In this paper I discuss the ways in which experimental and objective research from cognitive science and developmental psychology can help analysts evaluate the theoretical models of mental objects which we use; I indicate the ways in which such evidence tends to support models of internal objects as mental representations or developmental capacities rather than as wish-fulfilling expressions of instinctual drives. This land of empirical evidence is not just of academic interest but also has direct clinical relevance, particularly with borderline patients; such patients' sense of identity is totally dependent on the analyst's understanding of their internal world and for this to be misunderstood by the analyst can be catastrophic. An accurate theoretical model of mental objects can therefore help analysts to contain their patients more effectively.