The cytoskeleton, a lattice polymer network supporting shape and internal communication within living cells, may compute at the molecular level. We propose a set of models for computing within cytoskeletal filamentous polymers: microtubules ("MTs"), actin, intermediate filaments and cross-bridging microtubule-associated proteins ("MAPs"). Signals and information may be represented and transmitted via propagated conformational changes of these structures' subunits which locally interact via "cellular automata-like" interactions. Conformational automata (based on dipole-coupled coherent 10″ to 10′ sec cytoskeletal subunit excitations) may recognize and adapt to neuronal membrane and synaptic events (via second messenger systems) by changing conformational automata patterns, modifying MAP-MT connections (and thus neural architecture and synaptic function) and retrograde signaling. These cytoskeletal functions may subserve dendritic processing, ANN-like paradigms such as back-error propagation and provide a primary medium for neuronal information processing and storage.