1. Microelectrode recording and fluorescence measurement with voltage-sensitive dyes were employed in horizontal hippocampal slices from rat to investigate responses in the dentate gyrus to molecular layer and hilar stimulation. 2. Both field potential and dye fluorescence measurement revealed that electrical stimulation of the molecular layer produced strong excitation throughout large regions of the dentate gyrus at considerable distances from the site of stimulation. 3. Treatment of slices with the excitatory amino acid receptor antagonists 6,7-dinitroquinoxaline-2,3-dione (DNQX) and (+/-)-2-amino-5-phosphonovaleric acid (APV) unmasked dye fluorescence signals in the outer and middle molecular layers corresponding to action potentials in axons, presumably belonging to the perforant path. The spread of these axonal signals away from the site of stimulation was far less extensive than the spread of control signals through the same regions before blockade of excitatory synapses. Large control responses could be seen in regions distant from the stimulation site where the axonal signals were not detectable. A lack of correlation between control signals and axonal signals revealed by DNQX and APV supports the hypothesis that responses in distal regions of the molecular layer were not dependent on perforant path axons. 4. The perforant path was cut by producing a lesion in the outer two-thirds of the molecular layer. Both dye fluorescence and microelectrode recording showed that stimulation on one side of the lesion could produce signals on the same side as well as across the lesion. The lesion did not block the spread of excitation through the molecular layer. Across the lesion from the site of stimulation, negative-going field potentials were observed to peak in the inner molecular layer, which is the major field of projection of hilar mossy cells. 5. Electrical stimulation in the hilus adjacent to the granule cell layer evoked dye fluorescence responses in the molecular layer. Stimulation at this site evoked negative-going field potentials that peaked in the inner molecular layer. These signals were sensitive to excitatory amino acid receptor antagonists but not to gamma-aminobutyric acid-A (GABAA) receptor antagonists. 6. Activation of excitatory amino acid receptors in the hilus by focal application of (+/-)-alpha-amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA) and N-methyl-D-aspartic acid (NMDA) elicited negative-going field potentials in the granule cell layer and depolarization of granule cells. Field potentials were blocked by tetrodotoxin (TTX), indicating that they were not caused by direct activation of receptors on granule cells, but rather by synapses from hilar neurons on granule cells. 7. These results taken together with previous studies of hilar mossy cells suggest a fundamental circuit consisting of granule cells exciting hilar mossy cells, which then excite more granule cells. This circuit provides positive feedback and can be considered a form of "recurrent excitation" unique to the dentate gyrus. The robustness of this circuit in hippocampal slices under control conditions suggest that mossy cell excitation of granule cells could play an important role in the normal activity of the hippocampus, and, when inhibition is compromised, this circuit could contribute to the generation and spread of seizures.