1. Structural analysis and its ‘closed’ universes

In The Elementary Structures of Kinship, Claude Lévi-Strauss famously equated kinship structures with the three forms of cross-cousin marriage: restricted exchange, or marriage of the male Ego with the bilateral cross-cousin (MBD/FZD); ‘long cycle’ generalized exchange, or marriage with the matrilateral cross-cousin (MBD); and ‘short cycle’ generalized exchange, or marriage with the patrilateral cross-cousin (FZD).^{Footnote 1} But there is another, equally well-known form of marriage between cousins: that of marriage with the patrilateral parallel cousin (FBD), also known colloquially as ‘Arab marriage’ – which is not in fact limited to the Middle East, where it has primarily been studied, nor to Islamized countries, but extends across various parts of Africa completely untouched by Islamic influences. Why did Lévi-Strauss scarcely even mention this form of exchange in the more than 600 pages of The Elementary Structures? One need only read the definitions set out by Lévi-Strauss at the beginning of his preface to the first edition of the book (Reference Lévi-Strauss, Bell, von Sturmer and Needham1969: xxiii) to realize that both cross-cousin marriages and marriage with the patrilateral cousin fully fit the criteria of elementary structures, as systems that prescribe or give preference to marriage with a particular type of kin – that is, those systems that divide kin ‘into two categories, viz., possible spouses and prohibited spouses’. Why, then, did he exclude ‘Arab marriage’ from the category of elementary structures, given that it undeniably contains a definition of which kin types are potential, if not preferred, spouses? If one wished to provoke, FBD marriage could even be considered the virtual prototype of an elementary structure as it applies to marriage between kin, since FBD marriage occurs with a very strictly defined type of kin – that is between the Ego and the daughter of the father's brother, herself being part of the same lineage as the Ego.

If Lévi-Strauss did not discuss FBD marriage, it was certainly not out of ignorance, but rather because the introduction of this type of marriage into the category of elementary structures would have upset the ‘logic’ of his discourse, which, beginning from the prohibition on incest as the motor of matrimonial exchange, saw the marriages of cross-cousins as the only formulae capable of creating alliances between different social groups. FBD marriage obeys another logic, that of unity and solidarity within a lineage that contains both the Ego and his spouse, and the father of the Ego as well as the father of his wife.^{Footnote 2} But the fact that it obeys another logic is not a sufficient reason to erase it from the category of elementary structures as previously defined by Lévi-Strauss.

The silence to which Lévi-Strauss condemned this type of marriage in The Elementary Structures of Kinship – really, its repression – indicates an approach typical of his structuralism: that is, the creation of various ‘closed’ universes dominated by a single logic (that of alliance and exchange, in this instance), in which a determined number of possible alternatives are identified, such as the three types of cross-cousin marriage (MBD/FZD, MBD, FZD). These alternatives are discrete, but they refer to each other through simple inverse relationships. Structural analysis measures the connections among possible alternatives, the structure consisting of the logical passages between them. For Lévi-Strauss, structure does not consist of any particular one of these specific models; the structure is never that of a local system, but rather can be found ‘across’ the systems, since it constitutes a bundle of transformations that binds them all together. But in order for them ‘all’ to be bound together, they must constitute a ‘closed’ universe, which is what ‘elementary structures’ are. They are opposed, of course, by complex structures, which from the outset are defined by the presence of an opening in the system:

The opening in complex systems clearly derives from the lack of prior ‘determination’ of the spouse: the system ‘opens’ itself up to other mechanisms, other logics, other factors, allowing them to intervene regardless of the principles of alliance or matrimonial exchange. Pier Giorgio Solinas has rightly compared the opposition between elementary and complex structures – as they were defined by Lévi-Strauss – to the opposition between ‘closed’ and ‘open’ systems as defined by Ludwig von Bertalanffy (Reference SolinasSolinas, 2004: 162–163). But, significantly, Lévi-Strauss’ structural analysis stops at the threshold of complex structures. He foreshadows the existence of an intermediate category – what Reference HéritierFrançoise Héritier (1981) would call ‘semi-complex’ structures – those Crow-Omaha systems that act as a ‘connecting link’ between elementary and complex structures, forms that are hybrid, ambiguous, or a ‘compromise’ (Reference Lévi-Strauss, Bell, von Sturmer and NeedhamLévi-Strauss, 1969: xxxix, xli). Lévi-Strauss thus approaches complex structures, but does not delve into them. And the fact that he had declared his intention to avoid extending his structural analysis to complex structures (Reference Lévi-Strauss, Bell, von Sturmer and NeedhamLévi-Strauss, 1969: xxiv) is very indicative: structural analysis – as conceived and practiced by Lévi-Strauss, particularly on the subject of kinship and marriage – in fact requires a ‘closed’ universe, one dominated by a single principle, in which models or systems are all determinable and relatable to each other. When faced with ‘complexity’, Lévi-Strauss’ structural analysis beats a hasty retreat, and does so because otherwise it would be forced to recognize that something eludes it; that the field of investigation is, despite its best efforts, indeterminate, and that its analysis is thus incomplete and imperfect.

