are interpretations constructed by observers to account for the ordered, collective integration of such thinking—an integration (and hence, the relevance of these interpretations) which is attested by the systematic quality revealed in the study of social phenomena. The use of these constructions is primarily analytical: they provide structural models allowing for accounts of such phenomena as integrated into overall arrangements determined by a limited number of principles. From this point of view, hierarchy is hardly unique: other principles, such as those of dialectics or structuralism, allow for the elaboration of other types of models.

Dumont himself noted on several occasions the formal homology existing between hierarchy and dialectic, that consists in there being two levels, one of which transcends the other. But this is their only common feature. Indeed, whereas for dialectics the distinctive opposition between a pair of terms is a relation of contradiction, for hierarchy it is a relation of complementarity. While the former stems from a substantialist conceptualization and primarily deals with the terms themselves and their logical relations with another, the latter is grounded in structural [257]thought and is concerned with the “universe of discourse” formed by these terms and the internal relations that constitute it. Indeed, it is in light of this difference that we can understand the two-dimensional character of hierarchical oppositions. It is because the hierarchical relation establishes a relation of complementarity, not contradiction, that a superior/inferior relationship between a pair of term can encompass its opposite, entailing an inversion of these values. Such an inversion does not neutralize the relation between the elements concerned, but, on the contrary, enriches the discursive universe they comprise.

Contrasting these two principles from the point of view of the dynamics they bring into play only reinforces this fundamental divergence:

If we try to imagine a diachronic aspect, a process, corresponding to [complementarity], it will be differentiation, by which what was our “universe of discourse,” supposed to be at first unified, splits into two [complementary] opposites. Contradiction on the contrary has been used as a principle of process of development in time. (Dumont 1971: 77)

Hierarchy and dialectic are mirror images; they are, in this sense, true opposites:

[D]ifferentiation does not change the global setting, given once and for all; in a hierarchical schema the parts that nest one side the other may increase in number without changing the law.

The Hegelian schema based on the contradiction works quite differently. By the negation and the negation of the negation, a totality without precedent can be produced synthetically. In fact, in Hegel’s thought the deepest motive is to produce … a totality from a substance. In the hierarchical schema, in the contrary, the totality preexists and there is no substance. (Dumont 1980: 243)

These remarks allow us to emphasize, in passing, how important it is to distinguish between hierarchy and social stratification. While the latter denotes a relation of unilateral subordination between elements, hierarchy denotes a bidimensional (implicitly reversible) subordination between them that is based on a reference to a totality. In this manner, stratification and simple subordination, on the one hand, and hierarchy, on the other, represent inverse perspectives: in one, the sum of particular inferior/superior relations, taken as primary units, are aggregated to form a whole, whereas in the other, a differential reference to this whole determines the constitution of the relations between the elements it assembles.

At the same time, the hierarchical principle contrasts sharply with the underlying principle of the French structuralist approach:

Now, while in binary opposition the distinctive opposition used in its pure form both atomizes the data and renders it uniform, the hierarchical distinction unifies the data by welding together two dimensions of distinction—between levels and within a single level. (Dumont 1979: 813)

Developing this idea further, we may say that according to the hierarchical model, the ideological system, far from being monolithic is, on the contrary, both composite and contextual: its integration depends not on a commutative logic, based on permutation and analogy, and applied to homogeneous entities, but to an oriented logic, that of asymmetrical encompassment, which preserves the heterogeneity of the elements it [258]brings together. In short, the hierarchical principle differs from the structuralist perspective in that it takes the complex nature of cultural phenomena into consideration.

