About 4,000 gallons of water that might have been slightly contaminated with radioactive strontium 90 were dumped into the Susquehanna River from the Three Mile Island nuclear power plant last week, the Nuclear Regulatory Commission said yesterday.

The water came from a tank used to store leakage from the miles of pipes in Unit One, which was not involved in the March accident at Unit Two. Both units have been closed since the accident that overheated Unit Two's core and sent radiation into the air over Middletown, Pa.

Water has routinely been dumpled from the Unit One tank several times a week without being checked for strotium showed no significant amounts.

But three weeks ago a sump tank was filled with highly radioactive water from Unit Two, emptied into a holding tank, and then filled with water from Unit One, according to Robert Arnold of Metropolitan Edison, the utility that owns Three Mile Island. The Unit One water, probably mixed with some traces of Unit Two contaminated water, was then sent to the leakage tank to be discharged, Arnold said.

"It was a mistake," said John Collins, deputy director of the NRC team on the site. "It was planned but they realized later they should not have done it."

Arnold said the shift was made during efforts to juggle storage space so as to accomodate all the extra water contaminated during the March accident.

The apparent contamination was discovered when the NRC changed its rules last week to require that a test for strontium 90 and other emitters of beta radiation, called a beta scan, be performed before the water is dumped. Through a mixup, Collins said, a discharge began before the test was made.

The discharge was halted about halfway through, when about 4,000 gallons had been dumped. A test found "a slight over-concentration of beta emitters," Collins said, and the rest of the dump was halted pending further tests.

Other substances emit beta radiation and are less harmful than strontium 90, which settles in human bones when ingested, Collins explained. Tests due in two or three days will show whether the high beta count is from strontium 90 or some other source.

Whatever the source, Collins said, the amount of radiation released was insignificant and had no effect on public health and safety. Although the Susquehanna River is the source of drinking water for about 100,000 persons downstream, the low radiation level and the small amount of water released - a cube of water eight feet on a side - make the total impact on the river very negligible, Collins added.