2. FBD marriage: Lévi-Strauss’ defensive strategy

In 2000, the French journal L’Homme dedicated a special double issue entirely to the Question de parenté. At least three of the articles published (Barry, Bonte, and Conte) discussed Lévi-Strauss’ ‘silence’ (mutisme) regarding ‘Arab marriage’ (Reference BarryBarry, 2000: 69). After having described the vast geographic spread of this type of marriage, and the significance of the societies that practice it (from the Middle East to Africa and Madagascar), Laurent S. Barry declared himself in agreement with Claude Lefébure, who argued that FBD marriage constituted a veritable ‘challenge’ (défî) to the view outlined by Lévi-Strauss in The Elementary Structures of Kinship (Reference Lefébure, Breteau, Lacoste-Dujardin and LefébureLefébure, 1981: 195–196). Lefébure contrasted exogamous marriages with endogamous marriages, like that of FBD, which can be neither ignored nor left unspoken, nor marginalized. The importance of this type of marriage pushed Lefébure to contest the fundamental value and usefulness generally assigned to The Elementary Structures of Kinship (Reference Lefébure, Breteau, Lacoste-Dujardin and LefébureLefébure, 1981: 195–196). Along the same lines, Pierre Bonte argued that by now simply too much data had been accumulated which eluded the scheme laid out by The Elementary Structures:

For Bonte, too, cross-cousin marriages were not the only elementary structures of kinship. Such structures did not obey solely the logic of exchange and reciprocity, and parallel cousin marriage – by raising the question whether there were other logics behind marriage (Reference BonteBonte, 2000: 49) – introduced a plurality of principles and criteria into the universe of elementary structures. In other words, the recognition of FBD marriage transformed the category of elementary structures of kinship from a closed to an open universe, and compelled structural analysis to accept logics that were different and even divergent, if not in outright conflict.

In his Postface to the aforementioned issue of L’Homme, Lévi-Strauss admitted that the writings of Barry, Bonte, and Conte were ‘among the highlights of this compilation’ (Reference Lévi-StraussLévi-Strauss, 2000: 716). But, beyond reiterating the always relative character of the distinction between exogamy and endogamy, he only went so far as to mention a ‘tendency to turn inward’ (tendance au repliement) in marriage selection (Reference Lévi-StraussLévi-Strauss, 2000: 716–717), a tendency which he had always acknowledged in marriage dynamics and which still explained structurally secondary phenomena, ‘anomalies’ and ‘exceptions’ in an overall picture dominated primarily by the logic of reciprocity and exchange.

Is it true, then, that Lévi-Strauss tried to avoid tackling the question of so-called Arab marriage? While in the case of The Elementary Structures it might be difficult not to speak of repression, in 1959 he did explicitly address the subject.^{Footnote 3} On that occasion, he did not hesitate to argue that marriage with the patrilateral paternal cousin (FBD) constituted part of a territory – that of kinship and marriage in Muslim societies – considered almost ‘taboo’ by ethnologists, a sort of ‘forbidden kingdom’ at whose borders theoreticians of kinship systems came to a ‘halt’; more specifically, that type of marriage represented ‘a sort of exception, almost a sort of scandal’, exhibiting an ‘abnormal character, in regard to our classifications’, even a ‘sort of aberration’ from the point of view of the logic of exchange (Reference Lévi-StraussLévi-Strauss, 1988–89: 11–13). Rather than try to fit FBD marriage into the category of elementary structures of kinship, where it would have provoked a drastic reformulation of his theoretical apparatus, Lévi-Strauss interpreted FBD marriage as having a primarily political value. In so doing, he endowed this type of marriage with a significant degree of meaning in the sphere of ‘the history of human society’, while subtracting it from the realm of kinship systems: FBD marriage led us, in fact, ‘from the theory of kinship systems to the history of civilizations’ (Reference Lévi-StraussLévi-Strauss, 1988–89: 22–23). In order to liberate his theory of kinship systems from the irritating and troublesome existence of ‘Arab marriage’, Lévi-Strauss threw himself feverishly into outlining a vast historical movement, in which a number of societies – ‘from India to Egypt and Ancient Greece’ – supposedly ‘detached themselves from the common destiny’, that of matrimonial exchange, in order to practice ‘marriage between close kin’. FBD marriage was nothing other than ‘a sort of variation’ on the theme of endogamy, which characterized societies from the Ganges to the Mediterranean (Reference Lévi-StraussLévi-Strauss, 1988–89: 23). In so doing, Lévi-Strauss adopted a clearly defensive strategy: consigning FBD marriage to history, he thought he might keep the ‘closed’ universe of his own elementary structures intact. But, as we have seen, his maneuver was simply not credible. Today less than ever.