Indeed, recognizing hierarchy is to place oneself in the field of complexity that lies to either side of structuralism’s implicit assumption, which is that the whole can be reduced to the sum of the relations between its constitutive parts. With the restricted version of the principle of hierarchy, that is, the hierarchical opposition per se, the whole is taken as equal to one of its parts; in the general version of this principle, it is held to be more than their sum. Now, it is precisely this complex quality of the hierarchical model that underlies the integrative capacity of the systems it determines. Thus hierarchy produces, by definition, configurations that cannot be exhausted by their taxonomic interpretation. The latter, which systematically results in inconsistencies—“among men, there are men and women” is such an example—is necessarily partial. In other words, the principle of hierarchy acts not so much to register observable differences, but rather to establish relevant distinctions by incorporating them into global schemes—wholes—that cannot be reduced to simple classificatory principles. In sum, the hierarchical relation, as a source of complexity, generates structure.

We will come back to this, but let us note here that, by virtue of this complexity, a hierarchical model is also radically different from the normative models advanced by British structural functionalism: contrarily to the latter, a model based on the principle of hierarchy cannot define unilateral rules of organization based on the absolute value of particular concepts (e.g., unilineal descent, exogamy, etc.). In the perspective outlined here, not only does a given value-idea (concerning, for instance, group membership, residence, matrimonial prescription or proscription, etc.) encompass its “contrary,” but it is also integrated with the latter in a relationship of subordination which can similarly encompass its own reversal.

What sets hierarchical models apart from those founded on dialectics or (French or British) structuralism, that is, that which distinguishes the hierarchical opposition from contradiction, simple binary opposition, or unilateral subordination is its dependence on a reference to a whole. The hierarchical model rests on this holistic proposition: the integration of a system (or a relationship) is determined primarily by its incorporation in a larger system. Now, this raises a new problem: what about the system of cultural representations as a whole, the ideology itself? If we take the above-mentioned proposition seriously, an ideology’s integration requires that it also must be incorporated in a totality of a higher order. And for the latter to be truly ultimate, it must remain, at least in part, inaccessible. In other words, it must exceed or transcend ordinary experience. Thus, if the hierarchical relation depends on a holistic principle, this principle in turn requires a logical legitimation of a transcendental order. To partake of an ideology is to recognize a relation to a transcendental whole which, following the hierarchical scheme, at once encompasses the social order per se and, at another level, is distinct from it. This ultimate reference, in that it is presupposed by the very coherence of the ideological system, permeates it and animates it in its most minute details. This is surely how we must understand what Dumont calls the cardinal value of an ideology.

Based on the preceding comparisons, we may conclude that an ideological model based on the principle of hierarchy can be identified as having three interdependent characteristics: (1) refusing to separate ideas from values, and consequently [259]requiring an attention to contexts and situations, it is made up of elements (valueideas) connected by relations of bidimensional subordination (hierarchical oppositions) whose constitutive aspects are complementary; (2) by combining different levels, and combining identity and contrariety both within a hierarchical opposition and between the latter and its implicit reversal, it engenders complex configurations and sub-systems, that is, ones that cannot be reduced to simple classificatory schemes (e.g., taxonomies) or unilateral normative rules; (3) positing a holistic perspective, it makes a transcendental reference, that is, the recognition of a relation with something inaccessible (ancestors, spirits, deities, God, etc.) a necessary condition for the cohesion of these sub-systems and for the integration of the ideology as a whole.


Having outlined the hierarchical model, it should be clear that it raises a host of questions. I will now try to provide some answers, considering first the diversity and the significance of the hierarchical relation, and then the relationship between the hierarchical model and empirical facts. Finally, I will consider how the model might be justified on the level of the mental processes it supposes.

Our first problem has to do with the possible diversity of hierarchical arrangements and, more generally, the overall organization of an ideology. Since Dumont asserts that most oppositions have a hierarchical dimension, we should wonder how many distinct types of hierarchical relations can exist. Considering another of Dumont’s affirmations, that hierarchy is essentially the presence of an “encompassing of the contrary” (Dumont 1980: 239), we may also ask: how many kinds of “encompassing of the contrary” can be distinguished, and how are these different types aggregated?

