CLEVELAND'S SCHOOL voucher program, which provided 3,800 low-income students with state aid to attend private and parochial schools, was thrown into confusion last week when a federal judge blocked the program from entering its fourth school year. On Tuesday, the day before school opened, Federal District Court Judge Solomon Oliver Jr. agreed with plaintiffs that the voucher system "has the primary effect of advancing religion" because the schools at which the vouchers can be used are almost all religious schools. In issuing an injunction, he wrote that waiting to decide the case would only bring "greater disruption at a later time."

Faced with public outcry, the judge apparently thought better of his cavalier dismissal of the fate of so many children forced to change plans at the last minute. On Friday he stayed the injunction. So the case remains to be decided later this year.

But the anguish of kids and their families confronted with a return to the public schools underlines the conflict at the heart of the long-running voucher issue. On the one hand are concerns of constitutional principle and a long-term worry that voucher programs will sap money and initiative from the public schools they are designed to revitalize. On the other is the stark, down-to-earth plight of kids and families who want to get out of those schools now.

The specific facts in Cleveland make it hard to argue that the church-state issue in this case is inconsequential. More than 85 percent of the schools participating in the voucher program are sectarian; no measures are taken within the schools to allow some children to opt out of religious instruction. School mission statements quoted by the court include assertions that "Total religious instruction is the major focus of the educational program" and that "the life of the school is grounded in the word of God and dedicated to the purpose of showing the love of the Savior." If the voucher benefit is available only to parents willing to accept instruction under such strictures, that creates a real constitutional issue.

Another, more basic difficulty is created for taxpayers who do not wish, and ordinarily would not be compelled, to fund such instruction directly. After hearing of the injunction, several of the schools said they might close without the voucher money, heightening the impression of a connection between them and state support.

Is this degree of constitutional affront inevitable in voucher programs? The mixed legal record so far suggests otherwise. In Milwaukee, a program created in 1990 for nonsectarian schools only, then expanded to religious schools after the Cleveland one began, survived court challenge when the Wisconsin Supreme Court ruled that any religion-advancing effects were insulated by a layer of free parental choice. No taxpayer money went to religious schools except as a neutral result of a strictly educational decision. The U.S. Supreme Court declined to hear the case. When the Cleveland case is appealed, as it inevitably will be, appellate courts will have to weigh whether differing details of the programs' design are enough to distinguish their constitutionality.

The risks of an ill-designed voucher plan are not only constitutional. The larger practical question is whether they can, as advertised, help revitalize schools by introducing competition -- or whether they will instead make weak schools weaker, leaving behind only those students whose parents lack the resources to take advantage of choice plans. Not coincidentally, these are generally the lowest-achieving children as well.

Many voucher adherents say the market mechanism itself will save these schools, if not by fixing them then by supporting enough brand-new private schools to render the old ones irrelevant. Our sense is that, though it's possible this could happen in the long run, there's a risk that it would not. The problem then becomes whether a voucher plan is paired, as it plainly should be, with the commitment to put a floor under those schools left behind. Does it truly include incentives to improve, as with an intriguing Florida plan (also under challenge) that offers vouchers only to students whose public school has failed state performance tests for two years running.

Ensuring the quality of those new nonpublic schools is another conundrum: How to do it without getting the state too intertwined in institutions whose strength is supposed to be their freedom from public-school bureaucracy? It seems to us that a carefully designed, small-scale voucher program could manage to avoid these assorted pragmatic and constitutional pitfalls and save some kids while doing so.