Schools could be racially discriminating against pupils when deciding who should take high-stakes tests, Government-commissioned research suggests.
A report from PricewaterhouseCoopers (PWC) reveals significant discrepancies by ethnicity in the proportion of pupils entered for experimental single-level "progress" tests (SLTs).
All pupils looked at had already been judged by teachers able enough or "eligible" for one of the tests, designed to measure pupils against a single, national curriculum level in reading, writing or maths.
But much lower proportions of those from black African and Pakistani backgrounds were then entered for SLTs compared with those from Indian and Bangladeshi backgrounds.
Gloria Hyatt, a consultant on teaching ethnic minority pupils, said teachers lacked the necessary training to make sound judgments about whether pupils with "ethnicities that are outside their experiences" should be entered for tests.
"Unfortunately the education teachers receive - through the media, literature and influenced by the UK's cultural norms - tends to generalise the diverse and complex cultures of ethnic minorities in the UK, embedding prejudiced behaviours and thoughts," she said.
"Even the most well read and culturally competent individual may find it difficult to go against such pervasive generalisations."
The latest figures echo, to some extent, those of a 2008 Warwick University study, which looked at the proportions of ethnic groups entered for the higher tiers of key stage 3 national tests.
It led academics to conclude there was "institutional racism" in schools because, other things being equal, for every three white British pupils entered for the higher tier tests, only two black Caribbean pupils were put in.
But the PWC study suggests bias could be operating against different ethnicities. It shows that only 20.5 per cent of eligible black African pupils were entered for the June 2009 SLTs, being piloted as possible alternatives to conventional Sats in hundreds of schools.
Black Caribbean pupils fared slightly better, with 24 per cent of those eligible entered, the study shows. But only 19 per cent of black pupils from other backgrounds were put in and just 22 per cent of eligible Pakistani pupils.
At the other end of the scale, 31 per cent of both Indian and Bangladeshi pupils eligible were entered. The proportion of eligible white British pupils entered was 23 per cent.
John Bangs, head of education at teaching union the NUT, said: "I think teachers sometimes make unwitting choices based on unwitting assumptions. It is an issue of training rather than conscious racism."
Results from the June 2009 maths SLTs could contribute to more than 200 schools' league table positions this summer.
Nearly 33 per cent of eligible Bangladeshi pupils were entered for a maths SLT, a higher proportion than any other group. But when they took the test they were outperformed by Indian, black African, Chinese and white British pupils - all groups with a lower chance of being entered.
20.5 - Percentage of eligible black African pupils entered for the June 2009 SLTs.