If you want to test the accuracy of a square using another square, you actually need a total of three to do the job correctly. Using just two squares can give a false reading as they can both be out the same amount. But by checking three squares in all possible combinations of two will reveal any problems, but will not tell you, necessarily which one is out.

To check an individual square, machinists use a combination of a grade A (inspection grade) granite surface plate and a very high precision gadget called a cylinder square. A cylinder square is basically a length of stress relieved cast iron rod that has been turned on a high precision lathe to a true diameter (i.e., not tapered) and with the ends faced off to a perfect 90 degrees to the turned surface. The thing is then precision ground on all surfaces. These things are typically square to within 0.0002" per foot, or better, depending on the manufacturer. Here is one example that will set you back less than the cost of new Volvo:

http://www.use-enco.com/CGI/INSRIT?P...&PARTPG=INLMK3

Using a set of three squares and a precision surface plate, it is also possible to create a set of three "perfect" squares by comparing, then adjusting, combinations of the surfaces via a process called scraping using the principle called the "Symmetrical Distribution of Errors."

Considered something of a "lost art," scraping metal surfaces into perfectly flat or straight configurations is a fascinating, if highly skilled, procedure that has today been replaced by CNC enabled grinders. For those that may be interested, the Bible of the practice is the book Machine Tool Reconditioning by Edward F. Connelly (1955 by Machine Tool Publications, St. Paul, USA). If you do decide to check it out, get the book through inter-library loan as you do not want to even think about purchasing a copy!).

Brian
Taxachusetts