It is reported that chronic electrical stimulation of perforant path (PP) produces mossy fiber sprouting in dentate granule cells and as a result new reccurent excitatory circuits are formed. It has been assumed that the circuits are responsible for the well-known lowering in the threshold required to generate seizure discharges by the electrical stimulation. The assumption was tested in the present investigation. Current source density analysis of the field potential in the dentate gyrus was carried out in control (n=8) and kindled (n=10) rats under pentobarbital anesthesia during seizure discharges. The field potential was produced by stimulating the PP and recorded simultaneously from 10 sites which were 50 μm apart from each other. The field potential consisted of a monosynaptic excitatory postsynaptic field current (EPSC) in the middle molecular layer followed by a population spike current (PSC). In the kindled rats, following the EPSC and PSC, a late sink appeared in the supragranular layer (SGL). The amplitude of the SGL late-sink showed a positive correlation with the density of Timm granules (r=0.76). The latency difference between the onset of PSC and that of the SGL late-sink ranged from 0.5 to 1.5 ms. This suggested the possibility that the SGL late-sink is produced via the recurrent excitatory circuit newly formed by mossy fiber sprouting. The SGL late-sink, however, evoked no further firing in dentate granule cells. During 5 Hz PP stimulation for 10 s, the amplitude of SGL late-sink increased gradually at first, but later it decreased gradually and disappeared before seizure discharges produced. Accordingly, there was no evidence suggesting that the new recurrent excitatory circuit played any role in producing seizure discharges. During seizure discharges, the EPSC resembling the SGL late-sink appeared spontaneously only in the SGL and seemed to initiate repetitive synchronized-firing of granule cells in one of four kindled rats. The number of burst discharges and the duration of seizure discharges with the late sink, however, were not significantly different from those of seizure discharges without such a late skin. These results suggest that the newly formed recurrent circuit plays no essential role in decreasing the threshold required to generate seizure discharges or in developing them. (J Nippon Med Sch 1997 ; 64 : 154-162)