Abstract
Despite social change occurring over the 18th and 19th centuries in the Malagasy highlands, the same basic structure of the Merina circumcision ritual remained invariant, although some innovations and transformations have characterized its evolution. The issue is to render intelligible this co-existence of constant and shifting religious ritual forms throughout history. This case study argues that Lévi-Strauss’s ‘canonic formula’ can be reinterpreted as the law of permutation group organizing the whole series of Merina circumcision variants, thanks to relationality theories of ritual and to the structure mapping and multi-constraint theories of analogical reasoning. The canonic formula [Fx(a): Fy(b):: Fy(a): Fa−l(x)] formalizes a counter-intuitive proportional analogical thought and action inherent to the performance of Merina circumcision as a special agent ritual. Because this kind of analogy is based on core constraints, some of them structural and insensitive to the historical context, some of them being semantic and pragmatic and thereby depending upon cultural experiences and social situations, the permanence and variation of circumcision ritual form can be related to the hegemonic rise and decline of historical Merina polities.
The fundamental role of analogy in the making of culture has been emphasized by many anthropologists since Frazer’s work (Lévi-Strauss 1962; Bourdieu 1980; Tambiah 1985; Douglas 2001; Whitehouse 2004; Descola 2005). The issue I want to deal with concerns the extent to which analogy can render intelligible the performance of religious rituals. Contrary to the massive modularity hypothesis guiding the ‘new cognitive science of religion’ (Boyer 2001), I argue that analogy is the royal pathway to the articulation of cognitive and social sciences in the study of the religious domain. To go further on this line of reflection, I will go back to and rely on Bloch’s (1986) fascinating study of the Merina circumcision ritual. Despite social change over two centuries in the Malagasy highlands, the same basic structure of this ritual has prevailed on two levels: on the one hand, similar religious concepts have always been associated with its performance (for example, the notion of hasina – a holy power of growth – has remained significant); on the other hand, the same pattern of ritual actions has always been described (from the first account we have around 1790 until Bloch’s own fieldwork observations in the 1960s). 1 More important is that variable ritual actions and ideologies of hasina have been historically enacted in parallel to the expansion of definite highland Malagasy polities in the 18th and 19th centuries. My claim is that these variations and the persistence of Merina ritual forms and religious exegeses throughout history are both generated by the performance of a counter-intuitive proportional analogy: the canonic formula [Fx (a): Fy (b):: Fy (a): Fa−1 (x)] that Lévi-Strauss (1984) used to formalize the basic relational structure of such typical religious rituals and the permutation group within which these combinatorial variants are transformed into each other.
To proceed along these lines, I will first present the religious exegeses and pattern of ritual actions associated with the core of Merina circumcision. Second, I will argue that this invariance cannot be explained by the ritualization of human action (Humphrey and Laidlaw 1993), because what constitutes in fine this sequence of actions as a religious ritual is its relational structure established by analogical comparison and condensation between different units of actions, inside and outside the cult (Houseman and Severi 1998; Gell 1998). Consequently, I will propose to formalize this kind of analogical comparison under the prism of the Lévi-Straussian canonic formula that I will re-interpret thanks to Gentner’s structure mapping and Holyoak’s multi-constraint theories. I will suggest that this complex analogy is based on cognitive constraints, some of which are structural and insensitive to the historical context and others which are semantic and pragmatic and thereby dependent on social situations, cultural experiences, and political stakes (namely the hegemonic rise and decline of Merina polities).
The recurrent features of Merina circumcision
Merina are inhabitants of Imerina, a region of the central highlands of Madagascar. During the 18th century, this region became an area of dense settlement, in part because of the development of irrigated and terraced rice cultivation techniques (Berg 1981), and in part because of its position as the main distributive commercial centre of the island incorporated in the southwestern portion of the Swahili, Arabic, Indian, and Occidental maritime trading nexus (Campbell 2005). The Merina ethnonym never referred so much to a kind of homogeneous population as to a specific mode of government and subjection of intermingled populations, correlative of an historical mode of rooting in territory. Indeed, the term ‘Merina’ denoted membership, obedience, and loyalty to an imagined political community in which different levels of governance (fanjakana) on behalf of ancestors had organized means to control and move:
physical force (weapons, soldiers, dependants, relatives); groups of people (households [tokantrano], local extended families [fianakaviana], restricted ancestries [teraky], settlements [Tanana], demes [firenena], local assemblies fokonolona]); territories (ancestral lands and tombs [tanindrazana, fasandrazana], districts [toko], fiefdoms [menakely]); material resources (livestock, harvests, foodstuffs, houses, rice fields, manufactured goods, precious metals, money). Representatives of fanjakana choose the propitious day of performance during the cold and dry season, through the intermediary of an astrologer. The boy must be under the age of procreation. Representatives of fanjakana hide the circumciser from the view of everybody until the operation (this latter is affiliated to this polity and is called ‘the child’s father’). The real father and mother of the boy must avoid sexual intercourse during the ritual and for several days before and after it. Representatives of fanjakana at the residential level prepare the house(s) of circumcision:
They bring a special gourd before the ceremony; it becomes dry and looks like a phallic uterus after they hang it over the hearth. They empty the household from the hearth: mats, furniture and cooking utensils are removed from the house. They place prolific plants of living mother (reny velona) at the north east corner (the holiest part where one prays to the ancestors), such as rampant grass, reeds, and hasina (dracanae). Unmarried youths whose father and mother are still living (velona ray-amandreny) steal bananas and sugar-cane and bring them into the house at the northeast corner. Constituents of the deme gather and act in a united way: they all eat, drink, give food and money to the ritual organizer; they prepare their hair, ornaments and clothes; they dance, play music and sing together.
They make a collective procession in the build-up area with the boy and the banana plant, feasting in each other’s houses: the head of circumcision household gives the uncut silver coin
2
to the representative of fanjakana (act of manasina). Men of the deme communally cook together beef and rice outside the household. A group of young virgins (called ‘mothers of the child’), identified thanks to their distinctive hair-style representing the association of ancestries, take care of the child and dance with him during the ceremony. Unmarried youths tie up a powerful bull (omby mahery) after a violent bullfight, and they prepare it to be killed after the tsodrano ceremony. Elders of the deme (ray-amandreny) related to the child’s parents carry out a blessing ceremony (tsodrano, jôro, and kabary): when the sun is at the zenith, they invoke the deme’s ancestors and ask them by means of a formal speech to come and participate in the ceremony by blessing people and transmitting hasina through holy water (rano masina) placed in a bowl with uncut silver coins (ariary tsy vaky). The male elder from the father’s side blesses the child and the onlookers by blessing water and blowing it over them. Men of the deme try to steal the sugar-cane and the unripe banana placed in the house and they simulate a fight to take them away and eat them. Representatives of fanjakana prepare the gourd and the young male generation fetch the powerful water with it:
Unmarried youths burst into the circumcision house, hit the door violently with a pestle to overcome the mock resistance opposed to their entry, and then cross the threshold. Elders sing (father of a hundred, father of a thousand, of living mother, of living father) and bind the gourd by tying pieces of bark and grass around its neck. Representatives of fanjakana pierce the gourd with a spear. One of them slips an uncut silver coin or a silver chain made out of uncut coins into the gourd. A representative of fanjakana places the pierced gourd on the head of an unmarried youth acting as a bull in three stages: each time they incant ‘father of a hundred’ and the gourd is raised to the knee, then on to his shoulder/waist, then on to his head. Unmarried youths take the gourd away to running water (rano mahery). The leader of the procession threatens the gourd with a spear, while the young powerful ‘bull’ scoops up water with the gourd and carries it back to the house (before dawn, rushing water is thought to be pure and clean, unsullied by animal contamination). An elder takes a reed in the northeast corner and cuts it with a knife in three pieces while people whoop in the house: this action is called measuring the child (oha-jaza) because cutting is done first at his knee level, then at his waist level, then at his head level. The ‘mothers of the child’ stay with the real mother and her boy in the house, dancing in front of plants in the northeast corner and presenting them the child as a traditional gift, while the males of the deme go outside the house in the freezing night. Unmarried youths carrying the gourd assault the warm, woman focused-house and penetrate violently inside after a mock fight and a circumambulation around it to lay down the prepared gourd at the bottom of the plants in the northeast corner of the house. At dawn, the circumciser comes to the threshold to cut off the foreskin of the child, sitting on a rice mortar or a drum between his male relative’s legs: men stay outside of the house and form a semi-circle facing the door; women stay inside the house. A senior relative swallows the foreskin of the boy on a piece of banana. An elder of the deme pours the rano masina from the gourd on the boy and his penis to make it dry, and he washes his wounds, asking for blessings on the boy.
