Thanks tavert. I built a balance beam (like a see-saw with two identical girder constructs with ladders) and crawled a kerbal back and forwards along the ladders and it never tipped. It seems that a kerbal on a ladder might have a mass of zero??? Perhaps it would work differently if I took it off the KSP launch pad? Then I built a light rocket platform with two command seats on opposite sides and a small engine system at the centre. With a kerbal in only one seat the platform tips over very rapidly to that side on takeoff. I then added and subtracted parts in a cluster around the empty seat until it launched straight up. This way I found the mass of a kerbal when seated as approximately equal to 4x Z100 batteries plus 3x OX-STAT Solar panels. That's 0.095 tonne, so in this case the mass value given for kerbalEVA in the persistence file is correctly being modelled by the physics engine. So now if I need a crash test kerbal dummy I can use 4 batteries and 3 solar panels
