"Social structure." in Anthropology today: Selections. Edited by Sol Tax, 321-350. Chicago: University of Chicago Press. 1962 
Included here are some of the most famous and challenging quotes from this work. They have been some of the most influential, and now possibly most misunderstood lines from LÚvi-Strauss's work. You should also investigate the discussion of the orginal version of this paper by the main anthropolgoists of the time in Appraisal of Anthropology Today (Tax 1953: 108-118). It includes extracts from a letter from Radcliffe-Brown to LÚvi-Strauss sharpening the differences between their use of the term "social structure."
Passing now to the task of defining "social structure," there is a point which should be cleared up immediately. The term "social structure" has nothing to do with empirical reality but with models which are built up after it. This should help one to clarify the difference between two concepts which are so close to each other that they have often been confused, namely, those of social structure and of social relations. It will be enough to state at this time that social relations consist of the raw materials out of which the models making up the social structure are built, while social structure can, by no means, be reduced to the ensemble of the social relations to be described in a given society.' Therefore, social structure cannot claim a field of its own among others in the social studies. It is rather a method to be applied to any kind of social studies, similar to the structural analysis current in other disciplines.
The question then becomes that of ascertaining what kind of model deserves the name "structure." This is not an anthropological question, but one which belongs to the methodology of science in general. Keeping this in mind, we can say that a structure consists of a model meeting with several requirements.
First, the structure exhibits the characteristics of a system. It is made up of several elements, none of which can undergo a change without effecting changes in all the other elements.
Second, for any given model there should be a possibility of ordering a series of transformations resulting in a group of models of the same type.
Third, the above properties make it possible to predict how the model will react if one or more of its elements are submitted to certain modifications.
Finally, the model should be constituted so as to make immediately intelligible all the observed facts.'
These being the requirements for any model with structural value, several consequences follow. These, however, do not pertain to the definition of structure, but have to do with the chief properties exhibited and problems raised by structural analysis when contemplated in the social and other fields.
A structural model may be conscious or unconscious without this difference affecting its nature. It can only be said that when the structure of a certain type of phenomena does not lie at a great depth, it is more likely that some kind of model, standing as a screen to hide it, will exist in the collective consciousness. For conscious models, which are usually known as "norms," are by definition very poor ones, since they are not intended to explain the phenomena but to perpetuate them. Therefore, structural analysis is confronted with a strange paradox well known to the linguist, that is: the more obvious structural organization is, the more difficult it becomes to reach it because of the inaccurate conscious models lying across the path which leads to it.
It seems that both the reality and the autonomy of
the concept of culture could better be validated if culture were treated, from
an operational point of view, in the same way as the geneticist and demographer
treat the closely allied concept of "isolate." What is called a "culture"
is a fragment of humanity which, from the point of view of the research at hand
and of the scale on which the latter is carried out, presents significant discontinuities
in relation to the rest of humanity. If our aim is to ascertain significant
discontinuities between, let us say, North America and Europe, then we are dealing
with two different cultures; but should we become concerned with significant
discontinuities between New York and Chicago, we would be allowed to speak of
these two groups as different cultural "units." Since these discontinuities
can be reduced to invariants, which is the goal of structural analysis, we see
that culture may, at the same time, correspond to an objective reality and be
a function of the kind of research undertaken. Accordingly, the same set of
individuals may be considered to be parts of many different cultural contexts:
universal, continental, national, regional, local, etc., as well as familial,
occupational, religious, political, etc. This is true as a limit; however, anthropologists
usually reserve the term "culture" to designate a group of discontinuities
which is significant on several of these levels at the same time. That it can
never be valid for all levels does not prevent the concept of "culture"
from being as fundamental for the anthropologist as that of "isolate"
for the demographer. Both belong to the same epistemological family. On a question
such as that of the positivistic character of a concept, the anthropologist
can rely on a physicist's judgment; it is Niels Bohr who states that "the
traditional differences of [human cultures] in many ways resemble the different
equivalent modes in which physical experience can be described." (p. 295-6)