The basic polity was the deme (firenena), an endogamous descent group whose ancestral tombs were concentrated on irrigated rice valleys (the hierarchy of these tombs constituting its actual physical framework). Demes were segmentary, subdivided into residential kin groups (tokantrano, fianakaviana, teraky). They were the most historically relevant polities because rulers made a habit of negotiating political allegiance and autonomy with these rural communities as collective agents (Larson 2000: 34). In Merina polities, the ultimate legitimacy of fanjakana was based on monopolizing access and relationships to ancestors (razana). This involved control over the process of turning deceased people into ancestors and the capacity to establish or undo living people’s links to these ancestors (Bloch 1971; Graeber 2007). These ancestry relationships (karazana) were mainly established by funerals and circumcision. In the latter case, a single basic sequence of ritual actions can be described over two hundred years according to Bloch (1986):
From ritualized behaviour to religious ritual: The need for exegesis
Boyer and Liénard (2006) argue for the recurrence of ‘ritualization’ in this kind of religious ritual. This way of organizing the flow of actions is characterized by compulsion (feeling it is dangerous not to perform it), rigidity (following the detailed and prescribed liturgy), redundancy (repeating postures, gestures, and utterances), specific concerns (protecting against pollution and invisible threats, cleansing minds and bodies, purifying objects and sites), and goal-demotion (disconnecting actions from the intentions and motivations of its author, who has not entirely encoded its purpose). Boyer and Liénard give considerable data to show that ritualization is triggered by the activation of a cognitive module (‘hazard precaution system’), specialized for detecting potential danger clues such as predation, contamination, contagion, and social offence. At first appearance, it is true that purification and protection issues are predominant among immediate justifications for Merina circumcision: one must perform this ritual to avoid ancestors’ wrath, the child’s misfortune, and the spoiling of the boy’s destiny, but also to make this young male clean and pure, dry, stable, colder and fresher. Otherwise, one has to concede that the cost of non-performance is explicitly known by everybody: only a circumcised male can get married, participate in rituals, and be allowed to enter into ancestral tombs. This is a requisite for a man to pass down to his own descendants property rights over lands, houses, livestock, and ranks. It is furthermore the case that people circumcise their child because their parents and forebears did so (this is an ancestral custom to be imitated). Moreover, there are certainly no intrinsic and causal links between making the boy dry and pouring holy water from the gourd on him.
Nonetheless, serious problems go hand-in-hand with this modularity hypothesis. The principal stumbling block is that ritualization is not sufficient for the performance of Merina circumcision. Some ritualized behaviours may appear ironically as its most flexible and shifting parts; and when they are important and recurrent, they may not be proper and unique to this circumcision ritual: the detailed method for filling the gourd up and bringing it back home [11f] or the exact number of days or times for abstinence and circumambulations change considerably [3–14]; sometimes the boy’s foreskin is not swallowed but buried in a specific shaded place [16]. In addition, numerous sequences such as pulling a sheep corpse to pieces with bare hands appear and disappear without major consequences on the unfolding ceremony. And last but not least, specific ritualized behaviours can take place in other Merina rituals; they can even be common to those of the islands of Madagascar and the Comoros (think for instance of the jôro sequence [9]). In either case, what makes a Merina circumcision look like a religious ritual (and vice-versa) doesn’t rely uniquely on the performance of these ritualized behaviours: as Houseman and Severi (1998: 182) stated it, ‘in themselves, these procedures are unable to account for such events as specific totalities in the sense of Van Gennep’s internally structured, socially recognized ceremonial wholes’.
There seem to be two main reasons for this. First, what constitutes in fine a sequence of actions as a religious ritual is the discrepancy between its socially declared objectives and its internal structure: on the one hand, the ultimate goal and expected results are explicitly to create a fictional world based on imagined qualities (roles, status, cosmological features); on the other hand, the means utilized resort to the intervention of a ‘non-physical imagined’ (NPI) agent (the divine) or force (the occult) in a whole and complex ceremonial totality, including action chunks without goal ascription and possible alternatives (i). Second, what characterizes a religious ritual is that these fictional worlds and NPI agents/forces are displayed through specific systems of interactions which refer to everyday relationships and units of actions existing outside the cult and inside the ritual system of a religious tradition (ii).
Religious rituals as counterintuitive macro-actions
Religious rituals are an orderly sequencing of actions, some of them ritualized, some of them not, which are coordinated in demarcated time and space and whose relations are structured according to specific purposes which transform them into a counterintuitive macro-action (such as healing, burying, etc.). This is a macro-action in the sense that this set of interrelated acts holds its framing from the significant themes and overall project constitutive of its unity and totality as a self-referencing system. And this is a counterintuitive action because its project is connected to reasons and motivations while a subset of actions are not connected to these purposes in the same way as sub-actions are connected to sub-goals in ordinary behaviour. It is striking that the main term Merina people use to name the circumcision ritual as a whole is a metaphorical expression which literally means the ‘child’s umbilical cord cutting’ (fandidiana foitran-jaza). This expression explicitly connotes the idea of delivery and birth. If the removal of the boy’s foreskin (famorana, another label) is the technical purpose of this ritual, we find another set of explicit representations rationalizing its performance: people say that a circumcised boy will receive the holy power of growth (hasina) through blessing by the holy water in the gourd [17] and will be further able to transmit this power of obtaining prosperity (wealth, houses), fertility (good crops, herds), vitality (well-being), and fecundity (offspring). Moreover, he will become a sexually potent fertile male (lahy) and he will be recognized both as his father’s son and as a full member of the polity once this ritual is performed, being then allowed to participate in the generalized exchange of hasina as both taker and giver. This is precisely how Merina people conceive the circumcision ritual as a blessing ceremony during which the second birth of young boys confers on them the monopolistic ability to pick up, transfer, and canalize hasina on behalf of ancestors (Bloch 1986).
People expect thus from their ritual accomplishment that constituents of an imagined community (a Merina polity) will act towards the young boy in terms of an imagined role and status (a constituent and descendant able to receive and transmit hasina). But if it is possible for an ethnographer to understand these exegeses, it remains impossible for him and his informants to make justifications for each particular action considered from the point of view of their sequencing and association to the explicit finality of this ritual. Put another way, there is no empirical and direct link between transmitting hasina and removing the young boy’s foreskin, let alone the connection between these aims and the fact that specific categories of people hide the circumciser until the operation [2], avoid sexual intercourse [3–7], prepare the house of circumcision [4], violently steal and consume bananas and sugar-cane [4d–10], tie up and kill a bull [8], blow water [9], prepare the gourd and fetch water [4a–11], measure the child [12], cook, dance, sing, dress, drink and eat together [5–6], defend, penetrate into or escape from the house [10–13–14–15], pour powerful water on the penis [17], and so on.
According to the ritual form hypothesis defended by McCauley and Lawson (2002: 20), this link can be inferred because the representation of religious ritual relies on the same cognitive system deployed for the representation of action in general (the theory of mind modules), except that religious rituals involve a minima the insertion of a NPI agent/force in the agent or patient positions of represented action to guarantee in fine its efficiency and imagined impact on reality. Ritual’s structural description depicts thus a basic action representation system including different place-holders (an agent doing an act to something/someone [patient] potentially thanks to something [instrument]). The main differentiation, however, that marks it off, not only from everyday actions but also from the other sorts of routine religious actions (praying or singing), is the imagined involvement of NPI agents/forces in its performance and efficacy. In the Merina case, the representatives of fanjakana succeed in circumcising a boy to the extent of appearing to act on behalf of the divine (razana) and the occult (hasina). In participants’ representations of circumcision, ancestors are ultimately responsible for the child’s change in his reproductive role and imaginary status (his ability to channel hasina). Indeed, the notion of tsodrano consists in canalizing the holy power of the ancestors via the intermediary of the persons blowing, and the gourd pouring, holy water [9–17]: the main idea is that an undifferentiated group of ancestors transmit hasina to their descendants (Bloch 1986: 40, 51, 74, 88, 116). These latter take on different imagined roles and statuses; some act in place of ancestors, speak to them, behave as their indexes, and enact ritualized behaviours certifying publicly that they are connected to them [e.g. 8–9–17]. The fact that these ancestors are supposed to act upon humans and entities of the world, that they ‘are in on the action’ (McCauley and Lawson 2002: 33), thus facilitates the inferred inscription of the child in the graduated and hierarchical intergenerational chain of hasina circulation. This renders plausible why circumcision ritual brings about some change in the world in spite of the non-perceptible relation between means and ends, actions and sub-actions.
In short, up to this point, the representation of religious ritual can be derived from the intuitive ‘theory of mind’, which usually operates in ordinary and non-religious contexts. One way to formalize this point is to extend and marginally modify Gell’s theory of art (see Figure 1), by featuring the set of hierarchical embedding of agent-patient relations through which the behaviour of human/non-human indexes for recipients permits the abduction of divine/occult agencies and the attribution of prototypical imagined qualities to the intended targets of ritual-mediated agency. This consists in modelling the binary relation between these four components (H-NH index, recipient, divine-occult agency, and prototypical imagined qualities) in ‘special agent rituals’,
3
according to the notation proposed by Gell (1998), the agency arrows figuring the polarity of the agent/patient relation (A exercising agency over B).
4
Modelization of special agent rituals.
Religious rituals as relational patterns
If the overall ritual’s structural description is that ‘representatives of fanjakana’ (human indexes) ‘cut the boy’s umbilical cord’ (recipient) on behalf of ancestors (NPI agent/force), thanks to the sacrificed bull, the powerful water, and the prepared gourd (non-human indexes), it still remains that these NPI agents, indexes, and recipients appear and are recognized as such in the relational context of the ritual performance, as these prototypical imagined qualities are publicly manifested between people in the course of their interaction and expected mutual relationships. According to Houseman and Severi (1998: 208), this is what properly constitutes the ‘work of ritual’ and the ‘efficacy of its performance’.
