She filled the board with newly available numbers describing the link between education and employment in South Africa. With fifteen other women and men from South Africa, she had come to the University of Michigan Institute for Social Research (ISR) to learn how to use surveys "for the scientific measurement of human activities."

For a few minutes, everyone stared at the tables of percentages on the blackboard, including Couper, an ISR sociologist born in South Africa who was teaching the session. "So you found a positive relationship between education and employment status," he said. "Why? Is it really education that affects employment, or is it something else, like having rich parents? Even after you control for education, there are gender differences. Why? Are there any clues in the data set?"

A student named Nolunkcwe Bomela said, "In some Asian and African countries, men get more food than women. Boys get more food than girls." Several male students groaned.

"In the old times," Bomela continued, speaking louder and more precisely, "there were strong preferences for boys over girls. My perception is that this is not so any more. People have moved away from that attitude. A woman today won't feed her baby girl any less than her baby boy."

Using data on feeding practices from a section of the South African Living Standards Measurement Survey, Bomela had calculated that mothers of children age 6 and under fed 50.1% of baby boys and 49.9% of baby girls three times a day, and that they fed 21.5% of baby boys and 23.3% of baby girls four times a day. "That's my story," said Bomela, who works as a demographic researcher at the Human Sciences Research Council in Pretoria. She was taking courses at ISR's Summer Institute on questionnaire design and quantitative analysis as part of a three-year training program funded by the Mellon Foundation.

"You expected no difference, you found no difference. Is it over?" Couper said. "Or do you look at these results by race, by age, by socioeconomic group, by whether the person comes from an urban versus a rural area, since rural areas tend to be more traditional, and traditional societies tend to value boys more than girls?

"You don't give up when you find what you've found," Couper said, encouraging the class to consider the possibilities.

The sample included mothers who weren't breastfeeding, so could they calculate whether mothers were more likely to breastfeed boys than girls? Or whether mothers tended to breastfeed boys longer than girls? If so, was this necessarily an indicator of preferential treatment of boys over girls?

"Boys are often weaker than girls," one woman suggested, "so maybe mothers feel they have to feed them more." Loud groans erupted again from the men. "Also," Bomela said, "boys demand more food than girls. Every time the child cries, the mother gives it food. It is widely perceived that boys are more demanding."

"This is why I enjoy being a social scientist," said Couper, smiling.

ISR Director David Featherman and U-M colleagues Oscar Barbarin and David Lam developed the training program, along with South African professor Mala Singh. "The apartheid era in South Africa introduced many distortions into surveys conducted by government agencies," Featherman says. "At a moment when rapid and profound changes demand careful scientific analysis of population trends and socioeconomic inequalities, this legacy of distortion handicaps the democratic process."

Helping to develop a new generation of South African researchers who know what they're doing and can be trusted by a suspicious public is the challenge here, as Featherman sees it.

s ISR celebrates its 50th birthday with a year-long series of lectures, symposia, publications and exhibitions, the South Africa program points to a new trend toward international collaboration.

At the start of this decade, notes Director David L. Featherman, ISR helped establish its first sister institute, in Eastern Europe at the University of Warsaw. Research links with social scientists in Germany, Japan and China and elsewhere allow US social scientists to take advantage of "pockets of expertise" that have developed around the world.

International studies also allow researchers to sort out which features of society and individual behavior are local or cultural, and which are universal. "We can see ourselves better from the point of view of colleagues from abroad," Featherman notes.

The academic traffic certainly isn't one-way. More Michigan researchers travel overseas, but more foreign researchers also arrive in Ann Arbor to learn the secrets of the survey trade. The 1997 ISR Summer Institute in Survey Research Techniques attracted students from 29 countries, including Turkey, South Korea, Trinidad and Tobago, Zambia and Nepal.

Featherman is not for resting on ISR's laurels. "If we are to continue to play a leading role in social science research, we have to import as well as export," he says. Meanwhile, developing nations all over the globe are eager to obtain the kind of basic information provided by ISR surveys, and the findings actually lead to action in the form of heated political debates and real-world policy decisions.

