622837

research-article2016

MCU0010.1177/1359183515622837Journal of Material CultureSeveri

Journal of

MATERIAL
CULTURE

Article

Authorless authority:
A proposal on agency and
ritual artefacts

Journal of Material Culture

2016, Vol. 21(1) 133150
The Author(s) 2016
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/1359183515622837
mcu.sagepub.com

Carlo Severi

Laboratoire danthropologie sociale, Collge de France, Paris, France

Abstract
In this article, the author proposes a re-examination of the first example of agency given by Alfred
Gell in his famous Art and Agency (1998), namely the Zinganga nkisi (nail fetish) of Western Africa.
This article essentially argues two points. First, the relation between the artefact and the person
it incarnates is not, as Gell has defined it, bi-univocal (one-to-one), but can be better described as
one-to-many. In the authors view, the ritual artefact does not work as a mirror reflection, but as
a crystal. Secondly, the kind of distributed I that the artefact enacts is not composed of a single
identity distributed in several material occurrences, as Gell has described it in Art and Agency. In
the authors view, another concept of plurality, defined as a set of different identities condensed
in a single object, describes more appropriately the kind of agency that characterizes the nkisi.

Keywords
agency, artefact, nail fetish, ritual action, Western Africa, Zinganga

Introduction
In this article, I propose a re-examination of the first example of agency given by Alfred
Gell in his famous Art and Agency (1998), namely the Zinganga nkisi (nail fetish) of
Western Africa. I want to see whether we can identify some unexplored aspects of the
many possible ways to act that might be attributed to a ritual artefact. In order to do so,
I work on the specific kind of action attributed to that nkisi, namely a certain form of
authority, and look at the authority of the artefact from the point of view of what I have
defined as the pragmatics of what is shown that is, the exploration, in Bhlerian
terms (Bhler, 1990[1934]; Hanks, 2005, 2006), of the I, here and now of the ritual
Corresponding author:
Carlo Severi, Laboratoire danthropologie sociale, Collge de France, 52 Rue du Cardinal Lemoine, 75005
Paris, France.
Email: carlo.severi@ehess.fr

Downloaded from mcu.sagepub.com at Australian National University on August 29, 2016

134

Journal of Material Culture 21(1)

action attributed to the nkisi. I will essentially argue two points. First, the relation
between the artefact and the person it incarnates is not, as Gell has defined it, biunivocal (one-to-one), but can be better described as one-to-many. In my view, the ritual
artefact does not work as a mirror reflection, but as a crystal. The means by which it
refers to personhood is more comparable to refraction than to a mirror image. Secondly,
the kind of distributed I that the artefact enacts is not composed of a single identity
distributed in several material occurrences, as Gell (1998) has described it in chapter 4
of his famous book. In my view, another concept of plurality, defined as a set of different identities condensed in a single object, describes more appropriately the kind of
agency that characterises the nkisi.
In order to prepare the analysis of the nail fetish, let us focus first on the concept of
authority itself and clarify some points about it.

The concept of authority: Principles and sources

When the Florentines conquered the city of Pisa in 1406, one of their most prized war
trophies was a book: a copy of the Justinian code drawn up in the early part of the 6th
century and formerly preserved in Constantinople. Previously known as the Littera
Pisana, it was swiftly re-baptised Littera Florentina and transported to Florence, where
it was housed in the Palazzo dei Priori and placed in a tabernacle as if it were some
sacred relic. On the rare occasions when it was displayed in public, the book was placed
on a table and surrounded by candles. In a rite dating back to Antiquity, the monks and
magistrates responsible for the codex bowed when brought into its presence. This ritual
homage, so evocatively described by Settis (1986: 411412), sums up a number of traits
that together constitute the core identity of European notions of authority. Whether it is
defined as the right to command or influence others, or as a form of power that endows
with influence and the capacity to convince (Lalande, 1926), the European idea of
authority is intimately tied up with writing. As Jan Assman (1992) has shown, the authorisation and memorisation of the traditions of Antiquity are a result of the canonisation of
a corpus of texts, the veracity of which has been established once and for all. Thus the act
of wielding authority (whether it is religious, legal, philosophical or literary) is necessarily linked to the figure of an author. The origins of an intellectual tradition are always
to be found in the oeuvre of an author, from whom it derives its authority: whence such
formulae as the Platonic, Homeric, or Marxist tradition. The canonisation of a corpus of texts produces a principle of recognised veracity, which finds its source in a real
or mythical author. Whether the author occupies the role of ancestor vis--vis the tradition (think of Solon and Greek law) or whether he wields a contextually specific form of
authority guaranteed by that tradition, the relationship between a corpus of canonised
texts understood as a principle and an author qua source of authority remains fundamentally unchanged. One can, following Foucault, submit the notion of the author to critical
scrutiny, reducing it to a mere organising function. One can separate it off from everyday
speech where, Foucault again claims, there is no space for the figure of the author. One
can, with Florence Dupont, stress that the author-function is not always coterminous
with the person who actually composed the text1 (Dupont in Chartier and Calame, 2004:
171189). And one can argue that the term author refers rather to a specific utterance

Downloaded from mcu.sagepub.com at Australian National University on August 29, 2016

135

Severi

position, to a being of thought (Foucault, 1981[1969]) than to an actual individual.