The circumcision form consists roughly in choosing and announcing the propitious day, making invisible the circumciser, preparing the house, gathering and acting in a united way, making a manasina procession, tying up the bull, carrying out a blessing ceremony, stealing bananas and sugar-cane, preparing the gourd, fetching the water, measuring the child, segregating men and women outside and inside the house, mock fighting and assaulting the women’s house, cutting off and swallowing the foreskin, pouring the holy water and blessing the boy. This orderly sequence may be divided into sub-actions and ritualized behaviours which form a part of this relational configuration. On the one hand, the participants interact with each other on the basis of different ascribed roles (elders with constituents of the deme and the boy [9–12–17], mothers of the child with the boy and unmarried youths [7–13–14], representatives of the fanjakana with parents and unmarried youths [1–4–5–11], people with plants, water, house, bull, etc.). On the other hand, these sequences are articulated with each other while no intrinsic and logical links may be perceptible between encompassing goals and intermediary purposes: bringing plants into the house [4c] is a preliminary step for measuring the child [12] or stealing them [10], as one can fetch water [11e–f] only when the gourd has been pierced [11c], but no direct connections appear between them and the channelling of hasina by the prepuce removal.
Let’s focus on the first point, namely the nature of ascribed imagined roles. They are mostly contradictory identifications making protagonists behave as if they were simultaneously participating in two opposite and antithetical relationships usually separated outside the cult among different mundane activities and regroupings. 5 For instance, unmarried youth are depicted as constituent of both the ancestral polity and the bull in relation to the latter’s representatives [11]; the virgin women with their ancestral hair-style act as assaulted women in relation to unmarried youths [14], or behave as the ‘mothers of the child’ in relation to the boy [7–13]. Local exegeses are necessary to understand what are the main opposite relationships and imagined qualities which are combined along their association and conjunction in this ritual form. Bloch so finely restored the way Merina people contrast and compare ‘reny velona’ and ‘karazana’ concepts that we can further specify and re-qualify the tenor of these relationships and qualities.
The relation of ‘living mother’ (reny velona) between entities such as dracaena marginata, sugar-cane, banana, gourd, reed, rampant grass or human beings means the transmission of vitality (hery) and common essence along the uterine line (by parturition, cloning, vegetative propagation or plant cutting). Procreation is there assimilated to a paradigmatic and symmetric relation of substitutability and equivalence within each differentiated and divided uterine line. By contrast, the relation of ‘ancestry’ (karazana) between entities such as human persons and ancestors means the transmission of vitality (here called hasina) and common essence along undifferentiated descent line (through blessing, a sort of parthenogenesis). Procreation is there assimilated to a syntagmatic and complementary relation of contiguity and positioning within a part-whole set of undifferentiated and united lines. In central Madagascar, while the former is meant to prohibit sexuality between men and women who are related through their mothers, sisters, and matrilateral parallel cousins, the latter conversely prescribes sexuality between men and women who are related through their ancestors’ dry remains (in-marriages are thus expected within the descent group). Therefore, the imagined quality of reny velona is first and foremost instantiated in exogamous kinship groups that are networks of close relatives categorized as incestuous kinsmen: persons of one womb (kibo araiky), matrilateral parallel cousins (zanak’olompirahavavy), siblings (mpiray tam-po), and uterine half-siblings (mpiray-kibo). On the contrary, the imagined quality of karazana first and foremost stands for two main kinds of endogamous kinship groups that are segmentary and interlocked into the overall deme: the teraky, a three-generational level descent group of siblings and cousins managing ancestral tombs and lands (Vogel 1982); and the fianakaviana, an extended family living in the same settlement (tanana) and whose composed nuclei are the teraky and the household (Bloch 1971). Inside the deme occur the quadruple free choices that perpetually recompose the social morphology of descent groups: marital residence (regardless of the virilocal inflexion making wife-takers superior to wife-givers), filiation (in spite of the patrilineal bias for inheritance – anaran-dray), matrimonial alliance (even if the preferential marriage with patrilateral cousins or matrilateral cross-cousins assures that assets are not dispersed among kin – lova tsy mifindra), and burial place (according to the personal involvement in mortuary rituals and tombs maintenance). This everyday opposition in Merina kinship relations between uterine and descent links throws light on the ‘work of ritual’ proper to the circumcision form when, for instance, unmarried youths or virgin women behave as mahery and masina entities at the same time: this work consists essentially in establishing new linkages between reny velona and karazana relationships around the circumcised boy and beyond the limits of the elementary exogamous kinship group.
If we focus now on the second point (relations between orderly sequences), the issue concerns the nature of actions that condense these new linkages and bring together these contradictory identifications on the basis of pre-existing ties and acts drawn from a variety of mundane and religious domains. Here, the proposal is that these actions perform analogical comparisons which provide Merina circumcision with a relational form integrating both logical articulations and disconnections.
As regards the religious domain, there are similarities between different sequences common to the annual royal bath and funerals (for example, the tsodrano duty or the unmarried youths’ tasks of carrying the water and tying up the bull). More fundamental is the comparable act [5a–9] of presenting traditional gifts to representatives of fanjakana (‘manasina’, making something masina) in exchange for the latter’s blessings. Numerous redundant sequences derive thus from the act of ‘manasina’: the ‘mothers of the child’ present the boy to the northeast corner of the house [13], the ‘unmarried youths’ present the plants [4d], the bull [8], and the prepared gourd [14] to the ancestors. More generally, this act of manasina consists in paying homage to ancestors via a living representative, by giving them material goods, labour service, or uncut silver coins as tokens of loyalty and obedience, while these living intermediaries bestow blessings and transmit hasina upon those who honour them. Manasina takes place between people of unequal status every time there are rituals, transactions, and agreements regarding adoption, mortgage, land property allocation, slave sales, taxations, or tributes. As Berg (1988: 204) articulates, members of household present hasina to heads of households, who in turn perform manasina to heads of descent groups, and heads of descent groups give hasina to heads of other descent groups of superior status, local assemblies of descent groups eventually giving hasina to their sovereign. These traditional presentations occur many times a year and define a social hierarchy in which all individuals and descent groups find their places.
If we turn in the same vein towards what’s happening outside the cult, we face an important set of explicit representations about similarities and differences between mundane actions and circumcision sequences. More particularly, procreation and birth have their counterparts in ritual actions performed along the successive steps of the encompassing ceremony. Merina people oppose the propitious day chosen by an astrologer and the representatives of fanjakana [1], to the unpredictable coming of the new infant who may be born on a bad day and therefore suffer a bad fate (vintana). They compare the gourd not pierced yet (voatavo tsy mifandraka) to a place where the foetus is supposed to grow, and they deal with the prepared gourd as if it were ‘heavy’ (mavesatra), a word which is used to mean ‘pregnancy’ in everyday life. Moreover, it is taboo (fady) for pregnant women to break a gourd, to sit on the threshold or on a rice mortar (places where the child is circumcised), and to enter into the house of circumcision. Bloch (1986) outlines how people laugh at the evocation of sequences [11a] and [14], as if they were talking about a sexual penetration, while the parents and the young virgin women are taboo to perform any sexual intercourse [3]. Besides, people put side by side the little warm house made of mats (komby) which is built inside the household for the period of labour in pregnancy, and the preparation of the circumcision house clear-out of the warm hearth [4]. They contrast the united men cooking outside [6] with the divided women usually cooking each at their own hearth; they parallel the united deme [5] with the quite isolated woman delivering at home. The young virgin women are called the mothers of the child [7] as the midwife is during the labour (renin-jaza). People match up the way the gourd is placed on the head of the young bull in three stages [11d], to the way the baby goes out from his mother in three stages (first the head, then shoulders, then knees). This comparison is extended to the way the umbilical cord is measured and cut in three pieces at knee, waist and head level when a woman gives birth to a child, and to the way elders measure the child and cut the reeds during the circumcision [12]; people inside the house incidentally whoop as the newborn baby does when it’s delivered. Besides, people relate how the prepuce is swallowed [16] to how the placenta is buried next to the house after the birth. The way young mothers bathe and clean their babies with hot water (mifana) clearly stood in opposition for them to the manner in which the child is washed with cold water under the authority of elders [17]. In sum, most Merina adults are explicitly able to contrast the transmission of vitality (hery) to the infant from the female womb, to the transmission of vitality (hasina) to the boy from the tomb and the androgynous prepared gourd. These different exegeses consist of strong comparisons established between actions and relationships relating to child delivery and those relating to circumcision ritual, insomuch that the latter’s structural description seems to be partially analogous to the basic action representation system of maternal birth.
Analogical reasoning as a situated cognitive process
Holyoak defines analogy as a form of inductive reasoning that applies between different situations (called analogs), in which what is known about one analog (source) is used to infer new information (inferences) about the other analog (target) (see Figure 2).
Major components of analogical reasoning (Holyoak 2005).