3. The ‘open’ sea of the family and structural bewilderment

Lévi-Strauss’ structuralism requires the establishment of a ‘closed’ universe, but when it finds itself faced with an ‘open’ sea it becomes theoretically confused and uncertain. In an essay written in Reference Lévi-Strauss and Shapiro1956, later reprinted in The View from Afar (Reference Lévi-Strauss, Neugroschel and Hoss1992: 39–62), Lévi-Strauss grappled with the issue of the family. But it is obvious that here Lévi-Strauss was out of his element. As a theoretician of marriage and matrimonial exchange, he was also a proponent of the idea that families are secondary or derivative with respect to the relationships that connect them (Reference Lévi-Strauss, Jacobson and Grundfest SchoepfLévi-Strauss 1963: 50–51); but unlike Reference Radcliffe-BrownAlfred R. Radcliffe-Brown (1952) and Reference MurdockGeorge Peter Murdock (1949), Lévi-Strauss did not formulate a theory of the family built upon a common foundation, an omnipresent ‘atom’ on which all societies depend. Lévi-Strauss’ atom of kinship does not coincide with the family, but with the relationships created by matrimonial exchange between families. For Murdock, in contrast, the atom of kinship was the nuclear family itself, in the sense that any other type of family derived from the potential molecular combinations already present in this nucleus. With this unified theory, Murdock managed to present a very clear and straightforward portrait of family organization, something we do not find in Lévi-Strauss. Without a unified theory at his disposal, the subject of the family was dominated by a plurality and even an extravagance of forms: the result is a fundamentally ‘open’ universe. Unable to control this plurality, to transform it into variants of a single structure, Lévi-Strauss’ structural analysis is tripped up, becomes vague and indecisive, conveying a certain degree of confusion.

Conscious of the multiplicity that characterizes the field of the family, Lévi-Strauss first questioned the very idea of the family as ‘a union, more or less durable, but socially approved, of two individuals of opposite sexes who establish a household and bear and raise children’, understood to be ‘a practically universal phenomenon, present in every type of society’ (Reference Lévi-Strauss, Neugroschel and HossLévi-Strauss, 1992: 40–41). To give weight to this critical stance, Lévi-Strauss turned to the case of the Nayar of Malabar (Southern India), where marriage was a ‘purely symbolic ceremony’ (Reference Lévi-Strauss, Neugroschel and HossLévi-Strauss, 1992: 40–41) that did not lead to the creation of a family, and where the woman's children were raised in a home run by the mother's brothers and sisters. Unable at the time he revised his essay to consult more recent ethnographic case studies – one of the most recent is clearly that of the Na of Yunnan (China), studied by Reference Cai and HustvedtCai Hua (2001) – Lévi-Strauss grasped that they represented ‘an extreme form of a tendency that is far more frequent in human societies than is generally believed’ (Reference Lévi-Strauss, Neugroschel and HossLévi-Strauss, 1992: 41). But what tendency was this, exactly? Lévi-Strauss clarified himself immediately: ‘Without going as far as the Nayar, some human societies restrict the role of the conjugal family’ (Reference Lévi-Strauss, Neugroschel and HossLévi-Strauss 1992: 41 – italics mine). As we have already seen with regard to ‘Arab marriage’, Lévi-Strauss privileges one category as central above all others, and considers anything that conflicts with this view to be marginal or reductive. Alternative patterns to the conjugal family – even when considered from a perspective that leaves room for plurality – thus come to assume the guise of ‘bizarre institutions’ without their own authentic structural autonomy. In Lévi-Strauss’ view, the Nayar take the form of the most extreme, systematic example of a tendency to diminish the importance or, in this case, even the mere existence of the conjugal family (Reference Lévi-Strauss, Neugroschel and HossLévi-Strauss 1992: 41). But this tendency makes no sense if not in reference to this type of family: it is a tendency to ‘restrict’ the field of an institution that must necessarily be taken for granted (Reference Lévi-Strauss, Neugroschel and HossLévi-Strauss 1992: 41). He seems to say, then, that the plurality of family types – which Lévi-Strauss truly wants to acknowledge – is nothing other than the plurality of ways in which societies both recognize and misunderstand (altering, or even annihilating) the conjugal family.

Having engaged in such an argument (that societies can do without the conjugal family), Lévi-Strauss immediately feels the need to ‘define the family’; he does so not through induction, but rather ‘by constructing a model reduced to a few invariable properties, or distinctive characteristics, that a rapid survey has allowed us to discern’:

1. 1. The family originates in marriage.

