Teacher assessment will not provide a reliable alternative to Sats, a respected education academic has warned.
Peter Tymms, Durham University's head of education, said teachers' inherent bias meant their assessments could not be used as an objective measure of pupil attainment.
Various models of moderated teacher assessment are often promoted as alternatives by critics including teaching unions, who want to see an end to Sats.
Professor Tymms said teachers could also succumb to pupil or parent pressure to boost grades when assessments were "high stakes". He said external tests were a better option.
"It is the pupils whose parents are going to come and see you in the evening," he said. "Their lives are going to be affected by the grade you write down.
"In the United States there have been examples of riots. Teachers have been forced to give the B grade because the riot police have told them to do it in order to calm down the situation."
Professor Tymms said teachers could be subconsciously biased according to factors such as a pupil's gender, ability, social class or behaviour. Even if they were aware of their prejudices, trying to compensate for them would not make their assessments reliable.
"If you know you have got that bias and you react against it, you might go too far in the other direction," he said at a Cambridge Assessment conference. "Teacher assessment is not the answer and we need to accept that."
His analysis is bad news for the teaching unions which have been campaigning for the end of national testing and advocating teacher assessment as a replacement.
At the same conference, Mary Bousted, general secretary at the teaching union ATL, said Sats needed to be replaced by "properly moderated" teacher assessment.
Professor Tymms gave an example of biased teacher assessment from when he worked at a secondary modern, originally teaching pupils who failed an 11-plus test.
When the test was replaced with teacher assessment he "noticed a clear difference in the kids that started coming to me".
"I started to get badly behaved, clever boys who (would have) previously passed the 11-plus," he said. "But when they shifted it they didn't pass what the teacher set. They were biased in the way they were assessed."
He said pupils tended to be assessed as "good" or "bad" in general terms. There was not enough differentiation.
Unions have used the fact that teacher assessment levels have been broadly in line with those shown by Sats as an argument in their favour.
But Professor Tymms said they could be measuring different things and that in the early 1990s there was "considerable" variation between Sats English results and what he said were "honest" teacher assessment levels. Since then, the two had converged. He told The TES that teachers had reacted to the criticism they faced for being too lenient or too harsh where their assessments were out of line with test results.
"Teachers have adapted their methods to the tests," he said. "They will privately have their own assessments of the children but they won't dare put them down."
Professor Tymms said he sympathised with unions' concerns about Sats but that it was league tables rather than tests that caused harm.
Professor Richard Daugherty of the Cardiff School of Social Sciences has been developing the teacher assessment system replacing Sats in Wales, where league tables have also been abandoned.
He told the conference that teacher assessment should not be used as a "high stakes" measure.
But Professor Daugherty believes it can overcome potential problems of bias in other circumstances, as long as more than one teacher is involved.