In the Merina case, a familiar situation such as ‘women give birth to children’ (the source description) is matched with a less understood situation about how ‘representatives of fanjakana cut the boys’ umbilical cord on behalf of ancestors’ (the target description). The target situation serves thus as a retrieval cue for a potentially useful source analog (‘circumcision is like delivery and birth?’). This thinking implies that a mapping-finding is processed to establish a set of systematic and one-to-one correspondences aligning the elements of the source (B) and the target (A). The analogy can then lead to different kinds of learning and conceptual change in one or both analogs (Gentner et al. 2001): first, the projection of candidate inferences on the target (‘indexes of fanjakana/ancestors give birth to children?’); second, the generalization of explicit relational categorization yielding a more abstract schema for a class of situations of which the source (B) and target (A) are both instances: miteraka, or the relation of begetting by transmission of vitality and common essence; and third, the re-representation of the constituent elements of the analogs, each got rid of something to better match: the sexual role of the male genitor for example. Analogical reasoning requires some representation of the situation and its constituent elements being compared for comparison to take place. By convention (Gentner et al. 2001), the representation-building of situations as agent-patient relations in computer programs distinguishes between entities (place-holders such as agent or patient), attributes (entities’ sensible and imagined qualities), first-order relations (relations between two or more entities on the basis of their attributes), and high-order relations (relations between relations encoding causality, implication, similarity, and so on). But as suggested above, the shaping of religious situations into appropriate representations that are available for analogical reasoning as inputs can be considered as quite a separate issue because of their partial generation by the core cognition of agency.
The intuitive rules of analogy
If analogical reasoning involves four main cognitive sub-processes (retrieval, mapping, transfer, and learning), its entire process is governed according three core constraints (semantic similarity, structural isomorphism, and pragmatic centrality) applying in all of its components (Holyoak and Thagard 1997). The first one is the required existence of semantic similarity between the situations compared. Correspondences are established between analogues to the degree that their entities, attributes, and relations share common features and overlap in meaning (Holyoak and Thagard 1990). For instance, women and the representatives of fanjakana are similar because they are both human beings (superordinate-hyperonym), boys are also children (synecdoche), young virgin women and midwives are called the mothers of the child (synonym), the prepared gourd is treated as if it were pregnant (metaphor), mothers bathe and clean their babies with hot and not cold water, and men of the deme cook together outside the house and not separately inside (antonyms). The difficulty is that these correspondences presuppose semantic networks in which perceptual and conceptual categories are embedded, chained, contrasted, and sometimes organized into hierarchical taxonomies whose properties are still prone to debate between the classical and prototype/basic level theories of categorization. Nevertheless, the fundamental role of bodily and cultural experiences in the organization and representation of knowledge has been noted by Lakoff (1987: 99), who argues that culture-specific experiential domains can characterize links in category chains through providing binary oppositions and assigning distinct mutually-exclusive categories.
In Imerina, we find therefore an underlying cosmology that classifies entities according to their position in the hierarchical nomenclature of their sensible and imagined qualities: the more these entities are pure and clean (madio) and not dirty, impure (mandoto); dry (maina) and not wet (lena); cold, fresh and stable (hatsiaka) and not hot, dangerous, unstable (mafana); clear (mazava) and not dark (hizina); hard (mafy) and not soft (malemy); the more they are timeless, homogeneous and transcend the life cycle, the more they are holy (masina) and are prevented from decaying, the more they are placed at the top of the social hierarchy which is materialized in the cardinal space (the east for the sunrise, the top for the sky, the north for the optimum sunshine) and in the ecological environment (the top of mountains). Semantic similarity of the elements that are in the same way involved in delivery and circumcision relies partially on this overall classificatory scheme: bulls, cutting plants, flesh, komby, babies, women and ancient native people (vazimba) are all categorized as reny velona entities (common superordinate), being therefore associated with heat, humidity, instability, southwest direction, and the untamed power of growth (hery), whereas circumcised men, rice, bones, tombs, elders, and ancestors are referenced to cold, dryness, stability, northeast direction, and controlled power of growth (hasina).
Nonetheless, analogy is not literal similarity because perception of like relational patterns across different situations is more important than placing in correspondence identical entities and attributes (Gentner and Markman 1997). That’s why the second core constraint (structural isomorphism) has been described as the most fundamental rule of analogy (Gentner 1983). Parallel connectivity compels the pairs of corresponding attributes and relations to match their respective entities even if these are dissimilar. This mapping-finding occurs at different levels: first, among entities’ attributes (one role filler: e.g. the gourd is mavesatra as woman is mavesatra); second, among first-order relations (two roles filler: e.g. the male elder measures the boy as the midwife measures the child; the uncut silver coin is immersed into the gourd as the sperm is immersed into the womb); and finally, among high-order relations that take at least one first-relation as a role filler (the orderly sequencing of ritual actions for circumcising a boy is logically articulated like the necessary successive steps for giving birth to a child are logically implicated by each other: choosing/undergoing the D-day [1], impregnating the gourd/womb [3–11–14], preparing the house/komby [4], cooking [6], giving birth and measuring the child/boy [12–15], treating the prepuce/placenta [16], bathing and cleaning the child/boy [17]). This presence of a higher-order connective structure between the lower-order relations constitutes the ‘intuitive preference for systematicity’ (Holyoak and Thagard 1989): deep and interconnected systems of relations such as causal, mathematical, or functional relations influence both analogs retrieval and elements selection by constraining mappings between first-order relations and entities. This is the way Merina circumcision and parturition can also be represented as two sets of hierarchical embedding of agent-patient relations, whose overall structures have respectively the following form: Fx (a->m) and Fy (b->n). In this expression, the relation Fx (=circumcising on behalf of ancestors) holds between the agent (a = representatives of fanjakana) and the patient (m = boys), just as the relation Fy (=giving birth) holds between the agent (b = women) and the patient (n = children). Producing an alignment of their relational structure involves matching all the elements of the sub-sequences on the basis of their semantic similarity, and to coalesce these local matches into structurally consistent connected clusters which will be merged into mappings between ‘Fx’, ‘Fy’, ‘a’, ‘b’, ‘m’ and ‘n’ (Holyoak and Thagard 1990).
The last core constraint (pragmatic centrality) highlights how explicit or implicit knowledge about plans and purposes the analogy is intended to serve has a strong influence on its processing (Holyoak and Thagard 1989). Some commonalities and central elements can be so anticipated and expected that specific retrieval from long-term memory and mapping in working memory should be found for them. This stresses the decisive role of the specific vocabulary related to Merina circumcision as well as the importance of exegeses concerning its aims and awaited results: remember that this ritual is named the ‘cutting of the child’s umbilical cord’, that its main protagonists are called ‘of living mother’, ‘mothers of the child’, ‘child’s father’, ‘of living parents’, and that people are motivated to perform it in order to ensure the boy’s transformation into a fertile member of the descent group inscribed in the generalized exchange of hasina. Moreover, the father’s boy is publicly and officially recognized after he has paid homage (manasina) and organized a ritual circumcision for his son. Filiations are thus made effective, even in the case of fostered boys. All of these are so many clues that the situation when ‘representatives of fanjakana circumcise the boys on behalf of ancestors’ [Fx (a->m)] should actually be like the situation when ‘women give birth to children’ [Fy (b->n)].
The work of ritual as counter-intuitive analogy
The analogical thought and action which subsumes the ritual circumcision as an overall relational structure is more complex than a simple comparison with maternal parturition because of the ‘cross-mappings’ and ‘alignable differences’ occurring between these two situations (Gentner and Markman 1997). On the one hand, similar entities play different relational roles at the same time they embody contrasted and hierarchically opposed mahery and masina qualities (e.g. the women as female, mothers, mothers of the child, elders, or constituents of the deme). On the other hand, different entities play similar relational roles at the same time these latter imply opposite or hierarchically-embedded commonalities (e.g. the midwife/elders measuring the child/boy by cutting umbilical cord/reed; the mothers/elders bathing and cleaning the newborn/boy with hot/cold water). In both cases, the contrasted opposition between reny velona/hery and karazana/hasina features is put in correspondence as if these relations and qualities were similar at the same time that the latter were opposite and superior to the former because of partially negating them while including them. That’s why the basic analogical comparison between [Fx (a->m)] and [Fy (b->n)] doesn’t actually constitute ‘the work of ritual’ and the ‘efficacy of its performance’ because it does not depict this enacted transition from reny velona/hery to karazana/hasina features. More precisely, this matching does not embrace the sequence set of ritual actions that performs this transition by bringing together the contradictory identifications and condensed relationships of reny velona and karazana status while enacting the superiority of the hasina transmission of vitality: namely, the act of paying homage in exchange for instituting the filiation between the father and his (fostered) son [5a], the act of tying up and cutting the throat of a bull after a violent rodeo [8], the performance of a blessing ceremony [9], the act of aggressively stealing and eating cutting plants [4d–10], the one of binding and threatening the gourd to capture mahery water [11], the one of presenting the boy to the plants and ancestors [13], the one of violently assaulting the warm woman focused-house [14], and of course, that of cutting off the boy’s foreskin [15].
Here, the claim is that the work and efficacy of this ritual consist in the performance of a specific kind of proportional analogy [A is to B what C is to D], in which the similarity between the target A (ancestral circumcision, [Fx (a->m)]) and the source B (maternal delivery, [Fy (b->n)]) leads to a cognitive dissonance between the analogical transfer C (ancestral delivery, [Fy (a->n)]) and the source B (maternal delivery) that is resolved and transformed into a paradoxical synthesis (B implies C), on condition that the relation between A and B is like the relation between C and a fourth inferred and missing component D [F? (?->?)], this latter being derived from A in such a way that A implies D as B implies C. Along this four-part analogy, there are two kinds of dyadic relationships: first, comparisons of the similarity relations between A:B and C:D where the problem is to infer the missing component D that is related to C (ancestral delivery) in the same way B (maternal delivery) is related to A (ancestral circumcision); second, comparisons of the implication relations between A:D and B:C where the problem is to infer a missing common high-order relation, namely that of the whole to its parts.