Nevertheless, the fact remains that the Western concept of authorship depends on the
relationship between a textual principle and its real or mythical source. Within our tradition, it is inconceivable that an authority principle reposes on something other than a
canonised text or that authority lacks a clearly identifiable source. Nulla auctoritas sine
auctore, one might say.
What though of so-called societies without writing, lacking any form of canonical
texts? How is authority born of a tradition wielded in such contexts? When traditional
knowledge is purely oral and the idea of the author absent, how can we conceive of the
principle and the source of an authority born of a tradition (and not just in the political
sphere)? Must societies without writing be acephalous then, as we were wont to say
in the 19th century? Is their traditional knowledge condemned, as Goody (1987; Goody
and Watts, 1968) still sometimes claims, to a state of perpetual disorder and deprived of
any overarching structure?
It is right to be suspicious of negative definitions. Societies without writing do not
constitute some homogeneous bloc in stark contrast to an equivalent bloc of literate societies. They are, instead, highly diverse, and so our analysis must work up from specific
case studies. The approach I adopt here cannot, however, be defined simply in terms of
its empiricism. The degree of abstraction proper to the concepts at play is also important.
Whilst careful to avoid the ethnocentric extension of Western concepts to other situations, I nonetheless endeavour to identify higher-level abstractions that can help us to
think through a range of authoritative traditions. Thus, I do not seek to identify authors
and processes of canonisation within oral traditions, but focus instead on the unpredictable forms that the principles and sources of authority can assume in such contexts.

Evidentials, pragmatics and artefacts

Let us begin by underlining the fact that the act of endowing a proposition with authority
by describing its source is not restricted to written traditions. Such an evaluation of a
propositions truth value may be a simple by-product of the use of language, and this
long before a particular knowledge set is instituted qua tradition. For instance, in a number of Amerindian languages there exists a set of lexicalised particles, called epistemic
or evidential classifiers, that take the form of lexicalised particles and whose role is
explicitly to identify the nature of the experience that constitutes the source of a proposition and endow it with a particular truth value (Aikhenvald and Dixon, 2003; Dlage
2005; Landaburu and Guentchva, 2002). Thus amongst the Sharanahua (Dlage,
2005), the classifier -quia denotes a belief derived from hearsay, whereas -quin denotes
knowledge derived from direct experience, and -quian direct knowledge (derived, for
instance, from dreams) that cannot be shared with ones interlocutor. The use of evidential suffixes corresponds to what Landaburu (2007) calls the grammatical management
of truth. In other words, the nature of the source of authority is fully lexicalised and
constitutes an integral part of the semantic structure of language.
This evaluation of the cognitive value of an utterance can also be communicated pragmatically, by clearly defining the status of the utterer and the social conditions of the
speech act. Authority, understood as a process leading to the emergence of a belief, is the

Downloaded from mcu.sagepub.com at Australian National University on August 29, 2016

136

Journal of Material Culture 21(1)

product of certain types of utterance or speech act based on a particular definition of the
nature of the utterer. This is true of many Amerindian shamanic traditions where the
authority-function (which is entirely dependent on ritual activity) emerges via the definition of a complex shamansinger identity, itself based on the production of a series of
contradictory acts of ontological identification. The utterer, who may make use of
abstruse forms of communication (secret languages, esoteric metaphor, onomatopoeia
and voiced brawl breathing), is seen as simultaneously belonging to several different
ontological schemata. When he engages in ritual speech, the shaman is both human and
animal, animal and plant, etc. Despite this, these unusual types of utterance, which are
often restricted to ritual contexts, invariably attribute authority to a human being. Though
he is never author of his own chants (even improvised songs are the handiwork of supernatural beings), the Amerindian shaman is still, on a more prosaic level, a human utterer.
It is only his ritual function that lifts his speech out of the realm of the ordinary and
endows it with a particular pragmatic status (Severi, 2015[2007]: 207244).
In this article, I focus on situations in which the right to wield an authority-function is
not assigned to a human being (however exceptionally defined), but to an artefact. In
such cases, the object (by virtue of its mere presence) exercises an authority-function,
setting in train a process that leads to the emergence of a belief. In such cases, that which
makes a traditions utterances imperative is the product of a quite different constellation
of relations from the one described above. Traditional knowledge relies on the anonymous and non-human nature of the authority that lies at its source. It is remarkable that
the mere act of transferring authority to an artefact is insufficient to guarantee both its
independence of human actors and its anonymity. The Western tradition provides us with
clear examples of situations in which authority is transferred to an artefact. Classical
Roman law did not distinguish between people and objects when it came to the attribution of guilt and the act of sentencing. Statues of the Emperor, for instance (by virtue of
their mere presence) were able to legitimise or enforce contract (Bettini, 1992: 162).
There are also several examples of statues being prosecuted, as when the Romans
indicted a statue of the Greek general Philopemon, whose defence was entrusted to
Polybus (De Angelis, 2007: 3756). In Byzantium, the word legimus, used to sign official documents, indicated the emperors direct intervention in a world where scarcely
anybody still spoke Latin (Settis, 1986: 411). This almost incomprehensible word served
as an index of the emperors presence and authority. Closer to home (as Lea Dickerman
has shown, in Roth and Salas, 2001), the portrait of Lenin was used in the Soviet Union
as an icon, and sometimes even as an index, of his authority.
In all these cases, however, the authority thus indexed has already been defined. The
foundational schema upon which the authority rests (and which I have identified in the
relationship between a text understood as a principle and the figure of the author qua
source of authority) continues to function. The legitimising principle remains fixed to a
corpus of texts rather than being transferred to an object. Though the source of authority
is indexed or replaced by an artefact, it is not identified with it. What interests us here, in
contrast, are situations in which authority is uncoupled from any human source or avatar.
Unlike in the Roman and Byzantine examples mentioned above, such situations are not
simply a reflection, a response to, or the ghost of some prior coupling of a person and an
authority. Objects themselves can be made to bear full responsibility for the wielding of

Downloaded from mcu.sagepub.com at Australian National University on August 29, 2016

137

Severi

authority. In such cases, the object is anonymous and its identity is opaque, mysterious or
indecipherable insofar as it is not a reflection of any human identity: neither that of any
participant in the ritual, nor that of its actual author (i.e. maker). In other words, the
artefact has no external referent: it contains its own principle of legitimacy. As we have
seen, the source of a traditions authority can be expressed pragmatically, by endowing
the utterer with a particular status. As regards the ritual use of artefacts, this status
amounts to the attribution of an extremely minimal (indeed almost null) definition of
identity to the objects in question. Accordingly, the object is associated with a form of
intentionality seen as independent of any human will and, in certain cases, as uncontrollable. In this independence lies the problem to which I want to offer a solution.

