>So... 35 variables, 31 bytes used for their definitions? Variable names with negative length, then? ;) Or did you mean only the space used by variables' content - but then, at least a few of them were numeric, using at least 4 bytes (though I think VFP's numbers are in IEEE format, which is 8 bytes). Again, either some of them were using negative bytes, or had negative lengths... or this is a tiny little bug.

I mean space used by content and when that content is a reference type. I don't know the details of how foxpro counts it but looks like it doesn't count value types (maybe doesn't count until a prealloacted memory space is filled?). When it's a reference type it allocates more than needed bytes (so using some space for extra info, length, memhandle...?).

Cetin