At the first level (if ancestral circumcision goes to maternal delivery, what does ancestral delivery go to?), the challenge is to represent what makes possible the transformation of certain ritual sequences into the counterpart of successive steps in giving birth to a child. If the missing component D is arrived at by applying or transferring the relationship defined over a set of attributes of A and B (ancestral circumcision/maternal delivery) to the relevant attributes of C (ancestral delivery), then the intuitive path is to infer what the outcome would be if a similar transformation as the one that took A to B were applied to C: namely, the transition from the karazana transmission of vitality through hasina to the reny velona transmission of vitality through hery. We should then expect that the counterpart of C (ancestral delivery) would be a reny velona relation of transmission initiated by women. This is the position Mosko (1991) defended to analyse his ethnographic data among the Mekeo of Papua New Guinea when he revised the Lévi-Straussian canonic formula by deleting its double twist, and when he argued that Mekeo people linked their different conceptual categories resorting to ritual, kinship, economic exchange, bodily states, social classification, and political organization on the basis of a ‘recursively inverted dualism’ [Fx (a): Fy (b):: Fx (b): Fy (a)]. Nevertheless, I will put forward that this Merina case study underscores how the proportional analogy at stake is indeed counter-intuitive to the extent that A and B are not related according to the same path by which C and D are compared. The transition from hasina to hery transmission (A:B) has actually its correspondence in the transition from hasina transmission to hery predation (C:D): Merina ancestral delivery matches and goes to violent subjection of populations on behalf of ancestors, and the karazana relations (C) transform into condensed relationships of mixed karazana and reny velona status (D).
If we go back now to the second level (what kind of implication between A and D can be transposed to the contrasted opposition and logical contradiction between B and C?), the challenge is to represent a similar high-order relation respectively aligning maternal delivery (B) towards ancestral parturition (C), with ancestral circumcision (A) towards a missing component (D). As I just suggested, this latter component can be assimilated to a form of hery subjection/predation schema because of the implementation of two logical operations of opposition and inversion which will be discussed below. For the moment, let’s just consider that, on the basis of this transformational path, A implies de facto D to the extent that the way human and non-human indexes circumcise the boy implies that male constituents of the deme steal and eat cutting plants [4d–10], that unmarried youths violently sit astride and tie up a mahery bull [8], that representatives of fanjakana threaten, tie and pierce the gourd to capture mahery water [4a–11], that unmarried youths assault the warm woman-focused house [14], and that the ‘child’s father’ cuts the boy’s prepuce [15].
Typically, these ritual sequences are hierarchically subordinate agent-patient relations of the overall Merina circumcision’s structural description (A) which are not matched in the analogical comparison with B (maternal delivery). The implication link is thereby a metonymic relationship of the whole to its parts in which the performance of Merina circumcision is a necessary or sufficient condition for the occurrence of these ritual sub-sequences. And this is exactly the same inverted implication link (a metonymic relationship of the parts to the whole) which relates B and C, to the extent that maternal delivery (B) is conventionally associated in order to enhance procreation and maturation processes with institutional facts arising out of constitutive rules and specific ritual actions. On the one hand, the idea that indexes of ancestors give birth to children (C) is instituted through the necessary affiliation to a descent group and the obligation for its constituents to procreate and choose their marital partner within. On the other hand, this same idea is explicitly featured during cults performed on kings, chiefs, astrologers, or famous nobles’ tombs where women ask for blessings of fecundity; during the child’s first hair cutting ceremony (ala volo) where women swallow cut pieces with banana to enhance their fecundity; and last but not least, during circumcision sequences such as paying homage (manasina) in exchange of instituting the paternal filiation [5a], performing jôro [9], or else pouring the boy with the prepared gourd [17]. There is thus a tendency to systematic instituted co-occurrence in the birth domain between reny velona and karazana relations as if they were complementary.
In the end, this kind of proportional analogy does not strictly correspond with the intuitive rules of analogical reasoning because of minimally violating them. There is no absolute pragmatic centrality because the comparison of A:B with C:D doesn’t result totally from protagonists’ purposes, by being already encoded in the set of ritual actions and mundane practices related to each other through prefixed correlations and logical operations. There is neither strict semantic similarity between the relation A:B and the relation C:D, for the missing component D is not symmetrically inferred according to an inverted dualism. And if the structural isomorphism operates, the mapping-finding process is primarily pre-formatted to focus on similar high-order relation of implication between A:D and B:C to lead the comparison of A:B with C:D. What, then, about the hypothesis according to which this proportional analogy is counter-intuitive in the sense that its properties correspond to Levi-Strauss’s canonic formula?
At the crossroads of cognitive and social sciences: A new interpretation of the ‘canonic formula’
The canonic formula has an ambiguous status in Lévi-Strauss’s work. It rarely and marginally appears in his publications while it can be considered as the thread of his thinking upon religious and kinship domains (Scubla 1998; de Almeida 2011). If Lévi-Strauss initially defined its standard expression in 1955 as the ‘law of permutation group’ organizing a whole series of myth/ritual variants, 6 he did not explicitly apply this formula to empirical data for 20 years, before experimenting with different canonic expressions for Amerindians masks of the Pacific Northwest coast (Lévi-Strauss 1979), sacred kingship rituals and the Iatmul Naven ceremony (Lévi-Strauss 1984), Jivaro and Salish myth variants (Lévi-Strauss 1985, 1991), as well as architectural shapes of house roofs (the hourglass form) across Asia and America (Lévi-Strauss 2001). Consequently, its fate has been varied: ignored by the majority of scholars, disqualified by others as a ‘meaningless abracadabra’ (Leach) or a ‘personal mantra’ (Sperber), it has been used by some anthropologists as a powerful insight guiding their own theory of structural transformation (Maranda and Maranda 1971; Mosko 1985; Wagner 1986; Zempléni 1987; Viveiros de Castro 2009), while acclaimed by a few mathematicians as an ‘intelligent formula’ capable on a logical level of asserting a ‘Thomian morphodynamical modelling’ (Petitot 1988) or a ‘nontrivial anti-automorphism of the quaternion group of order eight’ (Morova 2005). This ambivalent status is easily understandable if we consider that Lévi-Strauss separately explored and dealt with the main logical operations constituting its analytical scope by elaborating and formalizing some of them in successive instantiations, and by setting apart some others for incompatibility with these latter (e.g. the temporary abandonment of the double twist in the Klein group exploration). The notion of ‘mediation’ is thus the first logical operation investigated in the analysis of the Oedipus myth and Zuni mythology (Lévi-Strauss 1958) or in the presentation of the culinary triangle – the raw, the cooked, and the rotten (Lévi-Strauss 1964). It comprises the replacement of a binary opposition between A and B with two subordinated isomorphic oppositions (for instance A/C and C/B or A′/A″ and B′/B″) that are comparable to each other on the basis of their common features with A and B. 7 This imaginary overcoming of contradiction can be also operated by transforming A into B through ‘relations of symmetry and inversion’. Lévi-Strauss (1964: 217) then put forward the notions of ‘chiasmus’ and ‘twist’ in the logical articulation of different Bororo and Ge myths versions, after he showed how the Pawnee myth of the pregnant boy had no counterpart inside the Pawnee ritual system but was symmetrically opposed and inverted to Mandan and Hidatsa age grade rituals (Lévi-Strauss 1958: 271).
If Lévi-Strauss (1966: 211) explicitly subsumed these logical operations under the canonic formula to formalize the passage of metaphors and exclusive paradigmatic sets to metonyms and compatible syntagmatic chains, he nonetheless preferred to rely on Boolean algebra for some time to model as a matrix the system of operations delimiting the relevant permutation group in the Amerindian mythology. It is therefore not surprising that several anthropologists reduced the canonic formula to a specific Boolean group for interpreting their ethnographic materials. Among them, Hage and Harary (1983) graphically represented as an n-cube the permutations of culinary and sexual Arapesh symbolism in respect to the ideology of marriage and kinship relations. And Desveaux (2001) used the Klein group as a way of featuring and bringing together the complex analogical relationships between myths, rituals, kinship nomenclatures, and social organization forms through the Amerindian continent. But Lévi-Strauss’s use of classic Aristotelian proportional analogy constitutes in fact one of the main foundations of the canonic formula beyond logical operations close to Boolean algebra. The savage mind is explicitly defined as ‘analogical thinking’ whose systems of classifications and taxonomies result from comparisons between similarities and differences of attributes/relations, but whose inferred predictions are not capable of empirical verification or falsification in terms of experimental observations. Lévi-Strauss (1962) deconstructed totemism by arguing that this cosmology was in fact the result of a human capacity for analogical reasoning and categorization asserting two kind of homological equivalences: first, an identification of kinship group members to specific plants or animals; second, an analogy between animal and vegetal species differences on the one hand and kinship groups’ significant contrasts on the other (the relation between species A and species B being analogous to the relation between kinship group C and kinship group D, in the sense that A differs from B on the basis of their physicality and ecological niches as C differs from D on the basis of their status and functions in the ritual division of labour).