The Fang case and the mvet

One fine example of this is the oral tradition of the Fang of Cameroon, analysed by
Pascal Boyer (1988). This tradition comprises a vast repertoire of songs that recount,
somewhat like a saga, the shifting relations of conflict and alliance that prevailed between
three tribes of giants when the world was yet young. The poets tell of the adventures of
the scions of Yo (ancestors of the Fang and of all humans) of the people of Oku (mythical, but mortal beings from far-off lands), and finally of the Engong (a race of immortals). The two mythical tribes who do not possess the secret of immortality wage terrible
wars against the others, and generally ally themselves with their ancestors and ghosts.
These songs (called mvet) are sung by poetmusicians who play a harp-like instrument
(called a stick-zither) (see Figure 1).2
This song cycle boasts a richness of both style and theme that makes it one of the
jewels in the crown of traditional African literature. Some of its chants are also highly
elaborate. We only need to look at the 400-page transcription of the saga sung by Zw
Nguma in the 1960s (Zw Nguma, 1972). This is not the place to detail the wealth of
this tradition, with its grand themes, genres and schools. Instead, I shall concentrate on
only one key aspect: the way in which this traditional knowledge is transmitted or handed
down and, more particularly, on the relationship that is established between singer, song
and instrument in the act of utterance. At first glance, the learning of these epic songs
seems to follow a fairly standard pattern. Each singer is situated in a chain of transmission in which each link is named and which reaches back from master to master until it
reaches a named, but less clearly identifiable, mythical character supposed to have
invented the tradition. The real learning of the song cannot, however, be reduced to a
mere process of musical or literary internalisation and reproduction. To master a mvet,
one must acquire an individual voice, and to do so, one must establish a specific relationship with a supernatural being. It is clear that in order to become a mvet singer one
must identify with (or rather acquire) a particular identity derived from an indeterminate
or paradoxical principle whose true nature remains unknown. In Fang terms, one must
possess an evur or a byang. These ambiguous concepts refer to supernatural substances
supposed to lodge in the pit of an initiated singers stomach. The term evur, normally
used to refer to a magical force related to the ability to bewitch, here describes a knowledge principle: in order to play a mvet, one needs evur. But this in itself is not enough.
One must also have received byang i.e. have undergone a precise form of initiation

Downloaded from mcu.sagepub.com at Australian National University on August 29, 2016

138

Journal of Material Culture 21(1)

Figure 1. The Fang harp, from Boyer(1988). Reproduced with the permission of Socit
dethnologie, Paris.

(Boyer, 1988: 63). The mvet itself (qua sung epic) is understood in terms of these substances, which represent in mysterious but physical form the songs veracity, efficacy
and faithfulness to a tradition.
A mvet is then, by definition, bewitching and a product of secret initiation. But the
relationship between knowledge of a song and the initiation process does not stop there.
Boyers interlocutors repeatedly stressed that in order to sing, one must really have
eaten the mvet. This alimentary metaphor condenses a whole network of relationships
that the poetmusician must establish with spirits whose voices find expression in the
mvet. What is striking when one reads the transcript of Zw Ngumas mvet are the
interludes where little by little the singer sketches out the history of his initiation. He
begins by stating that the mvet loved him like a woman. In the end, the mvet loved [the
singer] the ways girls love boys (Zw Nguma, 1972: 77). The poet then repeatedly
asserts that this contact with the words of the ancestors will make him sick and even
die: the village singer is killed by the melodies of mvet (p. 55), I am dying for the harp.
Why? I, Zw Nguma, am dying by mvet (p. 77). Ethnographers have explained this as

Downloaded from mcu.sagepub.com at Australian National University on August 29, 2016

139

Severi

an allusion to the risks of initiation (Boyer, 1988; Zw Nguma, 1972: 230), and the
dangers of revealing initiatory secrets are indeed mentioned several times in the transcription (I run the danger of revealing the relic of mvet, Zw Nguma, 1972: 231).
According to several native exegetes, this rather poorly guarded secret may involve the
symbolic sacrifice of a close relative of the singer, such as his wife or brother (p. 233ff).
Whether or not this is true, it is clear that this death by mvet transforms the singers
identity. In the interludes that accompany Zw Ngumas words, the dead poet first
transforms into a series of songbirds. Later, this contact with the mvet saps the poets
identity, transforming him into nothing more than an instrument for the spirits voices.
The acquisition by initiation of the mvets veracity and efficacy of utterance coincides
with the annihilation of the singers human identity. Various statements that appear in
Ngumas transcription support this interpretation. So singers declare that they are
merely a bar of a xylophone or nothing more than the string of a harp. After the initiation, the singer and his instrument, his speech and his voice (but also the song itself) have
become one composite being composed of artefact, utterance and utterer. The three are
collectively known as mvet. Once these different identities have been collapsed into
one, the initiate is defined by his dependence on a force that remains fundamentally alien
to him. As one mvet singer expressed it to Boyer: when I sing the mvet, what I say is a
dream, I cannot recall it (Boyer, 1988: 27). Or in the words of another, it is the mvet
speaking; these things are not in my heart. The veracity and the efficacy of the mvet are
not expressed through the person of the singer, nor via his knowledge of the words of the
chant (which constantly change), but by the nature of the artefact he wields: his harp. It
is only by analysing the artefact that we can understand the relationship between the
authority that founds the tradition (represented by the hidden voice of the spirits) and its
interpreter.
Actually, when the latter defines himself as an instruments bar or key, he refers
not only to the structure of the instrument, but also to a specific instrumental technique.
Singers often describe one particular stage in the learning process as crucial. Let us
describe this in detail. First, the young Fang musician must slowly and painstakingly
learn several melodic lines by heart. Once a certain degree of difficulty is reached, the
apprentice must make use of a different form of knowledge acquisition. Certain technical passages are crucial. The young musician will make a curious experience: rather
than gradually learning the lines, he struggles at length and seemingly fruitlessly to
acquire them until, one day, the ability to do so appears to materialise out of nowhere.
He just knows how to play them. Indeed, he will find that he is almost incapable of
playing the two lines separately. His pride at mastering the difficult sequence is tempered by the awareness that some aspect of the musics organisation escapes him. For
Fang musicians, it is precisely this type of experience that brings out the hidden voices
of the spirits. Let us examine more closely this musical form of the mvet spirit presence. The following are two transcriptions of a musical extract transcribed and analysed by Boyer (see Figures 2 and 3). Figure 2 separates the two melodic lines whilst
Figure 3 integrates them.
Each line is assigned to one of the players hands. When the two hands play together,
the two differently rhythmed lines (the left hand plays triplets, while the right plays a
highly regular duplet) intersect with one another, producing a surprising melodic

