Was It a Slab, a Slice,
or a Sliver?
High-tech innovations take food survey to new level
A meal as appetizing as
this one may prove memorable
indeed, especially if your
memory is jogged by the
new food-survey techniques
described in the cover story,
entitled, "Was It A Slab,
a Slice, or a Sliver?"
|According to the last USDA nationwide
food consumption survey, conducted from 1994 to 1996, Americans of all ages
averaged about 2,000 calories a day. So why are we overweight as a population?
Some Americans obviously eat more than they report, say Linda E. Cleveland and
Linda A. Ingwersen, of ARS' Food Surveys
Research Group, which develops and oversees the periodic survey. So often, we
tend to forget that soda, bag of chips, or candy bar we snatched yesterday when
our tummies rumbled. And we may think we ate smaller portions than actually
passed our lips.
So Cleveland, a nutritionist, Ingwersen, a home economist, and other survey
group members are working hard to catch those forgotten and underestimated
calories in the next nationwide surveyexpected to begin in 2002 (see
| It's Not So Much What You Ask, As
Staffers have improved the way the interviewers will probe for all the foods
and beverages a respondent ate during the previous 24 hours so that the
questions don't seem repetitive. "It sounds more conversational,"
says Cleveland, "like you're chatting about the meal, rather than
questioning their memory."
And Cleveland and Ingwersen believe they have improved the accuracy with which
people estimate the size of the portions they consume. The two scientists have
developed an easy-to-use Food Model Booklet that incorporates high-tech
graphics and research on how people perceive quantities and on what is and
In previous surveys, interviewers were armed with measuring cups, spoons, and
rulers when they visited households. But during the next survey, respondents
will be able to turn to life-size, two-dimensional pictureseach marked by
a numbered tabas well as the cups, spoons, and ruler.
Nutritionist Linda Cleveland
(left) and home economist
Linda Ingwersen designed
the Food Model Booklet,
which contains two-dimensional
measuring guides (foreground)
to help food-survey respondents
better estimate portion sizes.
"We're trying to provide a variety of ways for people to estimate amounts
in order to make it as easy for them and as accurate as possible,"
Respondents may find it easier to recall the size of that slice of pizza by
turning to one of the wedges pictured under tab 7 and adjusting it to just the
right width than to estimate its length and width with a ruler (see above
Under tab 6, the 5-inch-by-5-inch grid for estimating that serving of lasagna,
meat loaf, brownies, or corn bread may shake loose old memories of math class,
Ingwersen says, noting that focus groups have helped them fine-tune the
For instance, on the opposite page are blocks for estimating the thickness of
lasagna or meat loaf servings. "The focus groups wanted to know the actual
measurements, so we added them," Ingwersen says. One focus group of women
wanted a smaller wedge for estimating pie or cake servings. "And they
wanted it on a dessert plate," she says. The researchers obliged.
Survey participants position the
movable arrow on this model to
resemble the size of the pizza,
cake, or other wedge-shaped
food they ate.
|To put the servings in perspective,
Ingwersen and Cleveland had the grid, circles, and several amorphous mounds
printed on transparent pages that overlie a full-size dinner plate straddled by
a full-size knife. "The transparencies give the quantity a
three-dimensional appearance," Ingwersen says.
And the different-sized moundsfor estimating foods ranging from a dollop
of whipped cream to a heap of spaghettihave depth. A graphic designer
used a computer program to draw the mounds and then rotate them to a 55-degree
anglethe perspective one would have while sitting at a table, says
Developing the collection of glasses, cups, mugs, and bowls under tabs 2 and
3each marked at several different volume amountstook a lot of
thought, as well as some computer tricks.
These two-dimensional drawings represent
foods that mound on a plate, such as
casseroles, vegetables, or rice. The
plate and knife provide a size reference.
"We filled a table with glasses and bowls. We surveyed stores and looked
at what people had in their cabinets to determine which shapes and sizes to
include in the booklet," says Cleveland. And, adds Ingwersen, "we
included a wine glass. That says, 'Yes, we do want you to report any alcohol
you drank,'" because it contributes calories.
All their planning should produce more accurate data, according to tests
conducted by the ARS researchers. The tests involved 264 men and women from age
20 to over 60, who estimated the portion sizes of all types of foods and
beverages using the booklet or the cups, spoons, and ruler.
A new, computerized interview process,
developed by the Food Surveys Research
Group, prompts nutritionist and group
leader Alanna Moshfegh to ask survey
respondents about "forgotten foods."