More fundamentally, Lévi-Strauss (1962: 140–56) defended the proposition that there was a logic of conceptual change marking the passage from totemic groups to professional castes’ cosmologies. This logic resorted to the inverted and symmetrical analogical reasoning of the other, including the substitution of a component with another (the replacement of animal/vegetal species and ritual functions with manufactured objects and social functions). The proportional analogy between two relations of difference expressing the cosmology of totemic societies such as the Algonkin (species 1: species 2:: kinship group 1: kinship group 2) could transform into a analogy between two relations of similarity expressing the cosmology of caste societies such as the Natchez (manufactured object 1: professional caste 1:: manufactured object 2: professional caste 2), whilst differentiating into hybrid forms such as among the Chickasaw (species 1: professional caste 1:: species 2: professional caste 2). In all cases, these transformations could be at once correlated to the passage from exogamy to endogamy in matrimonial alliances, to the transfer from myth predominance to sacrificial ritual prevalence, and to the transition from an egalitarian ritual division of labour between totemic groups specialized in the reproduction of living species to a hierarchical social division of labour between professional castes specialized in the production of goods and services. Last but not least, these conceptual changes could be configured on the basis of a common analogy between sexual relations and eating (sexual/food prescriptions and prohibitions), at the same time as they could be parameterized under the pressure of non-cognitive factors such as technical progress, climatic change, demographic growth, population density, ecological degradation, and the history of the relations of power and production. In the following section, these are the fundamental insights and claims I want to develop and fit into an analysis of the Merina circumcision ritual.
The cognitive operations of the canonic formula
Lévi-Strauss employed different canonic expressions without exploring the overall set of its possible variations or defining explicitly the rules governing the selection and reduction of empirical data. Thereby, Mosko (1991: 137) could argue in favour of restricting its application for ‘non-myth’ to a ‘single socio-cultural tradition’. This idea is along the same lines as Lévi-Strauss’s use of Iatmul naven as a prototypical ritual among similar religious traditions such as Tupinamba cannibalism, Aztec sacrifice, Iroquois torture, or Jivaro head-hunting, which are all based on comparison between male capture of maternal procreation and male capture of enemies. For Lévi-Strauss (1984), the canonic formula [Fx (a): Fy (b):: Fx (b): Fb−1 (y)] was supposed to formalize the basic relational structure of this prototypical ritual by encompassing its variants set and the permutation group within which these combinatorial variants were transformed into each other. In the Merina case, there is also a central opposition between ancestral and maternal procreation along with a basic analogy between maternal delivery and ancestral circumcision, which can be superseded thanks to the mediation of these structuring oppositions along a specific transformational path of the following expression: [Fx (a): Fy (b):: Fy (a): Fa−1 (x)].
The point here is that the right-hand side of the formula is a transformation of the left side at the same time as it continues to be its equivalence: the induced comparison between the counter-intuitive action where ‘representatives of fanjakana circumcise the boys on behalf of ancestors’ [Fx (a->m)] and the intuitive action where ‘women give birth to children’ [Fy (b->n)] leads to an inferred counter-intuitive action where ‘representatives of fanjakana give birth to children’ [Fy (a->n)], whose antithetic opposition and cognitive dissonance with maternal delivery can be overcome by virtue of the double twist. The fourth component [Fa−1 (x->p)] is thus derived from two logical operations of opposition and inversion applied to the first component (A): the permutation of function and term values, and the replacement of a term by its opposite inverse.
Basically, the different canonical expressions require at least four properties:
first, the double equivalence between components A and B, C and D on the one hand; between the relation A:B and the relation C:D on the other hand, on the basis of an initial comparison between the two functions Fx and Fy and their respective terms (a) and (b); second, the exchange between the two initial functions (Fx and Fy) of their respective terms (a and b) in the third component (C); third, the replacement of one term (a or b) in the left-hand side by its opposite inverse in the right-hand side (a−1 or b−1);
8
fourth, the transformation of one term in the left-hand side (a or b) into a function (Fa or Fb) on the right side, and the transformation of one function in the left-hand side (Fx or Fy) into a term (x or y) on the right side.
9
The major difficulties for interpreting this formula are the meaning attributed to the functions, terms, and results of inversion operations, namely the last component D. Marcus (1997) argued for considering these symbols as an attribution relation of predicate to subject (i.e. the term ‘a’ possessing the property ‘x’; ‘a’ being an entity or a character, ‘x’ being a social elementary role). By reading the formula of Merina circumcision in terms of entities, attributes, and relations between entities as we suggested above, we can claim that Fx (a) is equivalent to saying that an entity ‘a’ has a predicate ‘Fx’ which can be an attribute or a relation, ‘a’ and ‘F(x)’ being both concepts. I propose to go further in this direction and to make a case for assimilating the terms ‘a’ and ‘b’ to entities capable of (inter-)acting, that is, taking the relational role of agency; and functions ‘Fx’ and ‘Fy’ to meaningful actions, that is, specific agent-patient relationships.
Consequently, the canonical expression [Fx (a->m): Fy (b->n):: Fy (a->n): Fa−1 (x->p)] means for each component that an agent ‘a’, ‘b’, or ‘x’ (an active entity) performs an action ‘Fx’, ‘Fy’, or ‘Fa−1’ upon a patient ‘m’, ‘n’ or ‘p’ (a passive entity). Additionally, it means that the relation between [Fx (a->m)] and [Fy (b->n)] is similar to the relation between [Fy (a->n)] and [Fa−1 (x->p)], to the extent that the transformation of Fx (a) into Fy (b) is similar to the transformation of Fy (a) into Fa−1 (x), because of relying on the same oppositions of agents and agent-patient relationships. But at the same time, the relation between the first and second component (A:B) is different from the relation between the third and fourth component (C:D), because of the latter mobilizing in its transformation an additional opposition of agent-patient relationship correlated to crucial changes in the world brought about by the ritual performance (namely, the apparition of a new agent, patient, and action: i.e. the double twist issue).
Let’s now break the canonic formula down into its constitutive operations and transformation rules, by translating into this theoretical framework its different components. First, we can say that transforming an action conducted by a specific agent into another opposite action consists in crossing their respective agents. For example, ‘women give birth to children’ becomes ‘representatives of fanjakana give birth to children’. Hence Transformation Rule 1: Rule 1: the permutation of ‘Fx (a)’ with ‘Fy (b)’ are either the Actions ‘Fx (b)’ or ‘Fy (a)’
Second, we can say that the inverse of an action is its patient becoming an agent (an entity capable of agency after this action has been performed upon it). For instance, the inverse of loving is a loved person as an agent (a lover). Hence Transformation Rule 2: Rule 2: the inverse of Actions ‘Fx’ and ‘Fy’ are respectively the Secondary Agents ‘x’ and ‘y’
Thirdly, we can say that the inverse of an agent is the action performed upon a patient to make it capable of acting as this agent. For instance, the inverse of a woman is the action performed upon a patient to make it capable of acting as a woman. Hence Transformation Rule 3: Rule 3: the inverse of Agents ‘a’ and ‘b’ are respectively the Actions ‘Fa’ and ‘Fb’
Lastly, we can say that the opposite inverse of an agent ‘a’ is the opposite of this agent as a patient. For instance, the opposite inverse of a woman acting up is a man acted upon. Hence Transformation Rule 4: Rule 4: the opposite inverse of Agents ‘a’ or ‘b’ are the opposite Patients ‘a−1’ or ‘b−1’
So, definitively, three main logical operations define the canonic formula and its double twist: the first one consists in exchanging in the left-hand side of the formula one action relative to ‘a’ or ‘b’ in the right-hand side. For instance, the representatives of fanjakana give birth to children instead of circumcising boys.
The second operation is to invert one initial action of the left-hand side components in the right-hand side. The inverse of the action ‘Fy’ or ‘Fx’ is the Secondary Agent ‘y’ or ‘x’ (rule 2). For example, the inverse of circumcising is a circumcised person capable of acting.
The third operation is to reverse the opposite patient to an initial agent of the left-hand side in the right-hand side situations. In other words, the opposite inverse of the agent ‘a’ or ‘b’ is the action performed upon a patient to make him capable of acting as the reverse of these agents; namely, the contrary of ‘a’ or ‘b’ as a patient (rule 4 + rule 3). For instance, the opposite inverse of a woman acting up is the action performed upon a patient to make him capable of acting as a man acted upon.
On this strict combinatorial logical plane, there are eight canonic formula versions (because of the two existing possibilities for each operation).
10
I will only focus on the canonic expression that is specifically relevant to the Merina circumcision as a kind of ‘special agent ritual’ [Fx (a): Fy (b):: Fy (a): Fa−1 (x)]. Some comparisons and contrasts have already been outlined between the overall structural description of ‘representatives of fanjakana circumcising boys’ on behalf of ancestors [Fx (a->boy)], and the representation of ‘women giving birth to children’ [Fy (b->child)]. If we integrate this basic analogy to the counter-intuitive proportional analogy described above, the implementation of the formula gives the expression of the circumcision ritual form that encompasses the whole series of its variants (see Figure 3).
The canonic formula of Merina circumcision [Fx (a): Fy (b):: Fy (a): Fa−1 (x)].