Downloaded from mcu.sagepub.com at Australian National University on August 29, 2016

140

Journal of Material Culture 21(1)

Figures 2 and 3. The musical cell, from Boyer 1988. Reproduced with the permission of
Socit dethnologie,Paris.

fragment. If we play the two lines on a piano, it becomes clear after a few repetitions (i.e.
once a distinction between the acoustic shape and the background is established) that
what we in fact hear in the foreground is a succession of fragments made up of a repeating demi-tone (here: B C B C B C, etc.). The result is an extremely simple melody that
emerges out of the combination of the different rhythms followed by the two prior melodies the musician has learned to play together. This is a very similar phenomenon to that
produced when two tambourines play slightly different sequences. In such cases, another
musical cell is produced one which is clearly recognisable, but which is played by no
musician as it is the result of the combination itself (Boyer, 1988: 121). In the sequence
played on the Fang harp, Boyer continues,
a similar effect is created. The instrumentalists hands appear to be operating independently
of one another, but the combination of the two produces a resultant melody that the
instrumentalist has not played and which it seems that he cannot play, except by repeating the
whole sequence.

Though this melody is, properly speaking, the outcome of an interaction between two
separate melodic lines, it seems to have its own independent, autonomous existence.
Theorists of Western music call this phenomenon (which has a rhythmic and a
melodic aspect) a resultant melody. Its technical aspects do not concern us here. Suffice
it to note that this mvet fragment embodies the vital shift from a sound intentionally

Downloaded from mcu.sagepub.com at Australian National University on August 29, 2016

141

Severi

produced by an instrumentalists playing of the strings and the emergence of a voice

that is not entirely his own. The hidden voice that comes of this is made up of traits
belonging to prior melodic lines and it is these lines that occupy the musicians attention, both during the learning process and later when he plays them. By a typical figure/ground reversal, the third voice underscores a convergence between the two lines
that otherwise tends to remain latent or implicit. In other words, this musical cell is not
the intentional result of the playing of the melodies, but seems to emerge from the
instrument itself. The artefact thereby acquires its own voice one independent of the
poet. For Fang musicians, it is this voice that expresses the source of the tradition and
so the authority on which it is founded. Although the Fang interpret this spirit-voice
as the sound-image of the spirits presence, they are aware that it is the product of a
particular technique. The musicians interviewed by Boyer are in no doubt about this.
Nonetheless, this voice the voice of the tradition that it authorises is seen as both
hidden and anonymous. The identity of those who sing, which is independent of the
player, remains somehow ungraspable. It is this paradoxical definition of identity (at
once evident, in the sense of being a bearer of truth, and undecidable) that transforms the artefact into a source of authority.
It is perhaps helpful, at this juncture, to explain the nature of this authority-wielding
process in an oral tradition by means of an analogy. Elsewhere (Severi, 2015[2007]), I
have analysed the role played by particular plastic or pictorial visual representations
(which I call chimaera-objects) in the transmission of traditional knowledge. These
objects are typically composed of fragments drawn from representations of different
beings. This double, or multiple, reference to representations of different beings in a
single image produces a form of presence that is not physically inscribed on the surface
of an object or in its sculpted form, but that emerges out of their conjunction. We see this
quite clearly in the following Haida mask (Figure 4).
The image is reduced to a handful of traits, thereby simplifying its structure and causing the observer to perceive a representation that mobilises both the figure of a crow and
a human face. This type of chimaera is thus a set of visual indices wherein what the
viewer is given to see necessarily calls to mind an implicit presence, which is entirely
produced out of indices present in a mental space. The images condensation into a
handful of essential traits relies on viewers projective interpretation of the final form:
they must fill in the gaps. We can describe this process as the intensification of an images
cognitive efficacy that, via an act of visual inference, sets its missing parts to work. This
kind of chimaeric representation, qua specific modality of human perception, is obviously neither a localised phenomenon, nor a formal configuration linked to just one
register of perception.3 In fact, if we return to the Fang harp, we could interpret this
instrument, whose independent voice endows it with a life of its own, as an acoustic
chimaera, as its voice is entirely composed of implicit elements of the real sound and
relies for its existence on an act of mental reconstruction, expressed by a reorientation of
aural perception. Here, as with visual chimaeras, we are faced with the production of an
indirect form of presence that emerges out of the playing of other melodic lines. This
model of a living artefact, which allows us to glimpse (behind the objects presence) the
definition of an authorless voice, can then be seen as an extension of the model of chimaeric representation: it produces an analogous form of salience, but in the aural domain.

Downloaded from mcu.sagepub.com at Australian National University on August 29, 2016

142

Journal of Material Culture 21(1)

Figure 4. A Haida mask. New York: American Museum of Natural History.

