Given sufficient training, any intelligent person can learn to conduct surveys, but getting people to use the results requires salesmanship of a high order. When I did research for the motion picture industry, I had my reports set in type and printed. I found that the Hollywood producers were less likely to argue with printed documents than typewritten memos. Size of sample Surveys can produce reliable results with amazingly small samples. If you want to know whether the word obsolete is understood by housewives, you don’t need an answer which will be statistically reliable within two percentage points. Twenty housewives will suffice. When, however, you are looking for trends over time, you had better use larger samples to be sure that any changes are statistically significant. You must also hold the composition of your sample and the wording of your questions rigidly constant. Pitfalls of research Some interviewers find it more comfortable to answer questionnaires themselves than to accost strangers. An enterprising London pub used to cater to them by setting aside a private room where they could drink beer while filling out questionnaires. Respondents do not always tell the truth to interviewers. I used to start my questionnaires by asking, ‘Which would you rather hear on the radio tonight – Jack Benny or a Shakespeare play?’ If the respondent said Shakespeare, I knew he was a liar and broke off the interview. When Gone With the Wind was a runaway best seller, we asked a cross- section of the adult population whether they had read it. The number of yes replies was obviously inflated; people did not want to admit that they hadn’t read it. The following week we put the question differently: ‘Do you plan to read Gone With the Wind?’ It was easy for those who hadn’t read it to answer yes, they planned to read it, while those who had already read it said so. This produced a credible result. Waiting for a train in Pennsylvania station one evening, I was accosted by an interviewer and asked questions which I had written two days before. They were impossible to answer. I went back to my office and