Put another way, this analogical thought and action constitutes the work and efficacy of Merina circumcision by comparing this ritual to maternal birth in such a manner that the reny velona relation of transmission implies ancestral birth like the karazana relation of transmission implies the transformation of mahery entities into secondary agents that are subjected and subordinated to masina entities (this transformation being realized by those created by the ritual). What is then the purpose of establishing a mapping between ritual actions that are conducted by indexes of nonphysical imagined agents and polities on the one hand, and everyday situations such as maternal delivery where human persons behave intentionally on the other hand? Indeed, Merina people circumcise their boys like Merina mothers give birth to their children in order that ancestral delivery of their children and political subjection of their subordinates may happen together because of their similarity and inferred mutual implication, in spite of their contingent, counterintuitive and fictional status of institutional facts. If there is a performed homology between circumcision and maternal delivery such that the ancestral transmission of vitality is conjectured as plausible, then the subjection of humans and non-humans that embody the opposite qualities of ancestors is logically judged as following, because it actually takes place inside the cult – think to sequences [4ad–8–10–11–14–15] – as it happens outside the circumcision ritual in the form of enacted techniques of power. The component D [Fa−1(x)] is indeed an induced schema for a class of situations that formalizes any Foucaldian technique of power, in the sense of making a patient capable of acting as an entity acted upon and contrary to the representatives of fanjakana. Counterintuitive representations and relations of subjection can thus flow from intuitive grounding, in spite of the fact that the fourth component (D) cannot be predicted from the occurrence of the three components (A, B, C) besides some constitutive rules accounting for such institutional facts, and besides logical operations getting stuck in such ritual actions.
We face therefore a ‘conventional persuasive analogy’ (Tambiah 1985: 70) that works as a political performance evoking, interlinking, and promoting hegemonic behaviours and ideological representations among Merina people. If this one differs from the scientific use of analogy based on predictions and causality relations, it has nonetheless an outstanding property: this counterintuitive proportional analogy is based on structural constraints that are insensitive to political and historical settings (the enduring and invariant circumcision form as a hegemonic performance), while it relies at once on semantic and pragmatic constraints that determine the periodic variation of the ritual form as an ideological endeavour. In this latter case, there is a flexible configuration of the indexes standing for ancestors (what kind of ancestral polity, of hasina exchange, are instantiated?); and there is the necessity of defining the parameters of the actions performed by circumcised men (what kind of technique of power, of ritual sequence, of patient are then instantiated?). This dual architecture promotes the formation and extraction of a new and explicit abstract schema (the component D), from which it is possible to reason and guide future inferences and new mappings (Gick and Holyoak 1983), and therefore to facilitate analogical transfers instantiating similar situations outside and inside the cult.
Hegemony and ideology in Merina ritual policies
The dual hegemonic and ideological aspect of the religious ritual form has been emphasized by many anthropologists (Tambiah 1985; Comaroff and Comaroff 1991). By considering rituals as efficacious actions on human persons that change, through inferences, the state of performers’ minds, Rappaport (1999) has related this dual aspect to the combined transmission of both ‘canonical’ and ‘self-referential’ messages, occurring through the respective use of symbols and indexes; the first one would be carried by the invariant aspects of liturgical performance, the second one being conveyed by restricted possibilities for variations inside this liturgical order (where, who, when and whether to perform or not, with what means?).
So far as the hegemonic dimension is concerned, the basic core of Merina circumcision is stipulated in advance as decorum to express the conventional idea of a transcendental power of growth (hasina) that is hidden in everyday life, ritually controlled, and transferred only by circumcised men. By performing this analogy, Merina people take for granted the conjectured hypothesis according to which human and non-human indexes of ancestors deliver children and create life, as well as its close connection with the propensity of the engendered circumcised men to predate, capture, consume, and appropriate the vitality (hery) and certain attributes of the secondary agents opposite to ancestors. This analogical reasoning depends on structural constraints that comprise the encoded alignment of the relational structure of the components A with B (the orderly sequencing of ritual actions for circumcising is parallel to the steps for giving birth), in addition to the preference for systematicity between the relations A:B and C:D on the basis of a similar high-order relation of implication between A:D and B:C. For this reason, the invariant sequences are those of choosing the D-day, preparing the house, impregnating the gourd, measuring the boy, treating the prepuce, bathing and pouring water over the boy. Moreover, what is also invariant is the required existence of the components C (ancestral delivery) and D (subjection schema), whatever their empirical instantiation (e.g. making manasina, cutting the boy’s prepuce …). By this process, hasina hegemony comes to have such a wide acceptance and long-term stability that no one is able to manipulate its essential meanings, namely the concept of a controlled power of growth whose exclusive ritual use by circumcised men is analogous to their practical subjection of human and non-human beings on behalf of ancestors.
But as far as the ideological dimension is concerned, no circumcision ritual is the same as another performance, everything is opened to contestation, change or manipulation, and the variable components make flexible its basic core. Because participants transmit deictic information pertaining to their imagined roles and status (rank, prestige, power, privilege), there are major historical variations in terms of participants’ profiles, scale of attendance and magnificence, place and duration of staging, or economic outlay (royal circumcision is thus different from that practised by commoners). Indeed, there are multiple ideologies of hasina in Imerina according to the agentive mode of particular descent groups or ambitious political factions (Larson 2000). We can then notice various referential indexations of the fanjakana in terms of its representatives, size, and strength. The instantiation of these representatives depends on which political leaders of which polity are/want to be legitimately recognized and placed at the apex of hasina hierarchy: among ritual variants, this is sometimes a sovereign (mpanjaka), a local royal judge (vadin-tany), a head of a descent group, or prominent male elders who choose the D-day, pierce and tie the gourd, receive the hasina as a mark of loyalty and obedience (Bloch 1986).
Semantic and pragmatic constraints determine therefore the way in which certain relevant relations, attributes and entities are perceived, retrieved and selected to be instantiated, represented, and processed in the circumcision ritual form. These intuitive rules of analogy intervene specifically to derive ideological meaning in the extent that it is impossible to instantiate apart from an historical context what are the indexes of ancestors/fanjakana and what are the actions performed by circumcised men upon a priori indeterminate patients. The inferred instantiation of these actions and patients can depend on, for example, semantic similarities between entities towards whom/which circumcised men behave as ‘conquerors’, ‘predators’, and ‘subjugators’ during the ritual, and entities towards whom/which they get a reason/duty to behave as such to transform them into secondary agents in the mundane life. Processing these kinds of goals and motivations can then lead to focus on specific instantiations of the missing component D by favouring certain correspondences and identifying empirical elements of this abstract subjection schema on the basis of pragmatic and ideological knowledge (Holyoak and Thagard 1990). That’s why these conquered and transformed entities may have been cutting plants, water, land, cattle, sheep, vazimba, women, children, witches and sorcerers, slaves, pawns, faded freemen and nobles, constituents of other demes, of other polities – all these patients having been acted upon during ritual life-cycle and blessing ceremonies, in the framework of domestication or hydraulic works, or at the time of military conquest, raids and reduction to slavery, as well as forced labour.
Inside/outside the cult: From the violent subjection of mahery entities to the making of secondary agents
We can point out typical ritual variants for each historical period that correspond to specific techniques of power intertwined with dominant relations of production in the development of precise Merina polities; namely, demes and chiefdoms under Andriamasinavalona’s descendants’ rules during the 18th century, divine kingship and early state under Andrianampoinimerina’s rule (1778–1809), and imperial, monarchical state under Radama I (1810–1828) and Ranavalona I (1828–1861).
After the unifying rule of Andriamasinavalona, the Merina kingdom was divided up into tiny chiefdoms at war, which were essentially composed of ‘robber barons’ with bands of retainers living on fortified hilltops and ruling over villages settled in valleys (Bloch 1989b). Two dominant techniques of power were then typical until the late 18th century. First, irrigated rice grew on the basis of derived water from mountain run-off that was distributed to terraces running from the valley head down the mountain slopes (Berg 1981): the capacity of opening/closing small holes in the paddies’ dikes permitted the control of the amount of water let into each terrace. Second, slave trade developed from the growing demand from Mascareignes, which enhanced both raids for war prisoners between chiefdoms, kidnappings between demes, and dependents pledged for debts (zazahova) who were sold abroad as slaves. Hence the consequent inflation of call prices for lost relatives compelled by feedback descent groups to engage in petty feuds to capture members of others groups, hold them to ransom or sell them to slave merchants in order to buy back their own parents (Larson 2000).
If we refer to the second-hand descriptions of ritual sequences prior to Andrianampoinimerina’s rule (Bloch 1986: 120), two main performances deserve our attention because of their similarity with these two mundane subjection schemas: they also respectively take water and other descent groups’ members as patients put in service of ancestors and fanjakana [11]. The first one is the invariant capture and fetching of mahery water under the threat of warriors [15’]. The second one is the innovated way of making a simulacrum of the child’s prepuce cutting, by bringing in front of a circumcised boy placed on a drum on the threshold a young red bullock in the courtyard whose ear is cut with a knife.