In his analysis of the mvet tradition, Pascal Boyer (1988) stresses its paradoxical
aspects, lack of rational determination and even its use of empty concepts. For him,
these sagas are typically made up of confused and contradictory statements that yield no
clear propositions (Boyer, 1988: 40). To the extent that one can, from this perspective,
speak of Fang harps in terms of their authorising a tradition, it is because particular
objects, or rather the specific uses to which they are put, display qualities that catch the
attention in such a way that the imagination can unfold itself through a process of indefinite repetition (1988: 17).
For my part, I think that we should focus our attention on the forms of the enunciation
of truth that this tradition confronts us with. To this end, we must understand not merely
their content, but also some more abstract and perhaps generalisable traits. In the Fang
case, we can identify three formal traits: (1) a series of identifications (collapsing of
identities) leading to the creation of a composite being, the mvet, which comprises a
chant, an instrument and a singer; (2) a weak definition of the identity of the source of
traditional truth (an apparently authorless voice, in the sense that it is associated with a
quasi-null identity); and (3) the transfer of the authority-effect onto an artefact, which is
constructed as autonomous and endowed with subjecthood.
We can use this initial model to explore further examples where artefacts are not only
capable of speech, but also of action liable to authorise a tradition. Across these various

Downloaded from mcu.sagepub.com at Australian National University on August 29, 2016

143

Severi

contexts (with their varying degrees of complexity), we find artefacts with similar ritual
identities and, above all, similar degrees of autonomy.

The Zinganga nkisi: A re-examination

In Art and Agency (1998), Alfred Gell made the point that those artefacts in our museums that we label primitive art are not merely instances of a universal instinct underlying artistic creativity. As well as being the product of a particular aesthetic, many of
these objects were also originally treated as living beings, notably in ritual contexts.
As such, artefacts endowed with agency express specific networks (skeins) of relationships between members of society. Whether it is a question of performing a sacrifice, marking out a symbolic space, or correctly accomplishing a rite of passage,
living artefacts play a crucial role. In such contexts, the authority wielded by an
artefact in a textless and authorless tradition such as that of the Fang can only be
understood as a specific expression of its own agency. With this example in mind, let
us now re-examine one of Gells own examples of ritual agency: the use of nkisi nail
fetishes (see Figure 5).
Nkisi are used throughout a wide swath of West Africa, comparable in scope to the
mvet zone. The ethnographic record suggests that these nail fetishes were mainly used in
legal contexts either as a record or as a guarantor of a pact or promise. The nkisi would
punish transgressors and could also be used to wreak vengeance. Gell draws his material
from the ethnography of Dennet (1968[1906]), who described the fabrication and ritual
usage of Congolese nkisi in the following terms:
A palaver is held and it is decided whose kulu (soul) it is that is to enter into the Muamba tree
and to preside over the fetish to be made. A boy of great spirit, or else, above all, a great and
daring hunter, is chosen. Then they go into the bush and call his name. The Nganga cuts down
the tree, and blood is said to gush forth. A fowl is killed and its blood mingled with the blood
they say comes from the tree. The named one then dies, certainly within ten days. His life has
been sacrificed for what the Zinganga consider the welfare of the people. They say that the
named one never fails to die People pass before these fetishes (Zinkici Mbowu) calling on
them to kill them if they do, or have done, such and such a thing. Others go to them and insist
upon their killing so and so, who has done, or is about to do them some fearful injury. And as
they swear and make their demand, a nail is driven into the fetish, and the palaver is settled so
far as they are concerned. The kulu of the man whose life was sacrificed upon the cutting of the
tree sees to the rest. (Dennet, 1906: 93, in Gell, 1998: 61)

Drawing up a table of the implicit relations (either active or passive) that the use of
the nkisi implies (see Figure 6), Gell (1998: 28) makes use of two main notions: the
active role of index, i.e. the material thing which motivates the abduction and the
passive role of recipient. In this analysis, the legal (or vengeful) Zinganga nkisi is
caught up in a chain of relations linking indices to recipients, and this chain describes
the objects cumulative agency. The authority wielded by the fetish (qua active index)
over its victim (the passive recipient) emerges out of the chain of actions comprising the
supplicants invocation of the fetish, the death of the hunter caused by the Muamba tree
and, finally, the priests sculpting of the fetish from the wood of the tree.

Downloaded from mcu.sagepub.com at Australian National University on August 29, 2016

144

Journal of Material Culture 21(1)

Figure 5.Thenkisi. From Gell (1998). Reproduced with the permission of the Muse du Quai
Branly and the Runion des Muses Nationaux, Paris.

Gells schema clearly shows how an object can imply a group of relations. In his own
words:
an instructed person, approaching such a fetish, does not see a mere thing, a form to which he
may or may not respond aesthetically. Instead, what is seen is the visible knot, which ties
together an invisible skein of relations, fanning out in social space and time. (Gell 1998: 62)

The idea, however, that networks of relations can be seen in the use of an object
leaves one central point unclarified. The fetish is here conceived of as a full agent and

Downloaded from mcu.sagepub.com at Australian National University on August 29, 2016

145

Severi

Figure 6. The skein of relations mobilized by the use of the nkisi. From Gell (1998).

