Published online by Cambridge University Press: 01 February 2010
The question of what categorical structure is required to give semantics to O‘Hearn et al.'s type system Syntactic Control of Interference Revisited (SCIR) is considered. The previously proposed notion of bireflective model is rejected as being too restrictive to accommodate important concrete models based on game semantics and object spaces; furthermore it is argued that the existing proof-sketch of the important property of coherence for these models is incorrect. A new, more general notion of model is proposed and the coherence property proved.