Coronation Street is embroiled in a ratings battle

They reckoned that 14.5 million people watched the soap - but Barb's figures put the audience at 11.2 million.

Carat said that Barb's figures could cost ITV �39m in lost revenue by underestimating audiences, meaning advertisers would be less willing to spend money.

But Barb said the two organisations used different methods to arrive at their figures - and that the two could not be compared.

Barb recently changed the households it uses to estimate the national TV audience.

Panel members record their viewing by pressing a button on a handset or "peoplemeter" during the show.

Average

Carat, on the other hand, conducted a telephone poll one or two days later asking people whether they had seen that particular episode.

That means the Carat figure includes everybody who saw a part of the show - no matter how much, Barb said.

Carat's figure could include people who watched a repeat or video

The Barb figure counts the average number of people watching at any one time, which could be considerably less than Carat's total of everybody who tuned in at all.

And Carat's figure could have included those who watched a repeat on ITV2 or who taped the show.

Barb says if it used Carat's criteria, there would only be a difference of 800,000 people between the two surveys - not the 3.3 million Carat said.

"Taking all of these factors into account, the results are surprisingly close," Barb's research director Tony Wearn said.

"Surprisingly, because a telephone snap-shot poll of this kind could not be expected to be compatible with Barb audience estimates."

Accurate

He added that Barb's "peoplemeter" method was "inherently a more robust measurement than a one-off question asked by telephone".

Carat had said "peoplemeter" data would be better - if it worked properly.

Carat intends to repeat the research in March when the new Barb viewing panel has been made up to full strength.

Barb, which is owned by all the UK terrestrial broadcasters and BSkyB, says its new panel, which will eventually number 5,300 homes, will be 3% more accurate than the old one.

Delay

But the transition, which started over Christmas, has had teething problems and attracted criticism.

The first figures from the new panel were delayed for two weeks, leaving broadcasters and advertisers unable to tell how popular shows were.

And when the figures did arrive, they showed large differences with previous ratings.

A spokesman for the BBC said the corporation's internal research had come up with figures that were "remarkably" similar to Barb's.