The headline is a grabber: “GOP support for Perry plummets after debate flub.”  The second takeaway from the latest NBC poll release also sounds explosive:

After his flub at a Republican debate on Wednesday, Texas Gov. Rick Perry has seen his support plummet among GOP primary voters, according to a re-contact of Republicans surveyed in this month’s NBC News/Wall Street Journal poll.

And after another charge of sexual harassment this week, ex-businessman Herman Cain’s support has stagnated or slightly declined.

“If this was not a good week for Herman Cain, it was an absolutely horrendous week for Rick Perry,” said Democratic pollster Peter D. Hart, who conducted this re-survey.

The beneficiaries: “Mitt Romney has slightly improved his position, while Newt Gingrich is the candidate who has gained most from the drops from the Perry and Cain camps,” Hart added.

Hey, all of those things might be true … but you can’t make that assumption from the NBC poll.  In fact, you can’t make any assumption from a national poll taken among 102 respondents.  Yes, it’s true, the new data comes from a survey of 102 people —  who had previously taken part in the prior NBC national poll of likely Republican primary voters.

How many people participated in the original poll?  Only 248, which itself is a highly weak sample size for a national poll.  That would be a weak sample in some states, let alone nationally.  Going back and cherry-picking less than half of a substandard polling sample in order to draw national conclusions about the direction of a presidential race — and presenting the results as a poll rather than an anecdotally-sourced story — amounts to polling malpractice.  How did this subsample get chosen for “recontact” anyway?  Did NBC go back and “recontact” respondents who were already inclined towards opposition to either Perry or Cain in the first poll?

And just who were these 102 Republican primary voters?  Only 60 of them identify as Republicans at all.  Another 22 are independents who lean Republican.  Three of them are Democrats, and another five independents who lean Democratic, with three more classifying themselves as “other” and another respondent who has no idea what his/her affiliation is.

This poll makes an impression, all right, but I’m pretty sure it’s not the impression NBC intended.