The mounds help the respondents
estimate portion size.
|"People estimated serving sizes
reasonably well with both types of guides," says Cleveland. "But they
did a little better using the booklet, especially the mounds." On average,
estimates improved by about one-third with the mounds compared to the measuring
Asking Without Badgering
The booklet is only part of the effort to improve the survey data, says Alanna
J. Moshfegh, who oversees the Food Surveys Research Group. During the last 3
years, the group has expanded and improved the method of questioning to help
respondents remember the foods they ate. They've also automated the whole
interview, computerizing the questions, prompts, and details about the food and
how it was prepared.
"Since the first nationwide collection of individual dietary intakes in
1965," Moshfegh explains, "the focus of research in USDA's nutrition
monitoring program has been the question, What is the most effective way to
collect a complete 24-hour dietary recall?"
This grid is used to estimate the
length and width of square and
rectangular foods like meat loaf,
brownies, or lasagna.
| In the 199496 survey,
interviewers used a triple-pass method, she explains, because her group's
research had shown that asking people about their intakes in different ways
helped respondents recall more of the foods they ate the day before. Since
then, Cleveland, Ingwersen, and their colleagues have been testing and refining
a better instrumentthe new USDA Multiple-Pass Method.
In addition to asking the respondent to remember all they ate, interviewers
specifically ask about "forgotten foods," such as nonalcoholic and
alcoholic beverages, sweets, snacks, or breads.
"Our research showed that beverages accounted for half of forgotten
foods," says Moshfegh. "Sweets accounted for one-third."
Survey participants use drawings of
different-sized glasses to estimate
amounts of beverages they drank.
|The method has a number
of built-in cues to help jog the memory. One step, for example, asks
respondents what time they ate the food and what they would call the eating
occasionlunch, snack, dinner, etc. The questions don't seem repetitive,
says Moshfegh. "The respondents remain engaged because they are still
adding foods throughout the interview."
Based on the results of pilot tests, "we believe this method does a better
job of collecting more complete food intakes," she contends. In the first
test, 383 women recalled eating an average 16 foods instead of the 14 reported
by their counterparts in the last survey. And they reported 300 more calories,
on average. In a larger study with nearly 800 men, women, and children, the
trend of more food and more calories continued.
No Easy Task
The fact that the questions and prompts are computerized is one reason for its
success. "It's easier to administer and more consistent," Moshfegh
says. "Participants said they liked the interview. And the interviewers
who previously collected data with paper and pencil liked the automated version
|| That's because the automated program
prompts the interviewers and leads them through the details they must ask about
each food. For instance, the program would prompt the interviewer to ask,
"Was that candy bar regular, king size, or fun size? Was that apple small,
medium, or large?" explains Moshfegh.
Prepared foods require more details. Take carrots, for instance. Were they
cooked, creamed, fried, pickled, or raw? the program would prompt. If the
answer is "raw," you get to move on to the next food. If it's
"cooked," however, the interviewer would ask if they started as
fresh, frozen, or canned. Were they cooked with fat or oil? And so on.
If you had beef stew, the interviewer might ask if it was home prepared,
canned, frozen, a restaurant entree, or something else. If it was a home
recipe, what were the ingredients? If it was a frozen brand name, what brand?
"There are 2,400 questions about foods and 21,000 possible answers,"
says Nancy R. Raper, who oversaw the automation. And the job is never ending:
"We have to keep up with the foods on the market to assess whether the
questions we ask are relevant," she notes.
Programming all these questions took about 2 years, Raper says, with input from
several scientists in the survey group. They used survey software developed in
The Netherlands and used by the U.S. Bureau of the Census and other government
agencies. So far, she says, the program developed for the food survey is the
largest and most complex application of this software.
Finding the Trends
As the data is collected, another programSurvey Netwill code each
reported food by number. "We have over 7,000 foods in the database, with
descriptions of their package sizes and weights, types of preparation, and
nutrient profiles," says Ingwersen, who ensures that the database is
When all the survey data is translated into numbers, it can be analyzed for
intake levels of either nutrients or foods, Ingwersen says. "And it can be
sorted by any variable you want: age, gender, socioeconomic group, geographic
region, food or nutrient, Food Guide Pyramid servings, foods eaten on weekday
versus weekend, or outside the home versus at home."
That versatility is what makes the survey data so valuable to researchers and
educators, says Moshfegh. And it's critical to government agencies in planning
food assistance programs and nutrition education programs.By
Agricultural Research Service Information Staff.
This research is part of Human Nutrition, an ARS National Program (#107)
described on the World Wide Web at http://www.nps.ars.usda.gov.
Alanna J. Moshfegh,
Linda E. Cleveland,
Linda A. Ingwersen, and
Nancy R. Raper are in the USDA-ARS
Research Group, Beltsville Human Nutrition Research Center, 10300 Baltimore
Ave., Bldg. 005, Beltsville, MD 20705-2350; phone (301) 504-0170, fax (301)
"Was It a Slab, a Slice, or a Sliver?" was
published in the March 2001 issue
of Agricultural Research magazine.