2. 2. It includes the husband, the wife, and the children born of their union, forming a nucleus around which other relatives can eventually gather.

3. 3. The members of the family are united among themselves by:

   1. a. Legal bonds.

   2. b. Rights and obligations of an economic, a religious, or some other nature.

   3. c. A precise framework of sexual rights and prohibitions, and a variable and diversified group of feelings, such as love, affection, respect, fear, and so on. (Reference Lévi-Strauss, Neugroschel and HossLévi-Strauss, 1992: 44)

After distancing himself from Murdock and Radcliffe-Brown with regards to their notion of the atom of kinship, it is disconcerting, to say the least, that Lévi-Strauss can think of nothing better than to repropose their very same definition of the ‘elementary’ or ‘nuclear’ family. The difference is that in Lévi-Strauss, families are not the building blocks that make up societies (Reference Lévi-Strauss, Neugroschel and HossLévi-Strauss, 1992: 61), since there is tension between family and society: every society, in fact, encourages the ‘incessant work of destruction and reconstruction’ of the family (Reference Lévi-Strauss, Neugroschel and HossLévi-Strauss, 1992: 60). In some cases, society reinforces and glorifies the families, giving them greater temporal stability; in others, it reduces them almost to the point of extinction:

This ‘halfway measure’, in between the tendency toward expansion and stabilization of the family unit on one side, and the countervailing tendency toward its contraction on the other, shows us that, for Lévi-Strauss, the family is a fundamental model that societies can only manipulate. That this conception of the family is rather unconvincing, however, becomes clear when Lévi-Strauss turns to consider the matrimonial unions of the Chukchee of Siberia, where adult women marry male children, or the Mohave of California, where it is the adult man who marries a young girl: ‘forms’, he argues, ‘that we would doubtless not be the only ones to judge incompatible with the aims that human beings conceive as the basis of the household’ (Reference Lévi-Strauss, Neugroschel and HossLévi-Strauss, 1992: 50). If we are not faced with only these two simple cases, two mere exceptions or anomalies, if ‘[s]imilar cases are known in the Andean and the tropical regions of South America and also in Melanesia’ (Reference Lévi-Strauss, Neugroschel and HossLévi-Strauss, 1992: 50), would it not be a sound idea to reconsider the overarching model with which the anthropologist seeks to define ‘the’ family, rather than dive into the perilous question of incompatibility with the goals that human beings have set themselves? Elsewhere in this rather troubled essay, Lévi-Strauss is forced to admit that ‘[i]t would […] be wrong to approach the study of the family in a dogmatic spirit,’ and that ‘[a]t each instant, the object that one thinks is in one's hands slips away’ (Reference Lévi-Strauss, Neugroschel and HossLévi-Strauss, 1992: 43). Well more than halfway through, he concedes ‘we do not yet know what exactly the family is’, even while at the same time that he claims to be able to perceive the ‘laws’ of its reproduction (Reference Lévi-Strauss, Neugroschel and HossLévi-Strauss, 1992: 55).

4. A family resemblance: Wittgenstein's contribution

How, then, can we come up with an anthropological definition of the family? The philosopher Ludwig Wittgenstein, who never concerned himself with the topic, comes to our aid with his image of ‘family resemblances’, a concept whose epistemological implications (Reference RemottiRemotti, 2009: 204–212) we might attempt to apply to the subject that interests us here.

1. 1. Plurality of criteria by which similarities and differences can be seen among various forms of the family, just like – for Wittgenstein – among various types of games. This plurality avoids necessarily reducing all families down to a fundamental nucleus. It should be noted that already by 1936 Ralph Linton, in his distinction between the ‘conjugal’ and ‘consanguine’ family, had avoided the concept of a common denominator (Reference LintonLinton, 1936: 159) – a direction not followed by the majority of theoreticians of the family.

2. 2. Observation, before categorization. Wittgenstein's injunction ‘Don’t think, but look!’ (Reference Wittgenstein and AnscombeWittgenstein, 1953: 31), for example, amounts to a mistrust of the move we have seen Lévi-Strauss make: the a priori elaboration of a ‘model reduced to a few invariable properties’ which, far from being heuristically effective, reveals itself instead to be a mistake.

3. 3. Openness of concepts. Anthropological concepts – like that of the family, for example – must be kept structurally open, and also accept the variations that ethnographic observation continually presents us with. They are bundles of relations, for which there are no exceptions or anomalies (such as, for instance, marriage between adults and children, which Lévi-Strauss was forced to discuss in terms of their ‘incompatibility’ with the general model of the family).

