Remembering a depressive primary object

Dear Sirs,

Marianne Leuzinger-Bohleber and Rolf Pfeifer (2002,  83(1):  3-33) have done the readers

of this Journal a genuine service by providing a clear exposition of the latest models of how memory functions, articulating the advances of these ‘embodied’ models from cognitive neuroscience over earlier ones based on computer science (‘classical cognitive science’). In the process, the reader gains a sense of the historical progression of such modelling, a movement from computer science to cognitive science to cognitive neuroscience.

Two additional qualities of their presentation, beyond its clarity, stand out as particularly helpful. First, the authors anchor the theoretical information in clinical material, enabling us as readers to explore abstract (and perhaps novel) concepts in a context that is meaningful to us. Second, the authors bring a critical reading to these models and their usefulness for psychoanalytic practice. As they themselves conclude, one wants to avoid ‘the danger for psychoanalysis of an over-hasty adjustment to viewpoints which are culturally and “scientifically” accepted but foreign to our profession’ (p. 29). It is in light of this concern that I make the following comments about ‘categories’ and ‘value’ (a concept about which the authors acknowledge that there is ‘extended debate’ (p. 18, fn. 9). As the authors point out, earlier computer science-generated models (cf. Rumelhart, Lindsay & Norman; Schank) were problematic. Those models postulated memory as static storage places for large databases that were structured by external experts (programmers), whose goals were to enable efficient access, search and retrieval of data, as well as the inference of new data. Early structures that focused on internal relations among data (such as ‘isa’ for taxonomy) were replaced by meta-structures such as ‘scripts’ (Schank & Abelson), ‘frames’ (Minsky) and ‘schemes’ (Rumelhart).

Unfortunately, more structure led to less efficiency and this approach was