An estimated rise in same-sex couples during the 1990s may actually be due to a change in the 2000 Census format, the Census Bureau reported on its Web site.

Live-in same-sex partners who checked off "spouse" as their relationship and "married" as their marital status on the 1990 form were reclassified in other categories based on other information on the form, including as heterosexual married couples, bureau officials said.

"In 1990, the logic then was that these are opposite-sex couples," said Martin O'Connell, chief of the bureau's fertility and family statistics branch.

In 2000, a same-sex partner identified as a spouse on a census form was counted in the unmarried partner category. The 2000 census short form did not ask specifically about marital status.

As a result, "estimates of same-sex unmarried partners are not comparable between the 1990 and 2000 Census," said the advisory posted on the bureau's Web site. "We believe 2000 estimates of this category are better estimates than those produced in 1990."

Several bureau analysts previously had said that comparisons of same-sex couple homes were valid, because the number of people who fell into the "same-sex spouse" category in 1990 was small. They did, however warn that 1990 figures were from a survey of households while figures from 2000 were based on a count of all homes.

Gay and lesbian advocates were angry at the Census' backpedaling. They contended that the explosive increases detected by the Census are real and resulted mainly from more homosexual couples becoming comfortable identifying their relationship.

"Statistically, what they did in 1990 was outrageous," said David Smith, a strategist with the advocacy group Human Rights Campaign. "The effect of that rendered gay families invisible," Smith said.

Releasing numbers detailing how many people who identified themselves as "same-sex spouses" in both counts would shed more light on the problem, Smith said. Regardless, Smith said gay and lesbian households were still undercounted in both censuses.

The bureau defended the 1990 count, saying the decision to reclassify same-sex responses was based only on statistical considerations and was not meant to relay a social message about same-sex partners.

They also said increased attention on gay and lesbian homes in the years leading up to the 2000 census led to changes in how same-sex spouses were counted.

"In 2000, because we were more cognizant that it could occur, it was then decided that those (same-sex) responses actually mean a (same-sex) couple living together," O'Connell said.

A study by the National Gay and Lesbian Task Force estimated that the census misallocated another 4,350 gay and lesbian couples in 1990 because they identified themselves as married.

Nationally, unmarried partner homes, regardless of sexual orientation, increased 72 percent from 3.2 million in 1990 to 5.5 million in 2000. Less than 5 percent, or about 145,000 of the country's unmarried partner households in 1990, comprised same-sex couples.

The Associated Press contributed to this report.