4. 4. Transformation and Adaptability of concepts. This is a consequence of their openness: The ability to welcome variations and new or unforeseen forms means the internal structure of the concepts themselves transforms in a more or less consistent way. The anthropological concept of the family cannot avoid undergoing a profound transformation as a result of, for example, the revelation of cases like that of the Nayar.^{Footnote 4}

5. 5. Resistance. Anthropological concepts face resistance – and it is good that they do; concepts cannot be completely open, or they would cease to be concepts, and would thus lose even the value of ethnographic innovation. But the rigidity of concept-categories, like the invariant model frequently invoked by Lévi-Strauss, is one thing; the conceptual structure that is formed through the accumulation of anthropological data and knowledge, and is thus by definition open to change, is quite another. There is a point of equilibrium between the internal organization of concepts (intension) and continually observed and accumulated phenomena (extension). The latter always requires a greater or lesser degree of revision, or even the elimination, of the concepts themselves (Reference RemottiRemotti, 2009: 206–207).

6. 6. The Permanence of Ethnography. From an epistemological point of view, Lévi-Strauss was compelled to move away from ethnography, as a sort of factual base, in order to proceed toward more generalized abstraction. From the perspective inspired by Wittgenstein, ethnography remains a limitless source for anthropological concepts, which change as they absorb ethnographic case studies or are abandoned because they are deemed incapable of accommodating its data (Reference RemottiRemotti, 2009: 206–207). I would argue that one of the concepts to be abandoned is that of the ‘elementary’ or ‘nuclear’ family (of Radcliffe-Brown and Murdock, respectively), as well as the reductive model put forward by Lévi-Strauss.

7. 7. Network of Connections. With Wittgenstein's family resemblance, anthropology takes the form of a network of connections: ‘a complicated network of similarities overlapping and criss-crossing’ (Reference Wittgenstein and AnscombeWittgenstein, 1953: 32). With respect to the family, anthropology constructs a web that collects the various forms of the family and, like a network, allows us to move from one form to another, both synchronically and diachronically.

8. 8. Arbitrariness of Borders. It is perhaps not difficult to trace the boundary between one form of the family and another (between polygynic and polyandric families, for instance), but where the limit of the general concept of the family lies is a question requiring an epistemological decision. ‘What still counts as a game, and what no longer does?’ (Reference Wittgenstein and AnscombeWittgenstein, 1953: 33); or, what constitutes a family, and what does not, for anthropologists? This requires an assumption of epistemological responsibility. For Wittgenstein, concepts are structurally ‘open’; yet he also teaches us that they must be provisionally ‘closed’. A concept is an area, and thus needs limits; but Wittgenstein implies that these limits can be understood and drawn in different ways. It is we who decide where to draw these lines, and it is we who decide if the borders are neat or porous: ‘Isn’t the indistinct one often exactly what we need?’ (Reference Wittgenstein and AnscombeWittgenstein, 1953: 34).

What does Wittgenstein have to do with Lévi-Strauss? Why turn to the Austrian philosopher's concept of family resemblances to try to resolve the aporiae of the French anthropologist's structuralism? It was Ernst Gellner who first highlighted the affinity between structuralism and Wittgenstein's theory of family resemblances (Reference GellnerGellner, 1985: 138). Both concepts strive to identify networks of connections between similar and comparable phenomena; to build a transversal body of knowledge that allows us to overcome morphological and cultural boundaries; and to create a dialogue between cases that, despite their apparent (historical, geographical, or cultural) extraneity, still demonstrate a certain air de famille. There is, however, one fundamental difference: Wittgenstein's family resemblance does not share the systemic ambitions of Lévi-Strauss’ structuralism; in particular, it does not require or admit a ‘closed’ universe of possibilities. Its conceptual ‘closedness’ – while necessary and unavoidable – is always provisional; beyond these provisionally established boundaries, there is an awareness of still unexplored possibilities. Translated into the world of anthropology, Wittgenstein's family resemblances contrast with both the typological, classificatory structuralism of Murdock, in that they perforate or even explode categorical boundaries, as well as the structuralism of Lévi-Strauss, to the degree in which the latter aims to grasp the generative principles of structural possibilities. There are thus many different ways of connecting anthropological phenomena, ‘more than what typologies, on the one hand, or the generative capacity of structural principles, on the other, allow us to predict’ (Reference RemottiRemotti, 2009: 203).

