This is an updated version of an older post, revived for students in my tutorials today.
There are infinite possible kinds or forms of long term intimate relations, or at least as many as there are possible combinations of persons alive at any given moment. It is unsurprising then, as feminist studies of kinship remind us, that there exists a great deal of diversity when it comes to socially recognised forms of intimate long-term relations. And yet pick up any ethnography, turn to the index, and there you will find the term marriage duly noted in between those things that come before and after. (Birth and death or maq– and mas-.) Flick back through to the relevant page and there you find an ethnographic description of marriage practices and marriage relations in the society or region in question.
In using these terms unproblematically anthropologists risk unwittingly translating over or effacing whatever it is that might be different, unique or idiosyncratic about socially recognised forms of intimate, long-term relations in the culture or region in question. Moreover, such habits of methodology unwittingly reproduce dominant norms when it comes to understandings and expectations of what is ‘normal, proper and right’ in intimate relations. (“You mean they don’t have marriage!?!? *imagination goes wild*)
What is marriage?
Marriage relations are contractual relations (i.e. alienated relations) between two persons recognised, conferred and regulated by the Church and/or State. The relationship between husband and wife is first and foremost an exclusive status recognised, conferred and regulated by the Church and/or State (i.e. the personal relationship existed before and outside the institution of ‘marriage,’ it did not come into existence as a result of it.) In conferring this exclusive status the centralised authority also confers certain exclusive rights to persons involved who agree, by way or means of a binding contract, to accept certain obligations in turn.
There are further defining characteristics of ‘marriage’ which include: the ceremonial ‘signing’ of said binding contract; the fact that the contract can only be terminated by the Church and/or State (i.e. a married couple cannot just ‘declare’ themselves no longer married), and crucially – the fact that the exclusive rights conferred upon persons involved include rights and interests to persons and property, which significantly affects (if not determines) socio-economic patterns of succession/inheritance (i.e. the intergenerational transmission of socio-economic rights and interests to persons and property).
This doesn’t really sound like an ahistorical, cross-culturally neutral concept or institution, does it? So why do anthropologists so often employ it as if it were? A number or combination of reasons I think, chief among which is the difficulty in describing terms and concepts that do not have direct equivalents in the language in which the anthropologist is writing. Writing ethnographic descriptions of culturally recognised forms of anything, is a lot more difficult and involved than drawing on familiar terms and concepts from one’s own language and culture. Thick description takes forever! 
From Schneider and Strathern to Keen and Morphy, many anthropologists have written about and proposed ways and means to avoid translating over social and cultural difference. The most recent I have come across and enjoyed is the Introduction in Howard Morphy’s ethnography Becoming Art: Exploring Cross-Cultural Categories (2008), where he writes, in part thus:
‘The development of a meta-vocabulary for anthropology, consisting of categories that are applicable cross-culturally, can be justified on a number of different bases. The pragmatic justification is that the meta-vocabulary is simply a formalisation of something that anthropologists do anyway. An anthropologist brings certain concepts with her or him into the field and uses them as an aid to analysing the data gathered from that society. The concept is then re-defined in terms of the categories and concepts of that society: so that gender in Melanesia, for example, is differently constructed or conceptualised than it is at a particular time and place in European history (see Strathern 1988).
It is better that anthropologists apply these terms in a considered way that reflects , even challenges the sense in which they are understood within the discipline than that they should apply them unreflexively and without definition.
A more theoretical answer is that the meta-vocabulary that anthropologists construct is indeed an acknowledgement of the existence of cultural categories that have general relevance. Not all cross-cultural categories are going to be universal; some may apply regionally or temporally. . . . To be valid in the sense in which I am using it, however it [the cross-cultural category] needs to be interrogated in the contexts where it is applied and to make sense to the people whose categories are under scrutiny. In practice such testing of the proposed category will proceed through ethnographic research . . . ‘ (2008, pp. 5-6).
In closing, a final note from Howard Morphy:
‘It is quite possible that as a result of investigation the anthropologist concludes that no such phenomenon exists in the society concerned, though it is more than likely that something will have some relationship to the phenomenon marked by the cross-cultural category. Of course it is precisely this mismatch that can result in a revision of the category itself’ (2008, p. 6).
#Occupy your own tools of analysis
 Assumed/dominant norms in the anthropologist’s culture or context (is what I mean here).
 As a cultural ceremony or ‘ritual’ most weddings follow Victor Turner’s three phases of ritual down to a tea: separation, transition/liminality and reintegration/incorporation.