"It's ironic," Featherman says, "but survey results are used far less here than elsewhere. The political process in this country uses social science to strengthen political decisions that have already been made, as if the findings are weapons or ammunition. We've moved away from the belief that there can be a scientific view of human activity. In other countries, like South Africa, there's an almost desperate need for information to make fundamental decisions about housing, jobs and the educational system. But until now, there hasn't even been a census that had any value."

s the Information Age enters a new millennium, anyone in America with a telephone and address can tell that surveys and polls are a growth business. Every year, the US government spends about $2 billion on surveys, not counting the decennial Census and the annual Current Population Surveys. States and local government agencies spend millions more. But public spending is a drop in the bucket compared with the private sector. According to Marketing News, in 1996 the top 50 market research firms in the nation, big guys like Gallup and Nielsen and Roper, reported revenues of $5.05 billion. Just over 40 percent of that revenue came from overseas business, which is by far the fastest-growing segment of the field.

Political candidates all over the globe want to gauge public opinion. And a whole range of corporations, selling everything from transportation and technology to health care and financial services, need survey information to meet the needs of existing clients and develop new markets at home and in developing nations. As a result, people who know how to do state-of-the-art survey research can pretty much write their own tickets.

In a field crowded with research firms, a few things make ISR distinctive. Along with the US Census Bureau and the National Opinion Research Council (NORC) at the University of Chicago, ISR boasts a national team of full-time field staff and part-time interviewers who are ready to roll. ISR surveys have a reputation for innovative design, up-to-date sampling and survey methods and top-notch systems to safeguard the quality and confidentiality of the data. So it's hard to argue with the findings.

In addition, 90 percent of ISR's survey work is generated by U-M research scientists using public money. Less than 10 percent is commissioned by private firms or foundations. So the results can't be dismissed as having been bought and paid for.

The breadth and depth of its expertise also makes ISR distinctive. Affiliated researchers come from a wide range of disciplines-from anthropology to epidemiology, engineering and evolutionary medicine. Among them are experts in some of the most esoteric details of survey work, from questionnaire and sample design to field data collection. Some statisticians even specialize in analyzing data that isn't there, using a technique called imputation.

As a result, despite the growing competition for shrinking government funding, ISR stays pretty busy. By the end of this year, telephone interviewers working from the ISR Survey Research Center's central phone facility should have logged more than 38,000 hours of interviews. In 1998, ISR field interviewers across the country are projected to spend more than 180,000 hours contacting and interviewing subjects by phone and in person. You may well be one of them.

In addition to the large-scale surveys, suggests Fred Bookstein, distinguished research scientist in the U-M Institute of Gerontology, "ISR needs to cultivate an intellectual milieu that encourages an attitude of social criticism and the pursuit of European-style investigations into the meanings of life, studies of class origins and ideology, delicate biographical studies of decisions and regrets, and many other types of inquiries. How about some surveys of the manipulation of public opinion, for instance, or of the effects of publishing survey findings?"

lsie Bremen, 84, worked as an ISR field interviewer for 40 years. She retired last year, but still comes down to Ann Arbor from her home in nearby Southfield to talk to new researchers about what the survey trade is like from the front lines.

A legend for her skill in tracking missing people, Bremen once found someone said to be in Dogpatch, Kentucky, by a helpful neighbor in another place altogether-Turtletown, Tennessee. Fifteen years after a group of mothers and children were first interviewed, she helped locate 98 percent of them.

Called the queen of refusal conversion, Bremen has a mix of tenacity and charm that could coax conversation from a stone. Over the years she has tackled her fair share of challenges, interviewing bereaved spouses, couples who have tried and failed to have children, parents of children with cystic fibrosis and city kids in a Chicago neighborhood so dangerous she was given an armed escort. "It was silly," said Bremen. "I didn't need one. Those boys were really nice."

No matter how uncooperative the person or how difficult the situation, Bremen usually managed to complete the questionnaire, which sometimes took up to two hours. Her secret? "I like people, so they like to talk to me." Some of the people she has interviewed still send her cards and photos.

Conducting surveys in the US is harder than it used to be. For the last 20 years or so, response rates have been dropping. Today ISR aims for a response rate of 70 percent or better in random telephone surveys, 80 percent in face-to-face surveys, and higher on panel studies that track the same people over time.

Finding people at home is the easy part of the problem to fix. All you really have to do is spend more time and more money, making repeat calls and visits.