not merely as the end-point of a chain of actions. The goal of ritual action is to transform the status of the artefact, moving it from a position of passivity to one of activity.
As Gell rightly observes, the fetish acquires the capacity to act because it has been
acted upon. When the nail is driven into the fetish, it is cast in the role of symbolic
victim and reacts against this. The force generated by this reaction transforms the fetish from object into subject of a violent act. The driving in of the nail also transfers
the priests symbolic aggression against the Muamba tree to the fetish. In this one act,
the object is both transformed from victim into avenger and its identity is partially
collapsed into that of the supplicant, who undergoes a parallel transformation from
victim into avenger of a wrong. In this sense, we can also say that the fetish acts as a
substitute of the avenger, since we have several examples of nkisi where a small mirror is attached to the statue. Yet even in the case of the nkisi, the objects actual
behaviour remains unpredictable: its intentionality always retains a certain ideal independence of that of the supplicant. Though it is hard to imagine a situation in which
the action of the fetish might run directly counter to the intention of avenging the
supplicant or obtaining justice, the latter nonetheless never exercises direct control
over the object. It follows that what we need to explain is not so much the series of
events leading up to the objects transformation, but rather the independence (or
autonomy) it seems to acquire in the ritual process.
Let us go back to Dennets narrative, and try to understand the kind of dynamic transformations it can involve. We are told that, once a collective decision has been made, a
priest is instructed to fell a tree. He strikes it and addresses it using the name of the young
hunter who is destined to die, at which point it begins to bleed like a human body. The
wood of the tree is then used to make the fetish. The blood of the wood is coupled with
the blood of a sacrificial chicken. In this sequence we see the young hunters kulu (his
ability to kill) transform from a passive role (qua element associated with the dead
hunter) to an active one (qua source of the fetishs power). The ritual action revolves
around the transformation of a tree trunk that represents the young hunter as victim of a

Downloaded from mcu.sagepub.com at Australian National University on August 29, 2016

146

Journal of Material Culture 21(1)

violent act into a fetish that represents him as the vengeful (and so active) party. But how
does this transformation occur? Gell writes that the tree brings about metonymically the
hunters death (Gell 1998:61). It is worth remembering, however, that in Dennets
description, the tree is not presented as an active party, or as an index. Rather, it is the
priest who sets the process in train by striking the Muamba tree. It is also clear that the
tree, being called the name of the young hunter, is ritually compared with a human body
by virtue of its bleeding. In other words, its identity is not seen as either active or passive.
It is progressively redefined as both.
For, if the young hunter dies certainly within ten days (Gell 1998: 61) following the
tree being struck by the priest, it is surely because his kulu has been assimilated to that of
the tree (their identities have been collapsed). This detail is crucial, because the collapsing of the two identities, the identification of the hunter and the tree, lies at the heart of
the ritual and is sequentially intensified and reinforced. Once this first identification is
established, a chicken is sacrificed and (like the tree and the human) its blood pours
forth. The priest mixes the blood of the chicken and the tree and so identifies the one with
the other. When the hunter dies, his symbolic identification with the tree is complete: the
blood spilt by the priest is indeed his.
Here, the ritual action produces changes that affect each element of the chain of identifications. The collapsing of the identities of the hunter and the Muamba wood destined
to become a fetish produced by two explicit acts of connotation: bleeding and naming
pave the way for another crucial (albeit indirect) identification, which both links the fetish to the hunter via a transfer of kulu and transforms the wood from passive stuff into
active agent. Thus the fetish is born as a ritual subject that partakes of both tree and
hunter and whose identity combines passive elements (felled tree, sacrificial chicken)
and active ones (the hunters kulu). The vengeful supplicant identifies himself (perhaps
by using the mirror) with this ambiguous or hybrid artefact/subject, which is simultaneously passive and active. We must conclude that the fetish has a hybrid, not a singlefaced, nature.
Subsequent ethnography of the nkisi, which complements and deepens Dennets
(1968[1906]) fragmentary and somewhat dated testimony, confirms this conclusion.
MacGaffeys (1986) extremely detailed analysis first suggests that we need to broaden
our estimate of the nkisis ethnographic scope, distinguishing between an official public
use of the fetish (close to an ancestor cult) and a private use, associated with witchcraft.
We also need to bear in mind that the ritual use of nkisi (which, as MacGaffey shows,
also play a part in initiation [Mc Gaffey 1986:111]) is always specific and revolves
around particular sets of clearly defined relations e.g. between father and son or mother
and future child. The legal use of nkisi is then, contrary to Dennets claims, just one
example among many (Mc Gaffey 1986:111112, 140, 264). Once these points are
granted, much of the rest of MacGaffeys analysis helps deepen and confirm my own
conclusions. He notes that the symbolic equivalence of body and tree (which underpins
my explanation of the use of nkisi) is so widespread in this ritual tradition as to be considered a constant and general characteristic of it (Mc Gaffey 1986:128130). He also
stresses that an nkisi is, first and foremost, the avatar of a dead person (Mc Gaffey
1986:137). Tombs, for instance, are invariably referred to as the nkisis house (Mc
Gaffey 1986: 145), which confirms our hypothetical identification of young hunter and

Downloaded from mcu.sagepub.com at Australian National University on August 29, 2016

147

Severi

felled tree. Most crucial, however, is the idea that nkisi are not figures. It is categorically
not (as our Western gaze would lead us to assume) the material representation of a supernatural being. Rather, it is a space of assemblage of several heterogeneous elements, each
of which refers back to a supernatural substance, like evur and byang among the Fang
(Mc Gaffey 1986: 122, 137, 139, etc.). The fetish is the plural image (Mc Gaffey 1986:
120) of this assemblage. This explains why the prototypical nkisi is not a statue, but a
basket or a shoulder bag (Mc Gaffey 1986:112). The statue is always conceived of as a
container of sorts. To take another example, Christian crosses in contemporary Congo
are often considered to be nkisi. Of course, the content of this symbol of Jesuss martyrdom is subject to a process of radical reinterpretation in the context of traditional cults.
As MacGaffey notes, Jesus is seen more as a magician than a victim in the Congo. But it
is not only the content of the symbol that changes. Its form is also reinterpreted. It is less
a unitary symbol than the sign of a crossing of paths. The horizontal line refers to the line
of demarcation between the visible and invisible realms, whilst the vertical line is a beam
of power that bisects it. This serves to highlight an extremely important point: the cross,
qua nkisi, is not a symbol, but a statement of relationships (Mc Gaffey 1986: 119). In
other contexts, the same symbol can represent the four matrilines (father, mother, mothers father and fathers father) that make up an individuals soul (Mc Gaffey 1986:123).
In other words, the nkisi always represents a composite being, or better the sum of ritual relations established between different beings. It is in this light that we should consider
it. MacGaffeys ethnography allows us to push still further, however. He claims that the
vengeful nkisi, whose use is linked to the spilling of blood, is always made up of antagonistic beings: the charm is both avenger and victim (Mc Gaffey 1986: 142). This confirms
our view of the Zinganga fetish as an entity composed of positive elements (corresponding
to the aggressors identity) and negative ones (corresponding to that of the victim).
One can only conclude that both in the example given by Dennet and in those analysed by MacGaffey, our analytical approach must account for the composite (rather
than, as Gell would have it, alternating) nature of the objects ritual identity. This analysis
casts the authority wielded by the fetish in a new light. This authority is wielded thanks
to the creation of a series of relations of identification and definition that are comparable
to those that emerge (only quasi-intentionally) from the strings of a Fang harp. These
identifications, which are serially articulated around a single connotation (of blood, sacrificial chicken, a name, etc.) are always partial. Each participant (hunter, priest, sacrificial victim, and legal or vengeful supplicant) performs his or her part of the rite, but none
of them has a direct relationship to the whole set of relations that together constitute the
fetishs composite identity. Though each contributes to the whole, the identity that inhabits this container is the result of the sum of partial identifications realised in a particular
sequence of actions.

