Juvenile delinquents who go on to commit murder as adolescents and young adults appear to share five common charactertistics that may make it possible to identify other disturbed youngsters before they kill.

In a presentation at the annual meeting of the American Psychiatric Association, Dr. Dorothy Otnow Lewis, a New York University School of Medicine psychiatrist, reported new findings suggesting it is not just a childhood filled with physical abuse and violence that seems to lead to the development of a murderer but rather a combination of these and other factors, including brain and mental problems.

Lewis draws her conclusions from a long-term study of more than 100 juvenile delinquents begun more than six years ago. At that time, she and Bellevue Hospital psychologist Alexander Simos teamed with five medical and psychology students to collect psychiatric, neurological and family background information from jailed delinquent boys, then aged 10 to 18.

During the study peroid, nine of the boys committed murder. Between the time Lewis analyzed the findings and made her presentation at the APA meeting, another two boys killed people. "As of today," she told the symposium, "it would be a study of 11 youngsters who went on to commit murder." Results of the study will be published this fall in the American Journal of Psychiatry. Those who kill differ from the other delinquents, Lewis said, in the following ways:

* All of the murderers showed psychotic symptoms six years ago. By comparison, only half of the delinquents in the control group were diagnosed as being psychotic.

* Almost 90 percent of the children who later killed showed major brain impairment, compared with only 27 percent of the delinquents in the control group. One third of the murderers had experienced grand mal epileptic seizures and showed abnormal brain wave tests. "Six had received severe head injuries in childhood, including falls from roofs and car accidents resulting in loss of consciousness," Lewis said.

* Abuse and violence "were rampant" in the homes of children who later murdered. Seven had been "severely abused by one or both parents," compared with 14 of the 24 delinquents in the control group. In the households of the children who later murdered, belt buckles, cords, broom handles, sticks and shoes "were used to inflict punishment." One boy's father had tried to kill him and his brother several times. Six of the boys witnessed extreme violence in their own homes.

*At least one close relative -- either a parent or a sibling -- of each of the boys who committed murder had been diagnosed as psychotic or had been hospitalized for a psychiatric illness, compared with only 58 percent of delinquents in the control group. One young murderer's psychotic 14-year-old mother had received electroshock treatments while pregnant with him.

* All nine murderers "had been extraordinarily violent long before they committed murder," yet only 62 percent of the control delinquents displayed extreme violence. At age 2, one boy had choked a bird. By age 4, he had thrown a dog out a window, and by middle childhood, had broken a sibling's arm. At age 16, he assaulted and raped a girl. At 18, he raped and then killed a woman by stabbing her 13 times. Another boy in the study had burned his bed at age 4. One boy was too violent to attend grade school. Another, at age 10, had "threatened a teacher with a razor."

Overall, 75 percent of the children who later murdered -- compared with only 9 percent of the control group -- showed those five traits, Lewis reported.

"It is somewhat surprising to note that neither early violence alone, nor a history of abuse strongly distinguish the groups from each other," Lewis told the symposium. "Rather, the combination of all five attributes most strongly distinguished the murderers from the ordinary delinquents."

* It seems, she said, "that severe central nervous system dysfunction, coupled with a vulnerability to psychotic thinking, created a tendency for the nine homicidal subjects to act quickly and brutally when they felt threatened."

The murders were largely impulsive, mindless and unpredictable acts. In at least five of the murders, "the victim was unknown to the assailant until just prior to the murder."

Although it is still impossible to predict with certainty whether children with these five attributes will murder, Lewis said that "it seems safe to conclude" that such a child "will commit further acts of violence."

"These are very early, very preliminary findings," cautioned Dr. Lenore Terr, a psychiatrist at the University of California at San Francisco, who is noted for studying the busload of children kidnaped in Chowchilla, Calif. "While Dr. Lewis has put together what may be a picture of the child who goes on to murder, it is not the picture yet. But it is an interesting picture."

"Murder is the most serious of crimes," Lewis told the symposium. "It is therefore the one we most need to learn to prevent." The discovery of these somewhat predictive characteristics may allow the medical community to "develop programs to recognize and treat" them before they murder.

The study of children who later murder was one of several hundred papers presented here during the 138th annual meeting of the American Psychiatric Association.

Another group of studies tried to understand the "voices" schizophrenic patients complain they hear -- what psychiatrists call "auditory hallucinations." New evidence from two studies at Harvard University's Medical School suggests that schizophrenic patients are actually hearing their own voices at very low levels and are attributing these voices to "others."

Dr. Peter Bick, a National Institute of Mental Health fellow working at Harvard University Medical School in Boston, studied 18 schizophrenics and found that their hallucinations were caused by "subvocalizations."

Subvocalization is commonly used by normal people, especially while reading. "Subvocalizing during silent reading decreases speed but improves comprehension," Bick says.

Schizophrenics, however, seem to misinterpret this subvocal voice as an outside person telling them what to do, or "commenting on their behavior," Bick says. People in the study had complained of hearing voices for two to 33 years.

In Bick's study, participants were asked to do three exercises and rate what effect they had on the "voices." The exercises included closing their eyes very tight; opening their mouths very wide, and making fists and squeezing very tightly.

Only the mouth-opening maneuver "abolished the hallucinations in most subjects," Bick reports. "We therefore infer that the patient subvocalizes listens to the covert speech and attributes it to another."

To further test this theory, Bick conducted another experiment in normal subjects -- 21 college students, who agreed to be hypnotized. While in a trance, these students were given a hypnotic suggestion to hear voices.

"These normal subjects typically heard a mixture of men's and women's voices in a low-volume murmuring which was difficult to understand," Bick reports. "They expressed amusement or bewilderment." Like the schizophrenic subjects, these normal college students were producing the voices themselves and, like the schizophrenics, they could stop the voices by opening their mouths very wide.

What was surprising was that neither the normal group nor the schizophrenics expressed any interest in using the open-mouth exercise to control the voices.

Bick thinks that the "voices" may be a way the schizophrenic copes either with psychological stresses or with a disordered thinking process. Removing the voices without correcting either the stress or the thinking process, he says, "would be inappropriate."