demographic responses. In this case, a computer creates a number. Although, points out Mr. Ferree, "There are drawbacks to even this technique. There is no way of knowing if the number the computer has selected is listed or even working. In addition, single women with unlisted telephone numbers, people who have moved and people with unlisted telephone numbers get systematically left out."

I was very interested to find out that pollsters need only ask 1,500 people (which is considered a large poll) to get an accurate result with a 3-4% margin of error. "This means that there is only a one in twenty chance that you will get answers differing by more than this in a sample, compared to what would happen if everyone in the population was asked the questions," Mr. Ferree told us. Twenty or thirty people can make up a poll, but the margin of error will be very high. A national survey takes about 10-15 minutes on the phone and costs about $60,000.

Even the most experienced pollsters can be wrong. For example, the pre-primary polls conducted before the New Hampshire primary were off base in understating how well Senator McCain would do against Governor Bush. According to exit-pollster for a major network, Marty Frankel, "Primary polls are the hardest to conduct because some people might say who they are going to vote for, but not actually show up to push the lever."

Polled information can have very varied results. Did you ever wonder why different television stations report different information on the same topic, like the presidential election? A television station reporting the latest polled information on the presidential race generally doesn't tell you what questions the pollsters asked respondents. These questions make a big difference to the inherent meaning of the poll and what it was intended to measure. "Even if no one is deliberately trying to bring about a particular result, what looks like similar questions can be answered very differently," added Mr. Ferree. In addition, information derived from a poll may have been achieved by a pollster manipulating a response. "Specifically," Mr. Ferry added, "how a question is asked and how it is worded makes a big difference in how it will be responded to. Different wording produces different responses." Yet, we have to wonder how the viewing/reading audience will respond to a reported poll. Will it influence their decision?

What are we to believe? We need the facts. We need to know how big and what population was surveyed; we need to know how the population was selected; what questions were asked and how the questions were worded. Also, we need to understand that we may not be getting all of the pertinent information from just polled responses. According to Mr. Ferry, "If it was not asked, it won't be reported." Therefore, a poll can be slanted in the presentation of the opinions it presents.