NavList:

Some overlap with what Peter Hakel wrote:  

I think maybe communication efficiency is the main payoff of a shared frame of reference.  If you're a land-roving beast moving in the 2D flatland of a small patch on a large sphere, then two orthogonal basis vectors are most efficient and allow a unique designation of any local point.  A third in-plane vector would be redundant and result in the same spot having multiple valid addresses.  

The shared orientation could be set by a local feature.  Many towns were laid out with an initial street parallel to a stretch of river, so the directions used for communicating were up river, down river, toward the river, and away from the river.  This was true even after theodolite survey methods were well established and there was no difficulty finding true north.  Nomadic groups could arrange tents relative to the rising sun or a dominant wind direction.  It would not matter if the orientation of one camp site was clocked a little relative to the previous.  In warfare, a temporary, site-specific coordinate frame would be established early on for reporting developments and commanding actions.  

Long-distance navigation and large-scale shared orientation frames seem like they would have co-evolved.  Cartographers and navigators may have progressively keened their concept of true north just as the need and benefit presented itself.  

If you're a small, neutrally buoyant drifter, I'm not sure how beneficial it is to establish a shared coordinate frame.  If you develop specialized propulsion features then you have a fore/aft orientation but not a left/right.  Even if your whole group agreed that toward the sunlight and away from the sunlight was an important axis, you would need a second vector to span 2-space and 3-space.  When the gravity vector starts to matter (large size and/or land-dwelling) in addition to fore/aft, then you get left and right at no additional charge.   

--Lyle Ruppert