Ependymin, a glycoprotein of the brain ECF, has been implicated in the neurochemistry of memory and neuronal regeneration. Three behavioral experiments (swimming with a float, avoidance conditioning, and classical conditioning) in the goldfish and one in the mouse (T-maze learning) indicate that ependymin has a role in the synaptic changes that take place in the consolidation step of memory formation and the activity-dependent phase of sharpening of goldfish retinotectal connections during neuronal regeneration. The ECF concentration of the protein was found to decrease after the goldfish learned to associate a light stimulus (CS) with the subsequent arrival of a shock (US): paired CS-US gave changes whereas an unpaired presentation of CS-US gave no changes relative to the unstimulated controls. Ependymin is present in ECF as a mixture of three disulfide-linked dimers of two acidic (alpha and beta) polypeptide chains (37 kDa and 31 kDa). Upon removal of its N-linked glycan fragment by N-glycosidase F, the beta chain yields gamma-ependymin (26 kDa). Determinations of the amino acid sequence of gamma-ependymin indicate that it is a unique protein with no long sequence homologies to any known polypeptide. There are, however, small segments (5-7 amino acids long) with homologies to fibronectin, laminin, and tubulin. Ependymin has the capacity to polymerize into FIP (after activation by phosphorylation) in response to events that deplete ECF calcium. FIP is insoluble in 2% SDS in 6 M urea, 10 mM Ca2+Ac2, 100% acetic acid, chloroform/methanol (2/1), saturated KCNS, and even 100% trifluoroacetic acid. FIP was found to be present in goldfish brain and to be formed as a labeled product in vivo. Ependymin's FIP-forming property was used to propose a molecular hypothesis for generating synaptic changes in response to local extracellular depletions of calcium at sites of "associating inputs." The model assumes that, following NMDA receptor stimulation, the translocated PKC that is generated activates extracellular ependymin by converting it to its phosphorylated form using presynaptically released ATP. The hypothesis was tested in studies of LTP of rat hippocampal slices at CA1. After LTP, new sites that stained with antisera to ependymin, visible at 100x, were obtained in its potentiated radiatum in the CA1 region but not in the unpotentiated CA3. Electron microscopic studies showed that the horseradish peroxidase reaction product obtained was localized at synaptic clefts and postsynaptic regions. The results suggest that FIP may be formed at extracellular and postsynaptic loci where multiple associating inputs interact at CA1.