Let us first consider two of these types of hierarchical relations. One, which I call the “extensive” type, is the hierarchical opposition per se (the case of the word “man,” for instance), that is, the case in which the extension of a set or category includes, at another level, its anti-extension (“woman”). The other, which I will call the “anti-extensive” type, is simply the reverse: the case of a set or category whose anti-extension contains, at a different level, the set or category itself. The distinction between the initiated and the uninitiated mediated by male initiation rites provides an especially rich example of the latter type. At one level, this dichotomy corresponds to the masculine/feminine opposition, where the (feminine) category of the uninitiated includes male individuals (boys) as well as female ones (women and girls). At a different level and similarly, it corresponds to the opposition between adults and children where the category of the uninitiated (children) includes both adults (women) and children (boys and girls).

From a dynamic perspective, this second type of hierarchy is a development of the first: while the hierarchical opposition properly speaking underlies—as it does in the Genesis narrative—the emergence of a duality from a singularity, in this case a duality, the initiated/uninitiated opposition, gives rise to a tripartite division that distinguishes between individuals having undergone initiation, those having yet to undergo it, and those who will never undergo it. And this three-fold distinction in turn leads to the emergence of a new dichotomy between, on one hand, persons naturally destined to undergo initiation and, on the other, those who, by their very nature, are excluded from it. This of course is but a confirmation, in cultural terms, [260]of the biological sexual difference, and, in that respect, initiation does not create new categories. Rather, it imposes a (culturally) reconfigured way of understanding preexisting (natural) ones. Indeed, by incorporating a distinction between the sexes, as well as that between adults and children into a unitary hierarchical scheme, it institutes these discriminations. In other words, by combining the categories of age and sex in a hierarchical relationship, it orients the taxonomic interpretations that can be elaborated with regards to them, thereby legitimating the emergence of genuinely new ideological propositions, such as: “among male individuals, some (initiated men) are more masculine than others (uninitiated men, boys),” “among adults, some (women, uninitiated men) are less adult than others (initiated men),” “among children, some (boys) are more childlike than others (girls),” and so forth.

From a synchronic point of view, the two types of hierarchies described above can be combined in what we might call a “modular” hierarchical relation: a hierarchical relationship than encompasses its own reversal. Such is the case of the left/ right opposition where one of the hands, the one that usually dominates, is subordinated to the other in situations that are associated to specific functions of the latter. We thus arrive at the definition of three elementary hierarchical configurations. They all include, in different ways, two elements or value-ideas: (1) the hierarchical opposition proper, which is an “extensive” hierarchy (that can be extended recursively through the addition of supplementary terms where A encompasses B, B encompasses C, etc.); (2) the inverse of this opposition, that is, an “anti-extensive” hierarchy; and (3) a “modular” hierarchy which is a possible aggregation of form (1) and (2), with specific properties (e.g., reversal) (see Figure 1).

Figure 1
Figure 1.

This typology leads to the following hypothesis: an ideology’s structure, though it may vary from one case to the next, is always founded on an articulation of these three modes of elementary hierarchical relation. Some features of the Biblical narrative of the Genesis illustrate such an articulation. In Figure 2, the story of the Genesis appears as a series of hierarchically ordered separations or differentiations, where various elements (sky, earth, man, etc.) are simultaneously arranged [261]at different levels so as to form sub-systems which encompass each other (e.g., creation/procreation). The “cardinal value” of this ideology, the divine, intervenes as a recurrent reference, underlying at once the constitution of these different subsystems and their overall cohesion. As the example suggests, from the perspective I have laid out, the hierarchical organization of a system of representations does not correspond to a fixed ordering of interrelations into a set layering of levels. Rather, it corresponds to an “open” articulation of hierarchical relations whose global integration is based, in the end, on an ultimate reference (the cardinal value) which, to a variable extent, is common to all of them.3

Figure 2
Figure 2.

