ABSTRACT
The work of Pollack (1990), although not dealing directly with syntactic issues, provides another important tool for tackling connectionist syntactic processing. He has developed a technique (recursive auto-associative memory, or RAAM) for representing recursive open-ended structures such as lists and trees in distributed fixed-width patterns, of the type one finds in hidden units of feedforward networks, or indeed in the fixed-width neural pathways of the brain. RAAMs can facilitate the encoding, decoding, and manipulation of complex tree-structures, which is a necessary pre-requisite for syntactic processing.