The maneuver illustrated here is something different with respect to the way Wittgenstein's work has been used in anthropology to date. Rodney Needham, drawing on Wittgenstein's family resemblances, has tried to shed light on the imprecision and vagueness of concepts employed in anthropology. To avoid getting lost in a jumble of local knowledge, resistant to all comparison, the anthropologist must rely upon a ‘vocabulary’ consisting of ‘purely formal’ analytical concepts able to express the ‘formal properties’ and ‘universals’ of the human mind (Reference NeedhamNeedham, 1983: 62). For Needham, family resemblances serve only to illuminate the unreliability of anthropological knowledge, to the extent it is based on local categories and cultural concepts. To save anthropology, we have to let go of these resemblances and opt for a ‘style of radical abstraction’ necessary to understand ‘the primary factors of experience’ (Reference NeedhamNeedham, 1983: 65).

Is it better to move on from family resemblances, then, or rather from anthropology itself, and its dreaded quicksands? It is no coincidence that Lévi-Strauss himself had foreshadowed his shift toward a sort of bio-psychology: ‘Ethnology is first of all psychology,’ he wrote; it is the ‘first step in a procedure’ that attempts to uncover the ‘invariants’ found ‘beyond the empirical diversity of human societies’ – that is, in its natural substrate, the brain. At the very same time, Lévi-Strauss had argued that ‘the ultimate goal of the human sciences [is] not to constitute, but to dissolve man’ (Reference Lévi-StraussLévi-Strauss, 1966: 131, 247). It is likewise no coincidence that Murdock, in a moment of ruthless self-criticism of his own work, had proposed a naturalistic solution in almost perfect parallel with that of Lévi-Strauss (Reference MurdockMurdock, 1972). Is this not the path that other illustrious anthropologists have, more recently, come to pursue: a sort of road to salvation, an escape from cultures, and the insidious customs of which they are made, to take refuge in laws and structures that are pre-cultural, or at least decidedly natural? For Dan Sperber, following on the heels of Needham, the reference to Wittgenstein and family resemblances is an obligatory step, necessary only to object that anthropological discourse, as it is generally employed, responds more to the demands of interpretation than to a basic theoretical commitment, and that its terms do not, for the most part, correspond to ‘precise concepts, but rather to “polythetic” notions, that is classes of phenomena having no more in common than a “family resemblance”’ (Reference SperberSperber, 1985: 24). Departing from our prior statement, regarding the close, organic nexus between the elaboration of anthropological concepts and ethnography, Sperber invokes a ‘divorce’ leading to their separation (Reference SperberSperber, 1985: 10): ethnography, tied to research in the field, must content itself with vague interpretive concepts (family resemblances, for example), while anthropology, explanatory and comparative, must look for the laws of human nature. The psycho-neuro-biological sciences are there just waiting for anthropologists, who, disillusioned by the tortuous, futile contortions of ethnography, aspire to more robust and steady generalizations.

There are those, on the other hand, who have privileged ethnography, deliberately abandoning all pretensions to anthropological generalization along the way. Clifford Geertz, for instance, understands anthropology to coincide precisely with ethnography. Here, too, we find Wittgenstein as a traveling companion. But whereas in the cases of Needham and Sperber the Wittgenstein employed is that of family resemblances, who shows anthropologists the almost irreparable imperfection of their discourse, the Wittgenstein deployed by Geertz has a much more reassuring function: ‘They have, wonder of wonders, been speaking Wittgenstein all along’ (Reference GeertzGeertz, 1983: 4) – not the Wittgenstein of family resemblance, but rather that of ‘forms of life’ (Lebensformen), that is of those sufficiently closed social and cultural worlds in which languages, customs, and behaviors acquire their meaning. This is the Wittgenstein that anthropologists have by and large turned to, as demonstrated in Veena Das’ essay on ‘Wittgenstein and anthropology’ (Reference Das1998), in which the notion of ‘forms of life’ plays a central role, while the concept of family resemblances is not even mentioned, much less understood as heuristically effective.

In his discussion of families, and in particular same-sex families (familles homoparentales), the sociologist Éric Fassin turned in systematic fashion to Wittgenstein's family resemblances. Anticipated by Reference Bourdieu and NicePierre Bourdieu (1990), Fassin argues that ‘the philosopher's contribution can be decisive for the social sciences’, and especially anthropology, in that ‘rather than invariants, anthropology could, by modeling itself after Wittgenstein, take as its subject the grammar of social usages’ (Reference FassinFassin, 2000: 402–403). This grammar allows us to see how social actors choose their own ‘definition’ of family; in other words, how they determine their ‘family of families’, widening or constricting the family space – including in their notion of family those formed by homosexual couples, for instance, as is the case in France (Reference FassinFassin, 2000: 405–406). For Fassin, the sociologist and the anthropologist need not ask themselves what constitutes a family, much less look for a common denominator or indivisible nucleus (noyau dur): their task is to investigate how different societies modify the family space, their own family of families. From this perspective, it is interesting to register Fassin's criticism of Françoise Héritier and her insistence on arguing that the difference between the sexes is an unavoidable, ineluctable fact, going so far as to make it the foundation of her ‘second fundamental law of kinship’ (Reference HéritierHéritier, 1981: 15–16, 50); civil unions (the Pacte civile de solidarieté, or PaCS) constitute a denial of this idea (Reference FassinFassin, 2000: 398–401). This is a criticism that, looking more closely, could also be made justifiably of Lévi-Strauss’ model of familial invariants (see section 3 above), which would allow us not only to shed the obligation to search for a definition of the family according to a nucleus, but also to study ‘the transformations in the definition of the family’ that societies themselves show us (Reference FassinFassin, 2000: 406). In essence, the question with which we began this paragraph, for Fassin, makes no sense: it is societies themselves that are responsible for defining what for them constitutes a family, and to change their ideas, if so desired, over time. Anthropologists and sociologists need only track these changes and describe them: It is a dangerous error to assume the task of indicating their limits, however large they might be, and to anticipate social change or attempt to teach societies how they should act (Reference FassinFassin, 2000: 403).

