Hi, dear sirs I'm trying to gain with mercury the next prolog idiom: % code f(A) :- A = a(B,B). % query ?- A = a(7,X), f(A). A = a(7, 7), X = 7. I tried to write something like: :- module ref. :- interface. :- import_module io. %:- import_module list. :- pred main(io, io). :- mode main(di, uo) is det. :- implementation. :- type a_type ---> a(int, int). :- inst a_inst ---> a(ground, free). :- mode a_mode == a_inst >> ground. :- pred f(a_type). :- mode f(a_mode) is semidet. f(A) :- A = a(B,B). main --> { ( f(a(7,Q1)) -> Q = Q1 ; Q = -1 ) }, io.write_int(Q), <a href="http://io.nl">io.nl</a>. But mercury (btw, I'm using rotd-2010-04-01-unstable) persistently complains on this like follows: $ mmc -e ref.m ref.m:023: In clause for `f((ref.a_mode))': ref.m:023: mode error in unification of `A' and `ref.a(B, V_4)'. ref.m:023: Variable `A' has instantiatedness `bound(ref.a(ground, free))', ref.m:023: term `ref.a(B, V_4)' has instantiatedness `ref.a( ref.m:023: free, ref.m:023: free ref.m:023: )'. ref.m:028: In clause for `main(di, uo)': ref.m:028: in argument 1 of call to predicate `ref.f'/1: ref.m:028: mode error: variable `V_8' has instantiatedness `free', ref.m:028: expected instantiatedness was `bound(ref.a(ground, free))'. For more information, recompile with `-E'. Yes, I know from documentation, that "The current implementation does not handle partially instantiated data structures yet.". Is this the case? I just wonder then what's the use from those (modes, insts)-magic if all compound insts except for ground and free are hardly useful... Your clarification is greatly appreciated. Is there any probability that this feature will be implemented in near future? Btw, interesting... If I understand correctly strong typed Tubo Prolog supported this feature by means of reference domains, but in last version of Visual Prolog they have dropped this capability. Sincerely yours, Vladimir.