Never mind the low-sodium diet. Some things a person just has to take with a pinch of salt.

One is the notion that scrupulous technique leads always to a dependable result. Researchers by the score learned scientific method in high school and practiced it faithfully throughout laboratory careers.

Yet their misses always outnumber their discoveries. Thomas Edison always kept his failures at hand. Maybe he wanted the inspiration they provided for him to keep inventing. Maybe he just had a lot of storage space.

Of particular note in the current environment are the findings of poll administrators, those folks nosy by nature, paid by politicians and the press, and railed on often because of errant results.

Prognostication, no matter the meticulous methodology, carries with it a risk. A Democrat waved me off last week on the question of party confidence going into the fall campaign. The degree beyond confidence, he said, is over-confidence, and that won’t do.

Besides, the man insisted, the polls often get it wrong.

“Look at Jeff Harris,” he said, referring to the state representative who ran for Missouri attorney general in this month’s primary. “The polls had him ahead, and he ended up finishing third.”

Surprising, yes. But the pollsters who compiled the numbers didn’t slop their way through the work. Their task took place in a time-honored way.

The outcome, provided by citizens going to a polling place, just proved them wrong.

This is not a mystery of the sort other polls offer. For instance, congressional approval surveys show that lawmakers in Washington barely manage 20 percent when a question gets asked about their job performance.

However, more than 90 percent of those holding seats in Congress get re-elected every two years.

Sometimes, the numbers just don’t calculate.

The respected Rothenberg Political Report recently ran the numbers on congressional elections in 2006 and the polling that preceded them. One incumbent, Rep. John Hostettler, an Indiana Republican, trailed by four percentage points in September and ended up losing by 22 points.

A poll showed Rep. Nancy Johnson, a Connecticut Republican, leading by four points in mid-August, then dropping the race by 10 points in November.

Then, the report also cites numerous races where incumbents barely lead in late-summer polling but triumph by large margins in the autumnal voting.

In short, polls that are interesting to read about never come with guarantees.

As with most things, technology has an effect on the business of polling. More people than ever have caller ID, providing a no-muss, no-fuss escape for anyone not caring to share their political views with a stranger on the telephone.

Beyond that, a growing number of people have no land lines on which to field polling questions. If those people skew to a certain demographic, say young people, the results might miss a vital cross section.

And a disgust factor weighs in. People washed over by political ads on television and radio and online and in their mailboxes might draw the line at extending the madness to their telephone, the thing they use to call loved ones.

The disgruntled form an under-represented voting bloc in telephone polls.

Or at least 70 percent of people surveyed say that.

Ken Newton’s column runs

on Tuesday and Sunday.