A total of 807 people were interviewed Wednesday and Thursday in The Washington Post poll on the governor's race in Virginia. Each had been interviewed between Sept. 15 and 20, in an earlier Post poll of 1,180 potential voters.

The reinterviewing of people is called a "panel-back" and is a commonly used polling technique, enabling analysts to track actual change and volatility among a set of potential voters. Since not all those originally interviewed could be reached the second time, a "ratio estimate," another commonly used polling technique, was applied to arrive at the figures used in the accompanying article.

In line with normal poll procedures, the age, sex and racial characteristics of the overall sample were adjusted slightly to have them conform to the latest U.S. Census Bureau figures for Virginia. Considerations such as a person's past voting behavior and statements about his or her likelihood of voting were taken into account to draw a "probable electorate" from those interviewed.

Theoretically, a poll this size is subject to a sampling error of about 4 percent in either direction, 95 percent of the time. Figures based on groups within the sample-- such as figures for Democrats, Republicans or independents -- are subject to a slightly higher sampling error.