The new study was based on answers to questionnaires filled out by 1,878 public-school students aged 14 to 18 in the inner Boston area, where Hemenway said crime rates were much higher than in the wealthier suburbs.
The overwhelming majority of respondents were Hispanic, African-American or mixed; few were Asian or white.
Among the questions were how much carbonated non-diet soft drink, measured in 12-ounce (355-millilitre) cans, the teens had drunk in the previous seven days.
They were also asked whether they drank alcohol or smoked, carried a weapon or showed violence towards peers, family members and partner.
What emerged, said Hemenway, was evidence of "dose response," in other words, the more soda was consumed, the likelier the tendency towards violence.
Among those who drank one or no cans of soft drink a week, 23 per cent carried a gun or a knife; 15 per cent perpetrated violence towards a partner; and 35 per cent had been violent towards peers.
At the other end of the scale, among those who drank 14 cans a week, 43 per cent carried a gun or a knife; 27 per cent had been violent towards a partner; and more than 58 per cent had been violent towards peers.
Overall, teens who were heavy consumers of sugary fizz were between nine and 15 percentage points likelier to show aggressive behaviour compared with low consumers, even when ethnicity and other confounding factors were taken into account.
This is a magnitude similar to the link found, in previously researched, with alcohol or tobacco.
Hemenway said the study had included a couple of questions aimed at taking a children's home background into account, including whether the teen had taken a meal with his family in the previous days.
As it was only intended as a preliminary investigation, the questionnaire did not ask what kind of sodas the teens drank, he said.
"This is one of the very first studies to examine" the question, said Hemenway.
"We don't know why (there is this strong association). There may be some causal effect but it's also certainly plausible that this is just a marker for other problems - that kids who are violent for whatever reason, they tend to smoke more, they tend to drink more alcohol and they tend to maybe drink more soft drinks. We just don't know.
"We want to look at it more carefully in following studies."
The study, published in a British journal Injury Prevention, will revive memories of the "Twinkie Defence," a US legal landmark in which a killer successfully argued that his behaviour had been swayed by eating junk food.
The defendant in this case, Dan White, had been charged with homicide. His lawyer's successful pleading led to conviction of a lesser charge, of voluntary manslaughter.
Several studies elsewhere have established a link between very high sugar consumption and lack of social bonding or irritable and anti-social behaviour.
Some diet research has also pointed the finger at the lack of micro-nutrients as a source of aggression, but this work is still in its early stages.
-AAP
