So, i’ve taken a closer look at the pointing response distributions that i commented on previously..

I split out the responses based on baseline, stay and rotate.

baseline: shows relatively little biasing. (yeah, straight ahead still shows it somewhat, but we’ve got a much smaller N.. )

200703071030

Stay: is pretty much the same.. with a slightly greater bias for straight ahead

200703071032

Now for rotate: Damn, look at that bias..

200703071033

So these are composed of control, verbal and spatial. How do things look under load?

I’ll save myself the trouble of the cut and paste: exactly the same. it’s rotation that produces the greatest bias, regardless of load or type of load.

So, updating (offline or online) is producing this bias.