The original PDP (parallel-distributed processing) volumes, though somewhat dated, remain remarkably relevant: RumelhartMcClelland86, McClellandRumelhart86.
An excellent collection of the important early papers in neural networks can be found in AndersonRosenfeld88.
For other views on the basic premises of cognitive neuroscience and levels of analysis, we suggest: Marr82, chapter 1; SejnowskiChurchland89; Shallice88, chapter 2; PosnerEtAl87; Farah94bbs; Kosslyn94.
For a developmentally-focused treatment of computational neural network modeling, see: ElmanEtAl96 and PlunkettElman97.
For other treatments of computational modeling using artificial neural networks, see: HertzKroghPalmer91, Ballard97, Anderson95, McLeodPlunkettRolls98, and Bishop95.
For an encyclopedic collection of computational neural network models and more general brain-level theories, see Arbib95.