types as set of values

Dear typers,

   interpreting a type of a language as the set of the _values_ of the
language that have that type is an idea that belongs to the type
folklore. I am trying to trace back this idea, and the oldest paper I
was able to find that mentions it is the Amadio and Cardelli paper on
subtyping recursive types, but I am pretty sure that older references
must exist. Has anyone a better (i.e. older) reference?

Thanks in advance