Barthes, the syntagmatic

The syntagm presents itself in the form of a 'chain' (the flow of speech, for example). Now as we have seen [earlier in the Elements], meaning can arise only from an articulation, that is, from a simultaneous division fo the signifying layer,and the signified mass: language is, as it were, that which divides reality [...]. Any syntagm therefore gives rise to an analytic problem: for it is at the same time continuous and yet cannot be the vehicle of a meaning unless it is articulated. How can we divide the syntagm? This problem arises again with every system of signs: in the articulated language, there have been innumerable discussions on the nature [...] of the word, and for certain semiological systems, we can here foresee important difficulties. [That is, the problem is not just in linguistics--it also has pertinence for semiology.] True, there are rudimentary systems of strongly discontinuous signs, such as those of the Highway Code, which, for reasons of saftey, must be radically different from each other in order to be immediately perceived; but the iconic syntagms, which are founded on a more or less analogical representation of a real scene, are infinitely more difficult to divide, and this is probably the reason for which these systems are almost always duplicated by articulated speech (such as the caption of a photograph) which endows them with the discontinuous aspect which they do not have.
-Elements of Semiology, III.2.2

For some reason this is a clearer explanation to me of why photographs need captions than all the others I hear (regarding that famous thesis people attribute to Barthes, citing his essays on photography). Barthes then says the following about the syntagm--a very concise and provocative formulation:

In spite of these difficulties, the division of the syntagm is a fundamental operation, since it must yield the paradigmatic units of the system: it is in fact the very definition of the syntagm, to be made of a substance which must be carved up.