[262]A second problem concerns the limits of the hierarchical model and, in connection with this, what we might call its general symbolic productivity. If we state that most oppositions have a hierarchical dimension, does “most,” in this case, correspond to an empirical assessment? A theoretical necessity? In other words, is the presence or absence of hierarchical relations determined by particular conditions? By specific classes of phenomena? Even if we accept (Dumont [1980] 2013) that any opposition is implicitly hierarchical—that the recognition of difference is ipso facto discriminatory, that distinction and equality are antithetical—this does not mean that hierarchy is similarly relevant everywhere and at all times. Thus Dumont notes, on several occasions, that modern ideology seems to work against an awareness of hierarchical relations. It remains to be elucidated whether there exists domains or conditions of knowledge in which hierarchy is systematically favored, and others where it is systematically minimized or concealed.

The development of the hierarchical principle itself suggests a possible answer to this question: while this model was inspired by a classificatory system of social categories (Indian castes), it does not generally fit well with indigenous systems of classification of natural species. Might this be an invariant for which totemism—in which natural distinctions provide templates for thinking about social differentiation—would be the exception that proves the rule? The hypothesis entertained here is that in a given system of cultural representations, hierarchy is privileged to order elements considered to be specifically human. Resorting or not resorting to hierarchical configurations draws as it were the boundary between a properly human world, Society or Culture, and the rest, Nature. For instance, by incorporating age and gender differences into a hierarchical scheme, initiation culturalizes or humanizes these discriminations. Having thereby become the legitimate objects of social determination, they are radically distinguished from analogous differences visible in the animal world.

If we accept this hypothesis, modern ideology, which minimizes hierarchy, does not move towards a humanization of the natural world; on the contrary, it aims towards a naturalization of the human world, all the way to the ultimate limit imposed by the individual subject, raised to the rank of cardinal value.4 Now, by proceeding in this way, modern society is confronted with a paradox: it strives to deny that which allows it to define itself. As a result, a determination of the social sphere can only be accomplished either by valorizing contradictions explicitly, or by hierarchical relations expressed in the absence of hierarchy, that is, as skewed relations of equality. Pushed to its extreme, this tendency leads to the totalitarian double-talk described by George Orwell: “war is peace,” “freedom is slavery,” “ignorance is strength,” the Ministry of Truth tells us in 1984; “All animals are equal,” affirms the pig-dictator of Animal farm, “but some are more equal than others.”[263]


Until now, the hierarchical model has been envisaged essentially in relation to systems of representations. Now, I would like to consider briefly the connection between this ideological model and empirical forms of social organization. Indeed, the principle of hierarchy may offer an original solution to the current anthropological dilemma regarding the relationship between ideology and observable social reality.

The most common approach to this problem was developed on the basis of African material by British structuralists. It posits a direct link between ideological models and social forms. Ideology defines a number of clear-cut structural principles governing norms of organization; those principles, expressed by (morally connoted) juridical rules concerning, in particular, group composition (recruitment criteria, rights and duties of group members, etc.), are deemed to determine actual social forms. The most important limitation to this otherwise fruitful approach is its presentation of an idealized image of the social morphology, which, all too often, conceals the variations of actual behavior: direct observation reveals significant divergences between the structural model which foregrounds ideology and the empirical reality it is supposed to account for.5

The other attempt to solve this problem is the outcome of the disillusion which emerged in particular when the first approach was confronted with Papua New Guinea societies that seemed to present an even wider gap between theory and practice. In this alternative perspective, the link between ideology and social form is held to be minimal if not non-existent. In short, it stipulates that society or at least some societies are intrinsically disordered; they are, as the expression has it, “loosely structured.” In such cases, an overall ideological model is held to be largely irrelevant; empirical facts, deemed elusive to structural analysis, are to be accounted for with models of individual decision-making based on the circumstantial intervention of factors external to social organization itself such as ecological, demographic, or war-related conditions.

Very schematically, we could say that anthropological research is faced with the following problem: either it accepts a structural model that does not entirely account for the facts, or it takes these facts into account, but in the absence of such a model. A structural model of the hierarchical type, by postulating an indirect relationship between ideology and social forms, provides a possible solution to this dilemma.

