Most alcoholics cannot learn to drink moderately, according to a study not yet released by the Rand Corp. This reverses conclusions Rand announced four years ago.

The earlier study said that some alcoholics could resume drinking -- a suggestion that left the alcoholism field reeling and took an unknown number of drinkers off the wagon.

The new study, done by the same researchers for the National Institute of Alcohol Abuse and Alcoholism says that fewer than 8 percent are successful at controlled drinking (not more than four drinks a day) after four years, and that the relapse rate among this group is higher than among those who abstain totally. The earlier study said that the relapse rate for those who drank was no higher than for those who abstained.

The findings are set forth in nearly 400 pages of tables and statements. What emerges, according to Loran Archer, executive assistant to the director of the National Institute on Alcohol Abuse, is that "those who were dependent [on alcohol] cannot go back to normal drinking .... The previous finding does not stand up over this period of time."

The study, which is to be made public next week, involved a sample of nearly 800 people from eight federal alcohol treatment centers, who were among the 1,304 people included in the first study.

Overall, it found, more than half the survivors (nearly 15 percent died during the study) at the four-year mark were showing symptoms of alcohol dependence such as tremors, blackouts or loss of control, or consequences such as liver disease, drunken-driving arrests, work problems or frequent arguments.

Fewer than one-fourth had been sober for a yera or more.

Eighteen percent were found to be "drinking without [recent] problems," but of those only 8 percent reported drinking less than two ounces a day -- a percentage that drops to 6 percent when corrected, the report says, for alcoholics' tendency to drink more than they admit.

The relapse rate after four years for those who had abstained at least six months at a the time of the first Rand study, was 30 percent, compared with 41 percent of those who had been drinking without problems (but in unlimited amounts), and 53 percent of those classified as short-term abstainers, that is, those who had eschewed liquor for from one to six months, the most fragile period in the recovery process.

The original study claimed a 70 percent improvement rate will fewer than one-fourth having abstained for at least six months at the time of the first study. But that study's statistics were muddied by mixing true alcoholics, who are physically addicted, with heavy drinkers and by lumping those who had not drunk recently with long-tern abstainers when comparing outcomes for abstainers and "normal drinkers."

That study based its finding that "some alcoholics do return to normal drinking with no greater likelihood of relapse than alcoholics who choose permanent abstention" on a small sample of alcholoics. The abstainers and drinkers indeed showed identical relapse rates of 16 percent in that study -- but the percentages represent five abstainers and three drinkers.

The new study largely avoids such pitfalls, comparing larger groups and applying somewhat more rigorous definitions and methodology. It finds that more than 73 percent of those with even one symptom of true alcoholism in the initial study were showing alcoholic symptoms or consequences after four years.