Why does soy milk curdle when salt is added? Does it have something to do with starch?
As noted in "History of Tofu and Tofu Products", alkaline earth metals (such as calcium and magnesium) in solution cause the proteins in soy milk to curdle. As I discovered by searching for "Magnesium in table salt" and "Calcium in table salt", a cup of salt contains about 3 mg of magnesium and 70 mg of calclum. While those concentrations aren't high enough to turn soy milk into full-on tofu, they should be enough to cause some curdling.
All of this hinges, however, on the soy milk being heated. There are a number of soy milk products available in the US (e.g. Silk) that contain added calcium for health purposes--primarily in order to compete with the calcium content of dairy milk--that, themselves, do not curdle because of the low temperatures of storage associated with consumer soy milk.
While some soy starch is contained in the curds of soy milk, the curdling reaction for any solution primarily involves the kinking of the proteins in that solution, trapping within them fats and starches, in order to form a stable matrix of such.
In east Asia, having soy pudding is a rather popular past-time. Some people call it soybean curd. It has the texture of freshly settled yogurt, or that of soft-grade tofu, which breaks up easily.
Soy pudding is deliberately curdled with a coagulant.
Intensifying the same coagulation process used on soy pudding, by repetitive compressive measures would produce the firm tofu we are familiar with.
Speculation: That must be how the Chinese discovered paper making - trying to make tofu out of plant celluloid - creativity in the midst of hunger desperation.
I believe that coagulation is due to action on the protein structure of soy milk. I also believe that similar protein coagulation process takes place in the curdling of cheese, but that soy and cow's milk are susceptible to different coagulants. Coagulation action of coagulants on soy milk is induced by below-boiling-temperature heat.
Which means that soy milk should not coagulate/curdle when salt is added to it at cold/cool temperatures.
Cow's milk coagulate naturally with the application of heat. Unfermented soy has a rich natural content of anti-coagulants, which needs to be deactivated, or compensated.
I found a study that a soybean engineered with lower phytate levels, required lower amounts of coagulant to curdle its soy milk. Whereas, phytates are known blood anti-coagulants. Therefore, am I allowed to surmise that soy naturally occurring radicals that cause anti-coagulation in-vivo humans also prevents coagulation of soy itself? Therefore, I should reach a reasonable line of reasoning that soy processing coagulants may not be totally coagulant but counters the effects of anti-coagulants, where thermal application is itself a significant contributor to soy coagulation.
Since non-fermenting soy processing does not actually remove anti-coagulants, but that "coagulants" merely compensate their action on soy protein, most of these anti-coagulants pass into digestion and blood stream, causing mal-absorption of nutrients as well as lowering ability to clot blood. BTW, repetitive anti-coagulant baiting is an effective way to poison rats and mammalian rodents - "non-toxic" rat poison. Which means having too much soy is anti-nutritional towards the absorption of minerals like zinc and calcium. Read toxicity of soy consumption, Soy is not a health food, Could Eating Too Much Soy Be Bad for You?.
I am confident that my line of reasoning on the coagulation process of soy milk is reasonable but incomplete. I hope someone with proper agricultural bio-chem expertise would reinforce or correct my opinion.