Getting people to cooperate is tougher. They say they're too busy, or they don't believe in surveys. People who live in central cities, gated communities or places with bars on the windows are the most likely to use delaying tactics and make negative comments, according to ISR research. Men are only slightly more negative than women, and people in their 30s are the most likely to say they're too busy. Lately, ISR and other survey groups have started to offer people $20 or so as an incentive to cooperate, devoting larger chunks of a project's budget to "refusal conversion."

Besides the obvious increase in the number of surveys and polls being conducted by everyone from political candidates to manufacturers of prophylactics, the American reluctance to be surveyed may be fueled by an increase in public concern about privacy and confidentiality, an increase in cynicism about social science, a decline in the public sense of civic responsibility, a growing suspicion of strangers, a widespread feeling of personal alienation, the conviction that one's attitudes and beliefs just don't matter-or all of the above.

What goes on in the first few minutes after potential respondent meets interviewer is crucial. Since experienced, confident interviewers elicit the fewest negative comments, delaying tactics or questions, selecting the right people and training them well is important. New ISR interviewers are flown to Michigan from all over the country for a week of training. They are taught to speak slowly, at the rate of about two words a second, and to give value-neutral feedback for acceptable responses between 30 and 50 percent of the time, by saying "Thank you" or "I see " instead of "good" or "okay." They go over the survey question by question, learning when and how to clarify meaning, how to probe for information without pushing too far, how to handle unwanted digressions when interviewing people who are lonely or disturbed or just have too much time on their hands.

So much effort goes into getting good response rates because what ISR and most other survey groups do these days are sample surveys, not head counts. According to James Lepkowski, a biostatistician who directs the ISR Summer Institute, a good sample needs to be both random and representative of the entire population being studied. "Once a person or household has been selected through the mechanisms of representation and chance," he says, "we don't want to lose them." In a sample of US adults, for example, a single person may represent 19,000 people. If you lose one, you lose them all.

Finding the right people for a newly sampled population presents additional challenges. In 1980, when ISR psychologist James Jackson set out to conduct the first survey of Black Americans based on a national sample, finding Blacks to interview in the rural South and any urban area in the country was simple. But locating them in Montana, Wyoming and other regions with low population densities and small Black populations was another matter. Knocking on doors would have taken too long and cost way too much money.

The solution came to Jackson one night in a dream. "You want to know where to find Black people to interview, and you want to do it fast and efficiently, for as little money as you can? It's simple. You ask white people. They know just where the Blacks are." Jackson named the technique the Wide Area Screening Procedure-WASP.

ampling began 100 years ago, but it really took off after World War II. In 1948 the men who started the Survey Research Center, ISR's forerunner, drew attention to the field when they were the only pollsters to predict a Truman victory.

"It was mostly luck. We had a very small sample," says Leslie Kish, an 87-year-old statistician and sociologist who was one of these pioneers. Kish still comes to his ISR office every day and received the top award of the American Statistical Association this past August for his contributions to the discipline, including his classic text on survey sampling, and for "using his knowledge and insight for the benefit of society." Certainly his background isn't that of the stereotypical statistician.

Kish was born in Hungary, but that was only one of the reasons he signed on to fight with a Hungarian brigade in the Spanish Civil War in 1937, when the FBI pegged him and his fellow volunteers as "premature antifascists."

"I met some Hungarians in a bar," Kish recalls, "who told me, 'You don't have to go through basic training. Go right to the front, they train you with the rifles and you start shooting. And also we have the best cooks.'"

After World War II, Kish went to work for the US Census Bureau then left Washington in 1947 with Charlie Cannell to join the newly started U-M Survey Research Center. That summer, he took the world's first course in sample design and worked on the poll correctly predicting the outcome of the 1948 US presidential election. The next summer, he began teaching the course to foreign statisticians, who now come from all over the world.

Used routinely in most countries, sample surveys are still being attacked by US politicians as "statistical guessing." Some senators are opposing the use of sampling to correct the systematic undercount of poor Americans. Kish worries that this kind of political skirmish deflects attention from the real issue-data obsolescence. "The Decennial Census was a great thing when it was invented by Jefferson and Madison," he says. "But unlike the US Constitution, it's not good enough anymore. We need, and can get, annual data, through a continuous census using a rolling sample."

Kish is convinced the rolling sample will become a reality in his lifetime, making up-to-date, detailed socioeconomic information on various segments of the current population easily available. Given the rate of social change, access to these kinds of numbers could have profound effects on politics, in America and elsewhere.