About 1 in 4 of those polled mistakenly think that national peanut butter brands are involved in the product recalls, but fewer than half are worrying about recalled snack bars, baked goods, ice cream and dry-roasted peanuts.
The recall of 1,900 products includes mainly minor-label peanut butter and a range of other items, but not major brand names of jarred peanut butter.
"A lot of people have taken some precautions but they're not looking at the ingredients in products not related to peanut butter," said Robert Blendon, the Harvard health policy professor who directed the survey.
About 93 percent know about the outbreak and recall, and most of them understood it was caused by salmonella bacteria an unusually high level of awareness for a public health issue, Blendon noted.
The poll also indicated little faith in corporations and the government. Only 1 in 3 Americans said they have a good or great amount of confidence in food manufacturers or government inspectors to keep food safe, the survey found.
Federal health officials are tracking a salmonella outbreak that has caused at least 636 illnesses in 44 states and has been linked to 9 deaths. The outbreak has been traced to a Virginia-based company, Peanut Corporation of America, that makes some minor-label peanut butter, peanut paste and other products.
Nearly 200 food makers who used or sold Peanut Corporation products are listed in a recall of more than 1,900 different items, making this one of the nation's largest recalls.
The telephone survey, which dialed both landline and cell phone numbers, included nearly 1,300 U.S. adults. The interviews were done last week.
Of those that knew about the outbreak, 70 percent knew that peanut butter crackers were part of the recall.
There was not a question about all brands of peanut butter. But a question about major national brands indicated 25 percent mistakenly thought they were involved and had been recalled.
Only about half correctly identified some snack bars containing peanut paste as part of the recall. Just a little more than a third understood that some candy and prepackaged meals were involved, and only about a quarter identified some types of ice cream as a risk.
Fewer than 1 in 5 people have gone to the U.S. Food and Drug Administration's online list of foods involved or sought other information about recalled products.
The survey also indicated extremes of concern and apathy: About 31 percent contacted friends or relatives to make sure they know about the recall, and about 15 percent stopped eating any foods containing peanuts. But 69 percent didn't contact loved ones, and 45 percent continued to eat all peanut-containing foods.
The survey also found that 33 percent of all survey respondents were very worried or somewhat worried about getting food poisoning, which was down a bit from the 38 percent who expressed such concern in a similar poll last June.
"We don't know why the level of overall worry about food safety did not increase," Blendon said. One possible factor: "About the peanut thing, some people say they are not worried because they're taking precautions," he said.
The poll also found that 37 percent had a good or great amount of confidence in government food inspections, down from 47 percent a year ago. About 48 percent had significant faith in grocery stores to safeguard food, down from 58 percent a year ago.
Only 32 percent had significant confidence in food manufacturers. There was no similar question on last year's poll to compare that result to.
The negative reviews may be due in part to increasing success in tracking food problems, said Glen Nowak, a spokesman for the Atlanta-based U.S. Centers for Disease Control and Prevention.
Although the number of confirmed food poisonings has held about steady in recent years, more advanced testing allows investigators to better link cases and identify national outbreaks.
"The system is going to look less safe," Nowak said.
Harvard is funded by the CDC to do a series of surveys on public health topics. The Harvard poll was conducted by ICR of Media, Pa., and had a margin of error of plus or minus 3.5 percentage points.