5. Anthropological ‘imperfectionism’

I am not convinced that anthropologists – or likewise sociologists – can free themselves completely from their definitional responsibilities. The idea that social scientists can act like the owl of Minerva, who, in Hegel's description, takes flight at dusk, after events have taken their course, emanates from Fassin's distinction between the citizen (citoyen) and the scholar (savant), and the dictum that the savant may not enforce its law upon the citoyen. How can one not share this principle described by Lévi-Strauss, which Fassin repeats in his article: ‘The choice of society does not belong to the savant as such, but – in so far as he is one – to the citoyen’ (Reference FassinFassin, 2000: 392)?

However, the anthropologist does not remain confined to only one society, and he does not register only the changes that occur in ‘his’ society regarding definitions of the family. If he were to do so, if he were to become an expert in only ‘one’ society, the one he had decided to study – his own, or his elective society, so to speak – why should he call himself an anthropologist? How would he be any different from a sociologist? The expertise of the anthropologist consists of the accumulation of knowledge of many local traditions, many diverse definitions and representations of the family, many different families and ‘families of families’. Certainly, Wittgenstein encourages (or, rather, we understand him to encourage) the anthropologist to keep a low epistemological profile.

So far, so good. I, too, have underlined the epistemological assistance provided by Wittgenstein regarding the definition of family (Reference RemottiRemotti, 2008: 102). But, so long as one looks at one society (again, whether his own or the elective), the ‘other’ games – or families – that we can describe do not pose a problem: If France in the 1990s deliberated over PaCS, the savant can truly describe a certain number of forms of family which took shape in that society (the form of the heterosexual family, for example, or the same-sex form), and his description would coincide with a definition, that given by the society in question (Reference FassinFassin, 2000: 406). But if – as in the case of the anthropologist – our perspective inevitably widens, if to propose some idea of the family, to explain to someone what one means by the word family, the anthropologist has at his disposal the entire gamut of social definitions of family (at least in theory), then the problem of choice inevitably arises. It is not immaterial if next to this family ‘game’ – the one normally ‘played’ in our society – the anthropologist adds the polyandry of the Nyinba, the domestic group of the Na, or the natolocal families of the Senufo: the resulting image will be different from that springing from other choices (if, for example, one takes the nuclear family and adds the two-spirit families of the Plains Indians, or the families formed through conjugal children, as among the Chukchee or the Mohave).

But it is not just a question of the choice of cases. It is also a question of the links that join these cases, of the modalities of passage or translation between them. Wittgenstein's theory of family resemblances does not invite us to produce only a handful of different cases (beyond our own type of family, that of the Nuer, or that of the Trobrianders), nor to indicate only that the list is open (‘this and similar things …’); it also raises the issue of the similar or different elements, traits, or characteristics that we perceive among the cases: not a single characteristic, but a plurality of characteristics that appear and disappear in various forms and that generate the air de famille. The assignment, therefore, is not only to compare one case to another, but also to identify the links between them. Upon closer inspection, this is where the anthropological significance of Wittgenstein's family resemblances resides. In effect:

Our responsibility goes beyond the mere drawing of boundaries. Here Wittgenstein points out another epistemological duty: we are in fact also responsible for choosing the fibers and the method of entwining them together. Spinning, entwining, connecting; these are the activities of anthropologists (Reference RemottiRemotti, 2008: 255). However, anthropology cannot satisfy its ambitions to cover all known cases or all imaginable possibilities. As I have already argued elsewhere, anthropology's networks of connections are partial and ‘do not make up a system’, not least because they do not make up a durable, perfectly integrated network (Reference RemottiRemotti, 2009: 213). The threads hold for as long as they can, they are robust so long as they do not slacken or break. Their adaptability and provisionality are in fact due to the opening up of anthropological concepts and the modifying effect of ethnographic observation. Those ‘and similar things’ contain all the transformative and, in some cases, even subversive potential of ethnography.