In such a model, structural principles, which relate to value-ideas in hierarchical relationships, do not define unilateral normative rules. As we have seen, not only does a given organizational principle encompass its opposite (or is encompassed by it), but furthermore, the relationship of subordination between these two “complementary” principles is implicitly reversible. The preeminence of one or the other will depend on the context.

From an abstract point of view concerned with the system of representations itself, defining organizational principles in terms of general situations favoring their [264]respective preeminence is fairly straight-forward; this is what allows for the elaboration of “ideal” indigenous models entailing absolute organizational prescriptions. However, from an empirical point of view that attempts to define particular, concrete situations with regards to the relative preeminence of these principles, the task is more complicated. Such a definition, which invariably involves circumstances specific to each case being considered, is necessarily more ambiguous. Of course it may happen that the evaluation of all participants concur, in which case the situation will be determined automatically, so to speak. However, more frequently, this determination emerges out of the confrontation of divergent evaluations, whose resolution—a “consensus” of value judgments (often imperfect)—is mediated through transactions (of words, of goods, etc.) among the actors involved. Either way, in both cases, the choice is the same: either one or the other of the principles of a “complementary” pair of principles will be applied; which one it is will depend on the definition of the situation the participants arrive at. Therefore, at the empirical level, the actualization of a hierarchical model, because it requires a contextual evaluation to determine the preeminence of the principles it incorporates, relies on particular choices. It is the aggregation of such particular choices that constitutes the social forms one can empirically observe; in turn, these forms provide a contextual reference which is of great importance in the determination of subsequent choices.

The relation between ideology and social forms, thus meditated by particular choices, does not, however, lead to an inherently disordered social organization. This is because the value-ideas organizational principles point to, but also the contextual situations that correspond to their relative preeminence, are hierarchically ordered among themselves. Consequently, for every situation considered, the majority of choices will point in one of the directions possible. Thus, (statistical) norms of organization are effectively realized; ideal models are indeed justified (both by anthropologists and by those they study) by the existence of these overall regularities. However, there remains a minority of choices, concealed by these idealized constructions, that run counter to these norms. The model proposed here also accounts for these; they also contribute, just like normative choices, to the ongoing elaboration of social structure.

In this manner, the hierarchical model takes into account the existence of (statistical) norms of organization, the elaboration of indigenous ideal models, and incorporates all actual behaviors—including those that run counter to these norms and models—into a structural analysis. Thus it solves the (false) problem of the disjunction between ideology and empirical facts. By doing so, it eliminates the unsatisfying notion of “loosely structured” societies.

The model proposed here is not less structural because it is not reducible to unilateral prescriptions, nor are the hierarchical configurations described above less classificatory because they cannot be reduced to taxonomies. However, it may be better indicated to qualify this model as “structuring” rather than “structural”: it does not describe a form of organization, rather—and this might be a superior level of explanation—it describes the conditions that determine the emergence of this form. Therefore it would be useless to attempt to draw from it generalizable rules of behavior: these are necessarily either contradictory (incorporating the encompassing of a contrary), or false (considering only one level). However, when combined [265]with knowledge of the diverse factors that can intervene in individual choices, it can help determine, and thereby explain certain overall systems of behavior. Earlier we saw how, in the case of initiation, hierarchy intervenes not to define categories of age and sex in a univocal fashion, but rather to orient thinking about them. Likewise, hierarchy intervenes here not by defining, in an absolute fashion, a set of norms for action, but by structuring behavior in reference to these norms. In the case of initiation, this results in a “system of thought” whose emerging properties cannot be deduced from the categories themselves: “among adults, some (initiated men) are more adult than others (women, non-initiated men).” The same is surely true of social systems.6


Now that we have considered the usefulness of a hierarchical model for the analysis of representations as for that of empirical facts, we must, finally, question the admissibility of this model at the level of cognition. We may ask ourselves: what do we have to assume about mental processes in order to affirm the relevance of an ideological system based on the hierarchical model? The more substantial these prerequisites are, the less relevant the model will be: the economy of the ideological constructions it proposes would be contradicted by the weight of the presuppositions regarding the mental processes it requires.