The complex artefact

In his book, Gell (1998) defines the relationship between the artefact and the human agent
as a one-to-one, binomial relation. He writes, for instance that there is an insensible transition between works of art in artefact form and human beings: in terms of the positions they
may occupy in the networks of human social agency, they may be regarded as almost

Downloaded from mcu.sagepub.com at Australian National University on August 29, 2016

148

Journal of Material Culture 21(1)

entirely equivalent (Gell 1998:153). In the same line of thought, he also adds that just like
people, artworks come in families, lineages, tribes, or whole populations. In other words,
artefacts are to be seen as distributed persons, defined as dispersed categories of material
objects, traces and leavings, which can be attributed to a person (Gell 1998: 222).
Let us now return to the comparison between the Fang harp and the Zinganga fetish,
and we will see that none of these two defining traits really apply to the ritual action
attributed to the nkisi. As mentioned, in the mvet tradition, the voice of the spirits seems
to emerge from the harp a phenomenon we can conceptualise through an appeal to
musical theory and the idea of resultant melodies. In a similar vein, we might say that the
nail fetishs identity results from the partial identifications that direct ritual action and so,
just like the Fang harp, it appears to exist independently of the ritual participants. Its
apparent autonomy can be explained in terms of the dynamic that connects the whole set
of relations implied by its use.
In both of these cases, we see a figure (and so a source) of ritual authority, which is at
once ill defined, almost graspable and attributed to an artefact. This type of authority is
wielded in the absence of a real person: the harp-player reduced to a string and the dead
hunter, who eventually becomes a nkisi. As such, its identity is negatively defined. It is
at once materially produced by an intentional action (plucking the harps strings or carving into the bleeding wood) and uncoupled from any direct form of intentionality.4 Be it
harp or fetish, the artefact wields an authority which is both provoked and constrained,
as much imagined as respected. The conclusion to draw is that in textless traditions,
authority is not wielded through reference to the figure of the author, but is defined as a
complex network of relations linked to ritual action. The artefact appears as the image of
a set of relations (rather than of an individual, whether they are mythical author or supernatural ghost) and this depends on the production of a series of partial identifications. A
tradition is thus authorised by a dispositif of anonymous utterance: though they emerge
out of a series of clearly defined interactions, its agentive power and its speech never
coincide with an actual intervention from a ritual participant. Behind the supposed presence of an utterer whose identity remains indiscernible, we can glimpse the objects
evidential function, which ties harp and fetish to an image of the truth. In other words,
the (paradoxical, at least from a Western perspective) space where the artefact is endowed
with agency is that of an authorless authority, where the ritual artefact does not work as
a mirror reflection of a human agent, but as a crystal capturing several identities in one.
The kind of distributed I that the artefact enacts is not formed by a single identity distributed in several material occurrences (as Gell would have had it). It is better described
as a set of different identities condensed in a single, but complex one.
Acknowledgements
This article, based on a revised version of Severi (2008), translated by Matthew Carey, was presented at the UCL Anthropology Department, London, 7 October 2015, and has benefited from a
lively discussion there. I wish to thank all the participants to that debate.

Funding
This research received no specific grant from any funding agency in the public, commercial, or
not-for-profit sectors.

Downloaded from mcu.sagepub.com at Australian National University on August 29, 2016

149

Severi
Notes
1.

In an article warning against the dangers of appealing to Latin etymologies, Florence Dupont
(2004) nonetheless points out that the Latin term auctor (author) is only ever used to describe
a historian and is only very rarely applied to poets. And even then, the historian in question
must have founded a tradition (like Tacitus, for instance). In this sense, an author is above all a
person who establishes a tradition and not (as in contemporary usage) somebody who creates
a literary work. Moreover, the author-function can also be decoupled from the person who
actually writes the text. In such cases (Dupont takes Virgil as an example) the true author is
not the person who composes the text, but the one who commissions it.
2. The mvet is in fact a stick-zither made from a branch of raffia-palm, whose bark is used for
strings. Four thin slivers of bark are attached to the branch at different points and placed under
tension using a small stick planted halfway along its length. This stick is notched at four separate points and each string slots into one notch. Thus the strings are divided in two, giving
each mvet eight strings in total (Zw Nguema, 1972: 10).
3. Such processes are extremely common and there are numerous examples from Africa. If we
restrict ourselves to West Africa, it is worth mentioning Lamps (1996) analyses of ritual
masks among the Baga Mandori of Guinea and, in particular, of the Tongkomba-Bana dyad.
The Tongkomba incorporates elements of both land and sea, whilst the Bana has a long, horizontal headdress composed of the jaw of a crocodile, the face of a human being (with Baga
scarification marks and a womans elaborately braided coiffure), the horns of an antelope, the
body of a serpent and the tail of a chameleon. (Lamp 1996: 144ff)
These characters are developed in what can only be described as a chimaeric dance, where the
dancer must indicate the different animals in turn (Lamp 1996:148).
4. For further discussion of this uncoupling of individual intentionality and ritual action, see
Humphrey and Laidlaw (1996) and Houseman and Severi (1998).

