Gregory Paul has come out, in 2000 actually the first time, and said that there is pretty much no way that Amargasaurus' spikes supported a sail structure; I have paraphrased grotesquely. I tend to agree with that idea, and with another part of his theory behind those spikes, but he raises one interesting idea which I find quite interesting and seems quite plausible that I, and most others probably, have not given much thought to when considering the neural spines possessed by Amargasaurus.One of the points of Paul's theory, mentioned briefly in his latest literary work The Princeton Field Guide to Dinosaurs, is that the neural spines, while used as a deterrent against large predators, could also be used to communicate. He notes that they "may have been used to create [a] clattering noise display" (page 188). Though his intent is to make the reader aware of this as an intimidation tactic, if this was a regular behavior for this species it more than likely would have also been used as an alarm to danger as well as way to recognize the presence of a herd or to call a herd; for example, in the case of an animal being separated during a predatory attack or having simply lost the herd while feeding. This would come in handy if studies (I'm looking for these purported studies actively) suggesting sauropod vocalizations were geared more toward hissing than lowing in the way of modern cattle are correct; sauropods have always had a rather decidedly bovine-like voice in Hollywood, so it is very hard to imagine them making hissing sounds actually. Paul does mention (page 37) that the long necked sauropods' long tracheas should have been able to create and emit low frequencies and therefore, using Paul's viewpoint, the rattling of the spines would have been an auxiliary mode of communication over distances rather than a primary form of communication.
Those spines, however, would probably have produced more of a thud sound due to their shape behind the cervicals. This is probably the reason that Paul has, in his illustrations, combined the idea of the camel-like hump behind the shoulders with the free keratin sheathed spikes on the cervical vertebrae. As noted yesterday in the linked SVPOW article, those dorsal neural spines are somewhat deep paddle shaped bones as opposed to the longer horn like cervical neural spines. It is pretty interesting actually that Amargasaurus, as interesting as it is already, would have such a diverse network of spines on its vertebrae.
Paul, G. S. The Scientific American Book of Dinosaurs, New York, NY.: St. Martin's Press., 2000
Paul, G.S., The Princeton Field Guide to Dinosaurs, Princeton, NJ.: Princeton University