A possible answer to this question, offered as an hypothesis, consists in only admitting one thing: thought is discriminatory, it inherently establishes superior/ inferior relations. This amounts to assuming that, by virtue of certain imperatives and criteria, some of which may be of a non-conceptual order, the mind effectively differentiates at once in value and in nature. Along with this assumption, which does not seem unreasonable (it matches what we know about the process of socialization for instance), we must also add the observation that an individual can make value judgments that are logically incompatible; he or she can affirm, at different moments and in different situations that A>B and B<A. The global accommodation of such judgments corresponds to what the hierarchical model implies. First, the coexistence in the mind of contradictory discriminations gives rise to a distinction between contexts, that is, between different wholes or “levels.” Then, given that the discriminated entities are in different relationships with respect to these contexts, they acquire a truly “hierarchical” value according to Dumont’s definition. Finally, in that these contexts are themselves differentiated from each other, they themselves become the object of a valued discrimination. It is thus sufficient to assume that the mind produces discriminations that are logically incompatible in a more or less systematic manner—discriminations that also apply to the contexts in which these valued distinctions occur—in order to arrive at the overall structure described by this model.

[266]Here we have merely tried to sketch a minimal hypothesis. What matters is not whether it is true or not, only whether or not it is possible. What does it reveal? Contrarily to what we may have expected, the cognitive prerequisites of the hierarchical model can be very light, lighter, in any case, than those assumed by a (structuralist) model in which facts and values are treated separately. Indeed, while the latter requires a set of distinct operations to establish a relation between elements, to assign these elements or relations to contexts, and to grant values to these elements and contexts, the hierarchical model integrates values from the start and thereby collapses these three operations into one. This simplicity speaks strongly in its favor. But let us consider the other implications of our hypothesis.

It suggests that it is possible, even likely, that the mental processes involved in the elaboration of a hierarchical system are qualitatively different from the dynamic that animates the system itself. Thus, according to our speculations, the hierarchical relation of complementarity stems indirectly from a relation of contradiction, and the “whole,” far from preexisting, actually results from a certain treatment of the elements that eventually come to compose it. Our tentative hypothesis also proposes that hierarchical value difference derives indirectly from a preceding, nonhierarchical discrimination. In this manner, the hierarchical model allows us to envision in a new way the connection between the ideology of a society and the individual minds of its members, not as a direct relationship in which the structure of one reflects the functioning of the other, but as an indirect relationship between two radically different forms of organization. This model suggests in short that the relation between these two planes is discontinuous and therefore complex: that the passage from particularized functioning (individual thought) to overall construction (a collective system of representations) is not based on an extrapolation of elementary principles, but on the emergence of a new synthetic logic. But let us describe further the divergence between these two levels.

The ideological system’s structure as described by the hierarchical model corresponds to a global integration of the various contradictory discriminations made by individual actors. Now, as we have already said, there is no reason to assume that this structure exists as such in the minds of these actors. In this sense, the hierarchical organization of an ideology represents a virtuality of individual thought, not its actual functioning. Individuals have in mind, more or less systematically, a multitude of contradictory discriminations, not their overall integration into a hierarchical ideological system. The latter, as potentially present in the minds of individuals, can of course find expression in the constructions of observers, and in the words of the actors’ themselves. However, it remains that the elaboration of such “indigenous models,” of such systematic ventures in self-interpretation, are not at all necessary for individuals to participate effectively in their culture. An analytical appraisal of cultural activities may demand a hierarchical perspective; but the activities themselves do not depend on it; while a hierarchical outlook may be necessary for the analysis of systems of representations, it is not required to inhabit them.7

[267]One of the limitations of the (French) structuralist model based on the principle of binary opposition is that it is sufficiently simple to allow unwarranted slips between mental processes and systems of representation, thereby eluding what distinguishes these two very different registers. While the first refers to a universal mode of individual functioning, the second refers to an overarching construction that varies from one culture to the next. From this point of view, the hierarchical model has the advantage of being sufficiently complex to dissuade from such illicit confusions, to effectively prevent them or at least greatly limit them. It is indeed easier to imagine that individuals have in mind binary oppositions and their transformations, than configurations of hierarchically ordered levels. By distinguishing clearly these two (cognitive and ideological) registers, the hierarchical model poses the problem of the relationship that links them. In doing so, it raises fundamental, unavoidable, yet all-too-neglected questions pertaining to the mental integration of cultural systems, the relations between individual minds and ideology.