References
Aikhenvald AY, Dixon RMW (eds) (2003) Studies in Evidentiality: Typological Studies in
Language, Vol. 54. Amsterdam: John Benjamins.
Assmann J (1992) Das Kulturelle Gedchtnis: Schrift, Erinnerung und Politische Identitt in
frhen Hochkulturen. Munich: CH Beck.
Bettini M (1992) Il Ritratto dellamante. Turin: Einaudi.
Boyer P (1988) Barricades Mysterieuses et Piges Penser: Introduction Lanalyse des popes
Fang. Paris: Socit dethnologie.
Bhler K (1990[1934]) Theory of Language: The Representational Function of Language.
Amsterdam: John Benjamins.
Chartier R and Calame C (2004) Identits dAuteur dans lAntiquit et la Tradition Europenne.
Grenoble: Millon.
De Angelis F (2007) Dei luoghi e della memoria: Pausania, Filopemene e la fruizione della
Periegesi. In: Cordovana O, Galli M (eds) Arte e Memoria Culturale nellEt della Seconda
Sofistica. Catane: Edizioni del Prisma.
Dlage P (2005) Le Chamanisme Sharanahua: Apprentissage et epistmologie dun rituel.
Unpublished doctoral thesis, EHESS, Paris.
Dennet RE (1968[1906]) At the Back of the Black Mans Mind: Notes on the Kingly Office in West
Africa. London: Harlem.
Dickerman L (2001) Lenin in the age of mechanical reproduction. In: Roth M, Salas C (eds)
Disturbing Remains: Memory, History, and Crisis in the Twentieth Century. Los Angeles:
Getty Publications.

Downloaded from mcu.sagepub.com at Australian National University on August 29, 2016

150

Journal of Material Culture 21(1)

Dupont F (2004) Comment devenir Rome un pote bucolique? Corydon, Tityre, Virgile et
Pollion. In: Calame C, Chartier R (eds) Identits dAuteur dans lAntiquit et la Tradition
Europenne. Paris: Editions Jrme Million.
Foucault M (1981[1969]) What is an author?, trans. DF Bouchard and S Simon. In: Bouchard DF
(ed.) Language, Counter-Memory, Practice. Ithaca, NY: Cornell University Press
Gell A (1998) Art and Agency: An Anthropological Theory. Oxford: Clarendon Press.
Goody J (1987) The Interface between the Oral and the Written. Cambridge: Cambridge University
Press.
Goody J and Watts J (1968) Literacy in Traditional Society. Cambridge: Cambridge University
Press.
Hanks WF (2005) Explorations in the deictic field. Current Anthropology 46(2): 191220.
Hanks WF (2006) Context, communicative. In: Encyclopdia of Language and Linguistics, 2nd
edn, section editor Mey J. Amsterdam: Elsevier, 115128.
Houseman M and Severi C (1998) Naven, or the Other Self: A Relational Approach to Ritual
Action. Leiden: Brill Academic Publishers.
Humphrey C and Laidlaw J (1996) The Archetypal Actions of Ritual. Oxford: Oxford University
Press.
Lalande A (1926) Dictionnaire de la Philosophie. Paris: Presses Universitaires de France.
Lamp F (1996) Art of the Baga: A Drama of Cultural Reinvention. New York: Prestel Publications.
Landaburu J (2007) La modalisation du savoir en langue Andoke. In: Landaburu J, Guentchva Z
(eds) LEnonciation Mdiatise. Louvain: Peeters, 2347.
Landaburu J and Guentchva Z (2002) Modalits Epistemiques. Paris: Peeters.
MacGaffey W (1986) Religion and Society in Ccentral Africa. Chicago: University of Chicago
Press.
Roth M and Salas C (2001) Disturbing Remains: Memory, History, and Crisis in the Twentieth
Century. Los Angeles: Getty Publications.
Settis S (1986) Continuit, distanza, conoscenza. Tre usi dellantico. In: Settis S (ed.) Memoria
dellAntico nellArte Italiana, Vol. III. Dalla Tradizione allArcheologia. Turin: Einaudi,
373486.
Severi C (2008) Autorit sans auteur: Formes de lautorit dans les traditions orales. In: Compagnon
A (ed.) De lAutorit, Colloque Annuel du Collge de France. Paris: Odile Jacob, 93123.
Severi C (2015[2007]) The Chimera Principle: An Anthropology of Memory and Imagination
trans. N Lloyd, foreword D Graeber. Chicago: Hau Books/Chicago University Press.
Zw Ngema (1972) Un Mvet: Chant pique Fang recueilli par Herbert Pepper, rdit par Paul et
Paule de Wolf. Paris: Armand Colin.

Author biography
Carlo Severi is Directeur dtudes at the cole des Hautes Etudes en Sciences Sociales and
Directeur de recherche at the CNRS. A member of the Laboratoire danthropologie sociale of the
Collge de France since 1985, he has been a Getty Scholar at the Getty Institute for the History of
Art and the Humanities in Los Angeles (19941995) and a Fellow of the Wissenschaftskolleg in
Berlin (20022003). In 20122013, he was elected to a Visiting Fellowship at Kings College,
Cambridge. Among his books are: Naven or the Other Self (Brill, 1998, with M Houseman),
Limage rituelle (LHerne, 2014, with Carlos Fausto) and The Chimera Principle: An Anthropology
of Memory and Imagination (Hau Books/Chicago University Press, 2015).

Downloaded from mcu.sagepub.com at Australian National University on August 29, 2016