These two ritual sequences instantiate the missing component D of the circumcision form at the same time that they are analogous to the first and second relations of power. With regard to the ritual capture and fetching of water, we can emphasize how warrior elites living on fortified hilltops could, in addition, control the flux of water by blocking its run whenever peasant communities dared to refuse their protection against tribute. The transformation of water into a secondary agent fertilizing the soil was then operated under the aegis of warriors and peasant elders deciding both at different levels to let it enter into the paddies’ dikes. In the circumcision ritual, the subjection schema of water was analogized through the role of elders piercing the gourd [11c] and the unmarried young leader of the procession threatening with a spear the operation of water scooping [11e–f]. As regards mock circumcision, the circumcised boy was replaced by a young red bullock that had been stolen or captured among wild cattle herds. This corresponds to the subjection schema of foreign deme members implemented in the circumcision form: these captured individuals were sold abroad to get silver money and their social death was the matching rebirth of lost kin. As newborn babies (strangers to the deme) were called ‘red children’ (zazamena), this young red bullock stood for raided and kidnapped members of other descent groups to the extent that circumcised men cut its ear on which traditional cattle markings (geometrical symbols of property and belonging) were carved when domesticated by a descent group in order to identify its owners. This ritual operation consisted then in cutting or impeding the bullock’s affiliation to a descent group in exchange for not harming the circumcised boy. As captured foreigners, red bullocks were transformed into secondary agents in the sense that their new disaffiliated status of sacrificial victim was the counterpart of the strengthening of circumcised men’s membership.
Andrianampoinimerina’s rule marked two important changes with the reign of robber-barons. From the mid-18th century, deforestation had encouraged the development of marsh production thanks to an irrigation system based on dyking up lowland rivers to store water in reservoirs and to divert it through canals to low-lying rice fields (Berg 1981). This labour-intensive technology required the coordination of hitherto isolated descent groups whose ancestral territories were henceforth linked through the common maintenance of dams and dikes. Consequently, these flooded rice paddies were managed differently from terraces or swidden plots, in the sense that land ownership was appropriated by descent group hierarchies at the expense of households, extended families, or restricted ancestries, and that heavy taxation was imposed by rulers who controlled the mobilization of labour for communal work and the codification of laws for access to newly irrigated lands (Bloch 1989b). This labour-intensive technology was also an incentive for importing slaves and requiring forced labour from dependents and political subjects. The second change concerned, therefore, the reorganization of the slave trade in Madagascar in parallel with the political reunification of Imerina. Andrianampoinimerina offered protection against enslavement to firenena, which recognized his sovereignty and accepted taxation due to their new territorialization. At the same time, he reserved privileges and monopolies to the leading long-distance merchants by reducing rebels, war prisoners, and rivals to slavery, and by conquering places that were situated at the crossroads of Malagasy commercial networks (Larson 2000). But he also boosted fokonolona by devolving to them the settlement of disputes and the right to reduce into slavery big men and slave traders who were hostile to their familial ethos of mutual responsibility (fihavanana).
Within this state formation process, royal speeches (kabary) were put in the service of the state orchestration and synchronization of circumcision ritual transmissive frequency (Bloch 1986: 114): a series of circumcisions took place every seven years, each deme waiting for its turn according to its rank, the king travelling through his kingdom to receive everywhere the gift of an uncut silver coin and the payment of a new tax (kirobo) by each prepared house of circumcision [5’]. The second most noticeable change in the ritual performance was that the whole population of the kingdom had to participate in a vast campaign of clearing the land of witches and sorcerers (mamono voalavo: ‘killing rats’) by making convicted persons drink the tangena ordeal [1’]. This use of poison ordeal was indeed the main technique of power to confiscate goods and lands as well as to dispose of one’s enemies, rich slave merchants, or supporters of rival political factions who were a threat to the unity of the kingdom and the demes. Manipulating accusations of witchcraft allowed the representatives of fanjakana to transform these emergent and independent chiefs or big men into secondary loyal subjects of the polity who were subordinated to its royal ancestors. These two relations of power (tribute and poison ordeal) were directly analogized in the circumcision form, through the payment of kirobo during the act of manasina [5’] and the administration of tangena during the witch-hunting [1’].
The following rules of Radama and Ranavalona in the 19th century were both in continuity and rupture with Andrianampoinimerina’s. Half of the Merina population was henceforth composed of slaves imported from East Africa and other parts of Madagascar to work in rice fields. But the interdiction on enslaving the king’s own subjects meant that royal armies in search of new populations to raid had now to go ever-greater distances towards the coasts, involving a degree of logistical organization beyond the scope of the army of farmers that highland rulers had previously used (Ellis and Randrianja 2010: 119). As a result, the creation of a professional army and the military control of commercial ports along the island’s coasts were among the first priorities of the new imperial state. Radama I made diplomatic treaties with the British governor of Mauritius to enact the abolition of slave exportation in exchange for silver money, weapons, and technology transfers in order to develop a state army and administration, a literate language, and a proto-industry for export. Ranavalona I carried on these mercantile policies but under the conduct of a new merchant and military hova oligarchy that controlled forced labour, manufacturing, customs tariffs, and patron-client systems (Campbell 2005). The competitive mix of religious traditions, first with Merina evangelization and then with the spread of revival movements focused on vazimba cults, led this oligarchy to transform major Merina rituals into a state religion that enhanced the interdiction and persecution of religious alternatives until the official conversion of elites to Protestantism in the 1860s (Raison-Jourde 1991).
The royal conscription of ‘freemen’ for forced labour (fanampoana) and the use of literacy for governing and communicating ancestral ideology were the two new techniques of power accompanying this culture change. Malagasy people were liable for taxations and to work free in manufacturing or porterage, as some of them were accountable for attending state schools and churches, or for performing military service. The exercise of bureaucratic power through written communication developed at the expense of kabary to census populations and garrisons’ provisions, to negotiate with foreign powers and regulate coastal trade by registering imports and exports, to collect charges and transmit instructions to remote state authorities, as well as to transcribe Malagasy language and standardize Merina ancestral customs (Berg 1995). The staging of numerous and elaborate national rituals based on theatrical acts of manasina at frequent and regular intervals included ceremonies of royal circumcision that army officers, career soldiers, ministers, judges, and ambassadors attended alongside the population. They were at once the opportunity for conducting a military census and the prelude for military campaigns. Among significant variants of circumcision, unmarried young soldiers went to fetch powerful water in the sea, every stage was accompanied by the firing of artillery, and the sovereign or his representatives visited locations beyond Imerina to be paid homage (manasina). But once again, the two main innovative ritual sequences were in phase with these new forms of domination and subjection: [6’] The sovereign and his representatives commandeered and distributed ‘vazimba’ cattle to slaughter among descent groups and localities of the imperial state. [5’] The sovereign, his family and tens of thousands of his subjects performed together the ‘writing dance’ (soratra) for several days, which took place in an area divided into differential paths for reaching the sacred stone on which royal boys were to be circumcised.
In the first case [6’], the wild cattle took the name of ancient native people without descent. They were captured in the grazing country by young soldiers in the name of their belonging to sovereign territory, and their vitality was consumed by territorialized groups that were loyal to the monarch and incorporated to the state apparatus. This was clearly an allegory of fanompoana, as conscripted freemen were recruited in priority among poor peasants made to serve the merchant and military state oligarchy and not anymore their own descent groups. In the second case [5’], this dance area was similar to the board for a Merina national strategic game (fanorana) whose principle rule is to advance on opponents along various lines and to take his pieces (Bloch 1986: 135). Common people were following different tracks from the royal family and the oligarchy, but all were as a body reproducing through their individual movements the wiggly lines of the writing process until they each reached the masina stone and covered the board of conquered Malagasy populations.
Conclusion
This Malagasy case-study argues in favour of an alternative to the massive modularity hypothesis of the ‘new cognitive science of religion’. In spite of claiming not to seek a psychological reductionist explanation of the religious domain, this evolutionary hypothesis has been until now focusing on domain-specific cognitive devices such as modules (hyperactive agent-detection device, cheater detection, hazard precaution system, theory of mind). These modules are supposed to explain cross-cultural religious thoughts and behaviours. But the result is incapacity of conceptualizing why people act and think as they do here and there but not everywhere, privileging the variation or persistence of certain religious traditions at the expense of some others. As Boyer (2005) himself seems to deplore, this ‘standard model’ is a general, probabilistic, and experience-distant one which still doesn’t provide any descriptive or explanatory model accounting for those factors that distinguish one religious tradition from another and determine their historical spread and evolution. This is precisely the point where social and cognitive sciences can combine and work together, on the condition that they articulate domain-specific and domain-general mechanisms as well as etic and emic observations and exegeses. This is the royal path followed by Descola (2005) when reconstituting animism, naturalism, totemism, and analogism cosmologies on the basis of analogical reasoning, schema theory, and intuitive dualism related to the theory of mind. This is the royal path the present Malagasy case study has taken by paying attention to the transformation of representational formats and the relevant context of activation of these cognitive processes, in order to render intelligible the variation and persistence of Merina ritual circumcision throughout history. In this vein, Lévi-Strauss’s canonic formula can be reinterpreted as a counter-intuitive proportional analogy that formalizes the law of permutation group organizing the whole series of special agent ritual variants such as Merina circumcision, and includes its own parameterization by mind-external factors such as relations of power and production. No doubt this formalization could be programmed and implemented into the main computational models of analogy such as ACME (Analogical Constraint Mapping Engine), IAM (Incremental Analogy Model), STAR (Structured Tensor Analogical Reasoning), LISA (Learning and Inference with Schemas and Analogies) or COPYCAT. But that’s another (just so?) story.
Footnotes
Acknowledgements
Author would like warmly to thank Luke Freeman, Maurice Godelier, Anne-Christine Taylor, and Andrew Walsh for their careful proofreading and insightful comments.
