Books on memory have an intrinsic appeal: We keep hoping there'll be a way of improving ours; that age will not take its toll; that we can erase the unpleasant, and retain the magic moments. In general, such notions presuppose a storehouse in the brain, a reservoir we tap for better or worse. Rosenfield, a physician Ph.D. who teaches at City University of N.Y., denies this popular view. Memory is a creative reconstruction, he argues, emotionally and contextually laden. There are no ""fixed"" cores or discs, to use the computer analogy. Rather, there are neural groups aggregated into large ""maps"" that interact in the act of recollecting, reworking, recategorizing. The original sensory features we abstract from the environment do have their byways and terminals in various locales in the brain, but the putting together of those features that become meaningful to the individual evolve in a fashion comparable to Darwinian selection. This theory has been developed in detail by Gerald Edelman, whose earlier work on selection in the immune system (for which he won a Nobel) laid the groundwork for the brain theory. Edelman's work also applies to morphogenesis: Selectivity operates in embryonic development so that while all brains have similar topographical landmarks, each brain, even of identical twins, is unique. In defending the new concepts, Rosenfield reviews some classic cases of aphasia with fresh insights, and examines as well artificial intelligence models, Freud's thinking, the foresight of Frederic Bartlett (writing in the 30's) and seminal studies of language and language acquisition. More than just about memory then, this is a provocative introduction to a view of the brain as constructive, imaginative creator and recreator.