This article considers the principle of hierarchy proposed by Louis Dumont from a particular angle, as the foundation for a general model of ideological systems. Once outlined, this model has been briefly evaluated in the way it treats cultural representations, in how it relates to empirical facts, and with respect to the type of mental processing it implies. To conclude, we should note that although this principle remains marginal in anthropology, this is less so in other disciplines. More and more, in fields as diverse as analytical philosophy, physics, psychology, biology, or cybernetics, do we find a growing concern with the mechanics of paradox, part-whole relationships, encompassment, an interplay of levels, holistic integration, and so forth. From this point of view, the hierarchical principle partakes in a general movement that seeks to elaborate complex models. This is why it is worthy of attention: as a possible and properly anthropological approach to this new and far-reaching concern.


Barraud, Cecile. 1979. Tanabar-Evav: Une société de maisons tournée vers le large. Paris: Éditions de la Maison des Sciences de l’Homme.

[268]Barth, Fredrik. 1981. “Process and form in social life.” London: Routledge & Kegan Paul.

Dumont, Louis 1971. “On putative hierarchy and some allergies to it.” Contributions to Indian Sociology 5: 58–81.

———. 1979. “The anthropological community and ideology.” Social Science Information 18 (6): 785–817.

———. 1980. “Postface: Toward a theory of hierarchy.” In Homo Hierarchicus: The caste system and its implications, 239–46. Chicago: University of Chicago Press.

———. (1980) 2013. “On value.” Hau: Journal of Ethnographic Theory 3 (1): 287–315.

Evans-Pritchard, E. E. 1951. Kinship and marriage among the Nuer. Oxford: Oxford University Press.

Galey, Jean-Claude. 1983. “Creditors, kings and death: Determination and implications of bondage in Tehri-Garhwal.” In Debts and debtors, edited by Charles Malamoud, 67–124. Delhi: Vikas.

Iteanu, André. 1983. La ronde des échanges: De la circulation aux valeurs chez les Orokaiwa. Paris: Éditions de la Maison des Sciences de l’Homme.

Kelly, Raymond. 1977. Etoro social structure: A study in structural contradiction. Ann Arbor: University of Michigan Press.

Madan, T. N., ed. 1971. “On the nature of caste in India. A review symposium on Louis Dumont’s Homo hierarchicus.” Contributions to Indian Sociology 5: 1–81.

Needham, Rodney, ed. 1973. Right and left: Essays on dual symbolic classification. Chicago: University of Chicago Press.

Tcherkézoff, Serge. 1987. Dual classification reconsidered: Nyamwezi sacred kingship and other examples. Cambridge: Cambridge University Press.

La relation hiérarchique: Idéologie particulière ou modèle général?

Résumé : La relation hiérarchique telle qu’elle fut pensée par Louis Dumont attend toujours de recevoir l’attention qu’elle mérite. En s’appuyant sur des passages de l’oeuvre de L. Dumont, ce texte se propose de présenter le modèle hiérarchique de façon systématique. Clarifiant d’abord certains problèmes de définition (les notions d’englobement, de valeur, de niveau, d’ideologie, etc.), cet article souligne l’originalité théorique du modèle et considère certaines questions qu’il soulève: Quelles formes cette relation peut-elle prendre? Comment peut-elle être identifiée au niveau des faits? Quels processus mentaux présuppose-t-elle?

