"Bill" wrote in message news
After about 25 posts, no one has come back to the OP's point that a level
can indicate that a surface is level ("both ways") even for a surface which
is slightly off-level. Different levels have different tolerances by
design. As I indicated in my post, a 100% accurate level would be a PITA.
But, I believe that there are different categories of bubble levels. I
think this is left to be addressed.

Stabila has specs for each of their levels and all the ones I looked at say
"±1/32in @ 72in." I did not find specs for Stanley levels.... Irwin
guarantees vial accuracy for the life of the product at .029 degrees but say
nothing about the accuracy of the placement of the vial in the level's beam.

One issue I haven't seen mentioned is that longer levels tend to give a
better overall indication of level as they average out the irregularities in
the surface being tested. For example, if you use a 2' level to plumb a 2x4
stud wall you run the risk of placing the level at different points in the
curve (no matter how slight) of the stud upon which the level is placed. A
6' level would tend to give a better indication of overall plumb in such
cases but a 4' probably suffices. Same kind of issue with laying bricks or
blocks... a 2' level would be subject to the variation within the few units
it's touching rather than giving an overall indication of the course...
again a 6' level would be better but a 4' probably suffices.

Put another way, there is the accuracy of the level itself (to include the
vial, the beam, and the placement of the vial in the beam) and then there is
the accuracy of the application of the level. I dare say that the level in
my Starret combination square is highly accurate but it would be pretty much
useless for use in plumbing and leveling the framing of a house.