I would like to conclude this essay in support of Wittgenstein's family resemblances with the conception of science proposed by Otto Neurath:

The ‘open’ sea is the context of Wittgenstein's family resemblances which, here as elsewhere (Reference RemottiRemotti, 2009), I would propose as a conceptual alternative to the ‘closed’ universes of Lévi-Strauss’ structuralism, as well as the ‘closed’ communities of Clifford Geertz. Family resemblances represent a concept capable of remedying the aporiae of structuralism, for the very fact that this concept does not only claim to be more deeply anchored in scientific practice, but is also more viable and productive.

But why, then, do anthropologists navigate the world, uproot themselves from one land to another, one island to another, thereby creating links between islands, societies, cultures? The most fruitful response to this question comes not from Wittgenstein but from Lévi-Strauss, and more specifically from his structuralism. As I have already noted, structure for Lévi-Strauss consisted precisely of the links that connect systems: ‘systems should not be treated in isolation’ (Reference Lévi-Strauss, Bell, von Sturmer and NeedhamLévi-Strauss, 1969: 155). In order to comprehend them, they must be integrated into the total of all structural possibilities – that is, the structure – of which they are part; that is, among those ‘transformations through which similar properties are recognized in apparently different systems’ (Reference Lévi-Strauss and LaytonLévi-Strauss, 1983: 18).

There is a point here that bears repeating, regarding the principle of incompleteness of the systems that characterize Lévi-Strauss’ structuralism, and it is a principle that might also apply and be extended to those cultural phenomena that we would hesitate today to call systems: cultures are in fact the fruit of shared choices and more or less conscious decisions; that is, they are always partial constructions, and need to be compared with other constructions to be understood in their particularity. The principle of incompleteness requires its exit from the system, and connection with otherness (Reference Remotti, Affergan, Borutti and CalameRemotti, 2003). When his belief in structuralism was at its fullest, Lévi-Strauss thought, however, that these connections could occur in a ‘closed’ universe, that the possibilities to be taken into consideration were limited in number. He conceived local systems as incomplete and thus open, on the one hand, and as part of a structuralist framework that contained finite possibilities and was thus closed, on the other. It is here that his structuralism failed: possibilities generally make up an ‘open’ universe, and are indeterminate in number; the list can only be considered provisionally closed, because anthropologists do not possess other data, cannot come up with broader theories, or simply do not realize that other possibilities could exist.

It is precisely at this moment that Wittgenstein's theory of family resemblances reveals its utility, allowing us to understand that anthropological knowledge must be structurally open, programmatically imperfect, incomplete, and contingent – like Neurath's boat. Instead of thinking, as Lévi-Strauss frequently did, of a sort of ‘perfect’ or ‘complete’ structure, in which all possible variables are accounted for, Wittgenstein invites us to make imperfect connections, convincing us to tolerate imperfection and even to appreciate the vagueness of concepts (the blurred lines of their boundaries), not as a sign of defects that must absolutely be corrected, but rather as the result of their reflection of real cultural phenomena.

For anthropological concepts, even those endowed with a certain transversality, getting closer to reality means ‘getting dirty’, losing the shine and clarity that Lévi-Strauss’ structuralism all too often tried to confer upon them. Getting closer to cultural context, moving through cultures and societies more slowly, means loading concepts with a semantic weight that renders them indubitably less tidy, agile, and self-confident. Combining these two principles – Lévi-Strauss’ penchant for transversality, and the acceptance that the tools of the anthropologist are like the planks in the hull of a boat that must continually be repaired and replaced, subjecting the boat itself to endless renovation – means valuing that which we might call imperfectionism, understood as both anthropology's fate and its good fortune: with use, tools not only get dirty, but they wear out; they must be constantly replaced, but their replacement is always only provisional, never definitive. Why do anthropological tools wear out so easily and so quickly? The reason is that they work not only with one type of reality, but in a multitude of situations, and that their job is not just to dig in one place, but to traverse multiple cultural boundaries and frequently to confront new and unforeseen phenomena (on Neurath's ‘open sea’). The so-called ‘imperfectionism’ of the tools of anthropology is the price, but at the same time the proof, of their use and effectiveness for a body of knowledge that can only be considered incomplete and unending, and thus by definition imperfect: a body of knowledge that never ceases to imperfect itself. The sailors both accept the state of their plight (conscious that the search for certainty is frequently a flight from reality) and yet never tire of trying to improve their vessel, knowing all too well that any renovation will at some point reveal its own leaks, and that every replacement will itself sooner or later need to be replaced.

Translated from the Italian by Richard R. Nybakken