Michael HOUSEMAN, a French-trained anthropologist, is a Directeur d’Études at the École Pratique des Hautes Études in Paris (chair of African religions) and member of the African Worlds Institute (IMAf). Author of numerous articles on kinship, social organization and ritual practice, his publications include Naven or the other self: A relational approach to ritual action (Brill 1998, with Carlo Severi) and [269]Le rouge est le noir: Essais sur le rituel (PUM 2012). He is currently working on New Age and Contemporary Pagan ceremonial and ritual dance.

Michael Houseman
Institut des Mondes Africains (IMAf)
27 rue Paul Bert
94204 Ivry-sur Seine


Editors’ Note: This article is a translation of Houseman, Michael. 1984. “La relation hiérarchique: Idéologie particulière ou modèle général?” In Différences, valeurs, hiérarchie: Textes présentés à Louis Dumont, edited by Jean-Claude Galey, 299–318. Paris: Éditions de l’École des Hautes Études en Sciences Sociales. We are grateful to the publisher for permission to publish this translation. We are grateful to the publisher for permission to publish this translation. We are grateful to the publisher for permission to publish this translation and to Philippe Descola and Anne-Christine Taylor for suggesting this seminal article for translation in HAU

1. We may nevertheless mention the series of articles responding to the publication of the translation of Homo hierarchicus (Madan 1971). Unfortunately, these comments suffer from a confusion as to the meaning of the term “hierarchy” as it is used by Dumont. It is interpreted, in particular, as akin to social stratification. I shall come back to this point. Concerning the exploitation of the principle of hierarchy based on reliable and extensive ethnographic materials, it is more or less limited to recent works by the members of the Eramse (CNRS, France) and of the Groupe de Travail d’Anthropologie Sociale (Barraud 1979; Tcherkezoff 1987; Iteanu 1980; Galey 1983). I would like to thank the members of this Group for their comments on an early draft of this paper.

2. For a classical approach to this question, see Needham 1973; for a detailed reassessment of lateral symbolism in a hierarchical perspective, see Tcherkezoff 1983.

3. The problem of the number of levels or sub-systems that can be incorporated into a system of representations is worth raising here. Dumont tacitly refers to a configuration with only two levels: an opposition and its reversal. Besides, with the caste system in mind, and not excluding the possibility that there may be more levels, he cites “the sophism of Zeno concerning Achilles and the tortoise: the hierarchical disposition entails that the successive distinctions [of levels] possible are of rapidly decreasing global significance; in fact, as we know, Achilles catches up with the tortoise.” (Dumont 1979: 808). If we consider the problem somewhat differently, we can say that given that a reference to a totality is taken as constitutive of the relations between the elements it contains, and given that this reference is ultimately—in the case of an ideology’s cardinal value—the same for all of these relations, the question of the number of levels loses its significance and becomes essentially a formal, methodological consideration.

4. In this perspective, the importance of hierarchical opposition within the individual proper in modern ideology should not surprise us: mind/body, will/desire, reason/ emotion, etc. On this question, we may refer to current studies on the functional lateralization of cerebral hemispheres.

5. The case of the “Nuer paradox” (Evans-Pritchard 1951: 28) has become a classic example of such a divergence between practice and norm.

6. Such a complex model has yet to be established; I have only tried to show here how the principle of hierarchy may contribute to its elaboration. Nonetheless, I would like to mention two attempts which, in different ways, are in keeping with the propositions I have laid out: on the one hand, Raymond Kelly’s model of “structural contradiction” (1977) and, on the other, Fredrik Barth’s generative approach (1981).

7. I have suggested that at the level of thought, discrimination may be realized without referring to a totality. Such a (contextual) reference can be seen as the consequence of the simultaneous presence of contradictory discriminations, where such value “reversals” give rise to distinctions between contexts. Alternatively, we may assume that thought is contextual by nature, and that discriminations, which always occur in relation to a whole (a context), are therefore hierarchical from the start; they are by nature bidimensional, implying their own reversal. In both cases, however, the same problem remains: as long as we assume that the global structure of an ideology does not exist as a general map in the minds of every participant, the overall hierarchical ordering revealed by analysis can only be but a virtual one.