Polls can be biased in many ways. You can bias a poll by the questions you ask, both in type and in wording. This poll is an example of selection bias. If the purpose is to show misconceptions about the Iraq war among the population, then why did they only select misconceptions from one side of the political spectrum?

Without seeing the exact wording of the questions, it's impossible to know if the questions themselves were leading or not. As an example, a poll might ask something like "Which would you prefer, Howard Dean for president or a sharp stick in your eye?" only to announce that "55% of voters state a preference of Dean for president." Without seeing the questions, the results should always be in doubt regardless of who does the poll.

You can bias a poll by how you select your sample population, both in size and makeup. A poll of 300 people from the "blue states" will likely get different results than a poll of 300 people from "red states." Likewise, the larger the sample size, the more accurate the poll results - so long as the sample accurately reflects the population as a whole.

Polls are at best an inexact art. As but one example, how many of the preelection polls came close to yesterday's recall election results in California?