If you are looking for polling numbers to help predict Tuesday's Democratic Primary in Pennsylvania, you won't have to go far. According to pollster.com's tabulation, there have been 26 statewide surveys of likely voters released since the beginning of April.

The three most recent polls, each conducted on one or more days between April 17 and April 19, all have Senator Hillary Clinton leading Senator Barack Obama by between three and five percentage points. Rasmussen has Clinton 47%/Obama 44%; Mason-Dixon finds Clinton 48%/Obama 43%; and Zogby/Newsmax reports Clinton 46%/Obama 43%.

Clinton supporters can draw encouragement from her small but persistent advantage. The approximately ten percent undecided also have a pro-Clinton demographic tilt based on exit polls from previous primaries. They are disproportionately older and less likely to have a college degree. In every April survey, Obama has never broken the 45% barrier while Clinton has occasionally polled above 50%, suggesting that she has the greater upside.

Obama backers can be relieved that there is no apparent bounceback from his celebrated "bitter" remarks. His performance has been largely is the low-40's both before and after his comments were made.

Making the three polls' similar findings even more robust is that none of the firms uses the same polling methodology. Rasmussen and Mason-Dixon use random digit dialing to select their sample, insuring that both listed and unlisted phone lines are eligible. But they have to rely on one or more questions to determine whether the respondent is both eligible to vote in the Democratic Primary (Pennsylvania has party registration and only registered Democrats can vote on Tuesday) and is likely to do so. But Rasmussen uses automated interviewing (press "1" if you plan to vote for Clinton) while Mason-Dixon employs live telephone interviewing.

Zogby follows Mason-Dixon on interviewing procedures but uses the voter registration rolls matched with listed telephone numbers as its sampling frame. That means it is certain that it's talking to a registered Democrat but systematically eliminates anyone with an unlisted number.

Nevertheless, polling remains much more dicey in primaries than in general elections as the New Hampshire experience amply demonstrated. A survey not only needs to get the preferences right but also must predict who is going to vote. That is a greater challenge when turnout is lower. In addition, party identification anchors opinions in general elections. In a primary, switching sides at the last minute is an easier journey since it does not mean abandoning strongly held issue beliefs.