If I remember correctly, a dogbones exercise involves collecting unordered control points (i.e. much like a Score-O) with the condition that all points come in pairs. That is, if you collect one control, you must then immediately collect its pair. Typically the pair is not the nearest control; if you were attacking the points in any order, you would not adopt the order given by the dog bone stipulation.
Running with the dog bone method gives a much longer orienteering session, and is particularly useful in small parks. The nomenclature emerges naturally because pairs of controls connected by a line resemble a dog bone.
Aha! well it does sound like something we could do for a summer training event - thanks!