Penn State Pulse's October 2002 survey on drug use seemed to indicate that there's a large discrepancy between respondents' own use of marijuana and how much they think other students are smoking.

This may actually characterize the situation, but we're not sure if the survey was crafted to correctly gauge such a gap. As any student of statistical surveys knows, what kind of questions you ask and how you ask them affect the results, sometimes greatly. According to the Pulse survey report, the available responses for both those interviewed and other students were "frequently," "infrequently," "never" or "don't know."

Sixty-seven percent of those surveyed said they never use pot, and yet 96 percent think that others use it -- frequently or infrequently. Consider the dilemma of one such non-pot-smoking student: He or she is aware that people are using marijuana, but is limited by the survey in being able to characterize that level of use.

Maybe the interviews should have asked people a question such as "What percentage of students do you think use pot frequently?" Then ask the same for "infrequent" use. Of course, the estimates would be vague, but the respondents would have more flexibility to express their perception of drug use.

The percentages for others' use in this survey was probably blown out of proportion because respondents didn't want to answer that others "never" use the drug, and thus were left with two poor choices.

Penn State uses the data collected from these Pulse surveys in a host of ways, most noticeably in the "Party Smart" alcohol-use advertising campaign.

It should be the responsibility of those conducting the studies to create questions that will be most effective and objective, if the statistics produced are to be believable and useful.